adaptation and clinical validation of a new handheld ... · iii conclusions: autofluorescence alone...
TRANSCRIPT
Adaptation and Clinical Validation of a New Handheld Optical Imaging Device (PRODIGI™) and Workflow for Real-Time
Intra-Operative Margin Assessment in Breast Cancer
by
Jenny Wang
A thesis submitted in conformity with the requirements for the degree of Master of Health Science
IBBME University of Toronto
© Copyright by Jenny J. Wang 2012
ii
Abstract
Title: Adaptation and Clinical Validation of a New Handheld Optical Imaging Device
(PRODIGI™) and Workflow for Real-Time Intra-Operative Margin Assessment in
Breast Cancer
Degree: M. HSc
Year of Convocation: 2012
Name: Jenny Wang
Department: IBBME
University: University of Toronto
Background: We report here early attempts of adapting a prototype fluorescence
imaging system (PRODIGI™) to be used as a surgical guidance tool to improve margin-
detection in breast cancer.
Methods: 36 patients were recruited to study the autofluorescence characteristics of ex
vivo specimens. 5-ALA (20 mg/kg) was used as a contrast agent in human breast cancer
cell lines and xenograft tumour models to detect PpIX fluorescence.
Results: Administrative approvals were obtained and a surgical drape was used for
sterilization. PRODIGITM could differentiate between normal and tumour tissues based
on autofluorescence alone in ex vivo samples. PpIX signal was detected in experimental
mice, and absent in control mice. The threshold of detection was on the order of 10 nM.
iii
Conclusions: Autofluorescence alone with PRODIGI™ was not sufficient for margin
assessment of ex vivo breast tumour surgical specimens. 5-ALA at an optimal dosage
may be adopted as a contrast agent to enhance tumour signal.
iv
Acknowledgements
I would like to thank my supervisor, Dr Ralph DaCosta and my co-supervisor, Dr
Brian Wilson for their guidance and suggestions throughout my thesis work.
I would like to thank my committee members, Dr Michael Rauth, Dr Wey Leong
and Dr David Jaffray for their knowledgeable inputs about the ongoing and future
direction of my thesis work.
I would like to thank Alexandra Easson, Jesse McMullen, Manoj Mathews, Viktor
Son, Susan Done, Dianne Chadwick and other members of the surgical pathology team
for their assistance in the clinical trial portion of my thesis work.
I would like to thank Jesse McMullen, James Bu and Azusa Maeda for their help
in the in vitro cell work and in vivo xenograft model imaging.
Lastly, I would like to thank my funding agencies, Cancer Care Ontario, Canadian
Institute of Health Research, the Canadian Institute for Photonic Innovations, University
of Toronto, and the George and Mary Turnbull Foundation for their generous support for
this important research.
v
Table of Contents
1. Breast Cancer and Surgical Treatment ..................................................................... 1 1.1. Introduction ....................................................................................................... 1
1.1.1. Basic Breast Anatomy ................................................................................ 1 1.1.2. Cancer Types and Stages ............................................................................ 3
1.1.3. Mainstream Treatment ............................................................................... 5 1.2. Best Practice ..................................................................................................... 9
1.2.1. Touch Prep ................................................................................................. 9 1.2.2. Frozen Section ......................................................................................... 10
1.3. Current Clinical Need ...................................................................................... 11 2. Current Imaging Approaches Applied to Breast Cancer Surgery ............................ 12
2.1. Treatment Planning and Margin Assessment ................................................... 12 2.1.1. Specimen Radiograph .............................................................................. 13
2.1.2. Intra-operative Ultrasound ........................................................................ 14 2.1.3. Optical Coherence Tomography ............................................................... 15
2.1.4. SpectroPen ............................................................................................... 16 2.1.5. Near-Infrared Fluorescence ...................................................................... 17
2.1.6. Point Fluorescence Spectroscopy ............................................................. 17 2.1.7. Problems with Existing Technologies ....................................................... 18
3. Objectives and Aims .............................................................................................. 19 3.1. Introduction to PRODIGI™ Platform .............................................................. 19
3.2. Principle of Operation ..................................................................................... 21 3.3. Application of PRODIGI™ in Breast Cancer Surgical Guidance – Related Clinical Engineering Issues........................................................................................ 22 3.4. Objectives ....................................................................................................... 26
3.5. Specific Aims .................................................................................................. 27 4. Aim 1 – Obtaining Administrative Approval for Intra-Operative Use .................... 27
4.1. REB Approval ................................................................................................. 27 4.2. Equipment Sterilization ................................................................................... 31
4.2.1. Available Sterilization Methods ............................................................... 31 4.2.2. Interference of Sterilization Process with PRODIGI™ Imaging ............... 33
4.3. Equipment Safety ............................................................................................ 36 4.3.1. Thermal Analysis ..................................................................................... 37
4.4. Conclusion ...................................................................................................... 39 5. Aim 2 – Determining Device Sensitivity ................................................................ 40
5.1. Introduction ..................................................................................................... 40 5.2. Apparatus and Methods ................................................................................... 40
5.2.1. Devices .................................................................................................... 40 5.2.2. PpIX Solution .......................................................................................... 41
5.2.3. Image Analysis......................................................................................... 42 5.3. Results ............................................................................................................ 42
5.4. Discussion ....................................................................................................... 45 6. Aim 3 – Evaluating Existing Filter Settings on Breast Cancer Samples Using AF Alone ............................................................................................................................ 47
vi
6.1. Autofluorescence ............................................................................................ 47 6.2. Materials and Methods .................................................................................... 49
6.2.1. Patient Sample Preparation ....................................................................... 49 6.2.2. Autofluorescence Imaging with PRODIGI™............................................ 50
6.2.3. Data Consolidation and Documentation.................................................... 51 6.2.4. Tumour Fluorescence Analysis ................................................................ 51
6.2.5. Point Spectroscopy with OceanOptics Spectrometer ................................. 52 6.2.6. Punch Biopsy ........................................................................................... 53
6.3. Results ............................................................................................................ 53 6.4. Discussion and Conclusion .............................................................................. 59
7. Aim 4 – Feasibility Testing of Using a Contrast Agent (5-ALA) ............................ 63 7.1. Contrast Agents and Image Probes .................................................................. 63
7.1.1. Image Probes............................................................................................ 63 7.1.2. Fluorescent Antibody Imaging Agent ....................................................... 64
7.1.3. Contrast Agent 5-ALA ............................................................................. 65 7.2. Clinical Implementation of 5-ALA at UHN ..................................................... 67
7.2.1. ALA and Clinical Interference – FISH Analysis ....................................... 67 7.3. Institutional and Governmental Approval ........................................................ 70
7.4. In Vitro Feasibility Test ................................................................................... 71 7.4.1. Materials and Methods ............................................................................. 72
7.4.2. Results ..................................................................................................... 75 7.4.3. Discussion and Conclusion ....................................................................... 81
7.5. In Vivo Feasibility Trial ................................................................................... 82 7.5.1. Materials and Methods ............................................................................. 82
7.5.1.1. Xenograft Model .................................................................................. 83 7.5.1.2. Maestro Imaging ................................................................................... 83
7.5.1.3. PRODIGI™ Imaging ............................................................................ 84 7.5.1.4. AxioObserver Imaging ......................................................................... 85
7.5.1.5. Point Spectroscopy ............................................................................... 85 7.5.1.6. Data Analysis ....................................................................................... 86
7.5.2. Results ..................................................................................................... 87 7.5.3. Discussion and Conclusion ....................................................................... 99
8. Summary and Future Work .................................................................................. 102 8.1. Summary of Results ...................................................................................... 102
8.2. Contribution to the Field ............................................................................... 105 8.3. Future Work .................................................................................................. 105
9. References ........................................................................................................... 108
vii
List of Figures, Tables and Equations
Figure 1 – Anatomy of a typical human breast. 1. Chest Wall. 2. Pectoralis Muscles. 3. Lobules. 4. Nipple Surface. 5. Areola. 6. Laciferous Duct. 7. Fatty Tissue. 8. Skin. Reproduced from [3] (with copyright permission). .......................................................... 2 Figure 2 – A PRODIGI™ prototype showing the back view (with the camera screen (A)). The protruding LED (B) with supporting heat sinks and fans are visible above and on the right side of the main camera body. ..................................................................... 20 Figure 3 – PRODIGI™ (A) and associated accessories in a carrying case. PRODIGI™ can be combined with an endoscope (B) and different filter sets for applications in colon cancer, oral cancer, chronic wound care and additional accessories can be developed for new applications. The package also includes additional batteries (C), a charger (not shown in image) and a USB connection (D) for the commercial camera. ....................... 20 Figure 4 – Comparison between undraped WL (a) and FL (b) images and draped WL (c) and FL (d) images taken by the prototype device. Normally, WL images were taken without the filter in front of the lens. However c) was taken with the filter and resulted in the red tint in the image. Otherwise, c) and d) showed a slight blurring artifact compared to a) and b), with no visible colour-shift. ....................................................................... 34 Figure 5 – Spatial resolution comparison between images taken under a) White light, without filter, undraped. b) white light, with filter, undraped. c) white light, without filter, draped. d) white light, with filter, draped. A ~10% resolution loss occurred with the addition of the camera filter, and a ~50% resolution loss occurred with draping. ........... 35 Figure 6 – Points measured in drape safety test. Point 1 and 3 were the heat sinks attached to the LEDs. Point 2 and 4 were the filters directly in front of the LEDs. Point 5 was the filter directly in front of the camera lens and was the point that would be closest to the patient during actual use. ..................................................................................... 38 Figure 7 – Results from the drape safety test. Each point was measured using an infrared thermometer at the following time points: every minute between 0-15 minutes, and every 10 minutes between 20-60 minutes. All points were below 70 degrees Celsius, which was 50 degrees below the Health Canada safety limit of 120 degrees Celsius, and was considered safe. The decrease in temperature as observed at time points beyond 40 minutes may have resulted from increased heat transfer between the drape and the surrounding due to a greater temperature gradient created by the elevated temperature within the drape. ............................................................................................................ 39 Figure 8 – Fluorescence signal intensity of PpIX solution at various concentrations as detected by the prototype device. A detection limit of 2.4 nM was observed, and a display limit of 625 nM was measured, above which the pixels in the red channel were saturated
viii
and were no longer representative of the PpIX concentration in the solution. Within these limits, there was an unexpected jump in FL intensity for concentrations below 4.9 nM. Each error bar indicates one standard deviation (N=3). .................................................. 43 Figure 9 – A re-plot of the truncated data collected in figure 34. The saturation region and concentrations below 4.9 nM were removed. A logarithm fit was introduced to the plot with a coefficient of correlation of 0.98. Each error bar indicates one standard deviation (N=3). ............................................................................................................ 44 Figure 10 – Point spectroscopy data collected from PpIX solutions at five different concentrations. Each data point was the average fluorescence intensity of 5 points between 635.14 nm and 635.93 nm. The six concentrations used were 0.12 mM to 12 nM, diluted by a factor of 10 in each subsequent dilution. A linear fit introduced to the data had an R2 value of 0.99. Each error bar indicated one standard deviation (N=5)............. 45 Figure 11 – Absorption spectrum of methylene blue. Absorption was lowest at around 405 nm, the excitation wavelength used by our device. Thus it did not attenuate any signal by absorbing the excitation light. Reproduced from [59]...................................... 55 Figure 12 – An example of WL and FL human breast cancer image taken by PRODIGI™. Under the WL image, tissue components could not be distingushed, and the tumours (identified by palpation, circled in red) did not appear any different from their surrounding tissue. Under FL, an array of components can be seen based on their unique AF signatures. In this image, green FL was associated with connective tissue, pale pink FL was associated with adipose tissue, and tumours (identified by palpation, circled in red) appeared darker than their immediate surrounding tissues. ..................................... 56 Figure 13 – Signal-to-background ratio of every ex vivo sample analyzed. Tumour signals were normalized to background in each image. A positive ratio was classified as “bright”, a negative ratio was classified as “dark”. Out of 29 patients, 17 patients were classified as “bright”, 12 were classified as “dark”. ....................................................... 57 Figure 14 – Tumour breakdown by its pathological diagnosis. 4 different types of tumours were present amongst all the patient samples - invasive lobular, invasive intraductal, adenoid cystic, invasive ductal. There was no statistical difference between “bright” and “dark” tumour under each type, suggesting that tumour signal as observed under AF was not indicative of its underlying pathology. .............................................. 57 Figure 15 – Point spectroscopy for patient #35 showing the peak emission wavelength for each core. T1, T2, N1, N2 correspond to the 4 cores in punch biopsy. T1 and T2 with 100% and 95% connective tissue had lower peak wavelengths at 499 nm and 500 nm, respectively. N1 and N2 had higher peak wavelengths at 508 nm and 524 nm, respectively. .................................................................................................................. 59 Figure 16 – FISH images for each respective cell line. The number of antibody stains remained constant between the 5-ALA+ sample and the 5-ALA- sample. MCF7/ALA-
ix
image was re-processed by the technician who performed FISH analysis because this image did not have enough contrast for unknown reasons. However, the number of antibody stains in MCF7/ALA- image remained constant before and after processing. .. 69 Figure 17 – SK-BR-3 imaging using AxioObserver (40x magnification). Autofluorescence channel was used to calibrate between different images. The levels of FL intensity in AF were normalized across different image sets, and the corresponding, adjusted PpIX fluorescence were compared. Cells incubated with 5-ALA produced bright red fluorescence due to 5-ALA-induced PpIX. Control cells without 5-ALA did not produce any fluorescence in the PpIX channel, suggesting that 5-ALA could increase breast tumor-to-normal fluorescence contrast in vivo. 5-ALA-induced PpIX fluorescence can be detected using the existing filter settings on the prototype device. ....................... 75 Figure 18 –MCF7 imaging using AxioObserver (40x magnification). Similar to Figure 14, AF channel was used to calibrate between different images. In this cell line, 5-ALA-induced-PpIX fluorescence was only visible in incubated cells, and was absent in control cells. The WL image for 8 hour timepoint was lost due to image corruption during saving. ........................................................................................................................... 76 Figure 19 – Quantitative comparison of average PpIX FL intensity between MCF7 cells and SK-BR-3 cells at various time points. Both control cells had comparable fluorescence intensity. SK-BR-3 had a higher FL intensity than MCF7 cells at 8 hours, and SK-BR-3 had a higher FL intensity at 6 hours than at 8 hours. Each error bar indicates one standard deviation (N = 5). .......................................................................................................... 77 Figure 20 – SK-BR-3 breast cancer cells incubated with 5-ALA produced PpIX fluorescence at both 3 hours and 6 hours. 5-ALA- cells did not produce any PpIX fluorescence. The AF channel was used similarly to trial 1, as a calibration channel to compare the 5-ALA intensity across different images. (40x magnification) ................... 78 Figure 21 – MCF7 breast cancer cells incubated with 5-ALA produced PpIX fluorescence at both 3 hours and 6 hours. 5-ALA- cells did not produce any PpIX fluorescence. The AF channel was again used as a calibration. (40x magnification)....... 79 Figure 22 – MDA-MB-231-H2N cells incubated with 5-ALA displayed an increase in PpIX fluorescence over time from 1 hour to 6 hour post-incubation. PpIX fluorescence signal could be observed starting at 1h post-incubation. (40x magnification) ................. 80 Figure 23 – Normalized fluorescence intensity over time in MDA-MB-231-H2N. There was a general increase in fluorescence over the period of 6 hours when the 5-ALA-induced-PpIX fluorescence was normalized against the corresponding autofluorescence intensity. A linear fit had an R-square value of 0.75 (75% “fit” to the data). Each error bar indicates one standard deviation (N=3). ......................................................................... 81 Figure 24 – Maestro images of a) an 5-ALA-injected mouse and b) a control mouse. White light images are at the top, and their corresponding fluorescence images are at the
x
bottom. Mouse was facing down, with the tumour on its right flank (indicated by an arrow in WL image). In the 5-ALA-injected mouse, there was a visible increase in FL over time, with a more dramatic increase in the liver (indicated by an arrow in FL image). In the control mouse, the whole body FL signal was similar at all time points. Liver had a higher signal than surrounding skin (indicated by arrows in FL images). ....................... 88 Figure 25 – Whole body mouse imaging with Maestro. 5-ALA-injected mice displayed a general increase in FL intensity over time, while control mice showed no change in FL intensity over time. Each error bar indicates one standard error (N=3). .......................... 88 Figure 26 – Whole body mouse imaging with PRODIGI™ of a) an 5-ALA-injected mouse and b) a control mouse. Tumour (indicated by an arrow in WL) was not visibly different from skin AF in the FL channel. There was no visible increase in FL over time in either a) or b). This either indicated that 5-ALA-induced-PpIX mechanism did not work in in vivo xenograft model, or that the signal was too weak to be detected by PRODIGI™ or by eye. .................................................................................................. 89 Figure 27 – Tumours from control and injected mice. In control mouse there was a faint AF signal throughout the tumour; in 5-ALA-injected mouse #1 (ALA+ #1), there was a spot of bright FL (indicated by arrow) on the surface of the tumour. However, when both tumours were sliced, there was no visible difference in FL signal between control and injected tumours. ........................................................................................................... 90 Figure 28 – Tumour (without skin) and tumour bed imaging using PRODIGI™. Tumour location is circled in blue in FL channel. In both cases, there was no visible signal from the tumour that was different from the surrounding muscle signal in either injected or control mouse. Skin AF was much brighter than either tumour FL or muscle AF. .......... 91 Figure 29 – Organs from control (left) and injected (right) mice. There was no visible difference in FL signal between the two. In both cases, kidney and liver had the strongest singles, while the remaining samples had no visible signals. .......................................... 92 Figure 30 – Quantitative analysis of organs and muscle FL signal taken by PRODIGI™. In all cases, the controls had a stronger signal, although both were weak. Kidney and liver had significantly higher signals than heart, spleen and muscle. This was in agreement with the observations made directly from the images. Values on the y-axis indicate the average pixel intensity. Each error bar indicates one standard deviation. ........................ 92 Figure 31 – Point spectra of xenograft tumours. In both curves, the max peak was at around 500 nm. Control tumour had a secondary peak at 600 nm. 5-ALA-injected tumour had a secondary peak at 635 nm (arrow), which was in agreement with PpIX emission peak. This indicated that if the signal intensity reaches the detection threshold of the camera in our prototype device, PpIX signal could be picked up by the camera in its current filter settings. Peak amplitudes were normalized. ............................................... 93
xi
Figure 32 – PRODIGI™ images of 5-ALA-injected (top) and control (bottom) mice. Tumour was circled in red. There was visible skin red FL in the injected mouse, and no such signals in the control mouse. However, there was no visible FL in the tumour in either case. This indicated that while PpIX was being produced in mice, it was not produced specifically in tumours. .................................................................................. 94 Figure 33 – Tumour cross section fluorescence from all 5 mice in trial 2. All 3 5-ALA-injected mice had a small secondary peak at around 635 nm (arrow), coinciding with PpIX emission peak. This peak was absent in both control mice. This indicated that 5-ALA-induced-PpIX was a viable mechanism in xenograft model, producing a detectable signal (by point spectroscopy). ...................................................................................... 95 Figure 34 – PpIX signal in various components in a single injected mouse. All components showed a peak at 635 nm (red dashed line), with varying intensity. This suggested that PpIX was present throughout the mouse, but at varying quantities. ......... 96 Figure 35 – AxioObserver images for a) an 5-ALA-injected window chamber mouse and b) a control window chamber mouse. Both tumours showed a similar GFP signal. Signal was detected in the PpIX channel in control tumour as well as in 5-ALA-injected tumour pre-injection, indicating a flood of GFP signal in the PpIX channel. Under this constant background signal, PpIX signal was not detected in the injected tumour over time. The FL intensity in injected tumour remained relatively constant visually. ........................... 98 Figure 36 – PpIX FL signal comparison in all 3 mice as measured by ImageJ. This analysis aimed to visualize any difference in trend between control and injected mice. All data points from the same mouse were normalized against pre-injection intensity. Between mice, the two 5-ALA-injected mice points were normalized against control to give a common starting point. The general trends for all 3 lines were consistent, indicating that the signals were similar between control and injected mice. If injected mice showed an increase in FL intensity over time, the normalized graph would display a positive slope much greater than one from the control mouse. ....................................... 98
Table 1 – Numerical measures from resolution test with drape. Between images with and without filter, a 10% loss in spatial resolution was observed. Draping added another 50% loss in resolution. .......................................................................................................... 35 Table 2 – Histopathology analysis of the biopsy cores obtained from patient # 35 indicated that the peak emission wavelengths as measured by point spectroscopy were indicative of the % of connective and adipose tissue in each core. Higher % of connective tissue corresponded to a lower peak emission wavelength (closer to 475 nm, the AF wavelength of collagen). T1 – tumour core 1, T2 – tumour core 2. N1 – normal core 1 (green). N2 – normal core 2 (red). ................................................................................. 58
xii
Abbreviations
AF – Autofluorescence
ALA/5-ALA – Aminolevulinic Acid/5-Aminolevulinic Acid
ARC – Animal Resources Centre
AUP – Animal Use Protocol
Bcl2 – B-Cell Lymphoma 2
BCS – Breast Conservation Surgery
BP – Band Pass
Bs – Background Signal
BT – Background Threshold
CCD – Charge-Coupled Device
CPD – Central Processes Department
CT – Computer Tomography
CTA – Clinical Trial Application
DCIS – Ductal Carcinoma in Situ
DMEM – Dulbecco's Modified Eagle Medium
DMSO – Dimethyl Sulfoxide
EGFR – Epidermal Growth Factor Receptor
ER – Estrogen Receptor
FBS – Fetal Bovine Serum
FISH – Fluorescence In-Situ Hybridization
FL – Fluorescence
GFP – Green Fluorescent Protein
xiii
H&E - Hematoxylin and Eosin
HER2 – Human Epidermal Growth Factor Receptor 2
IBBME – Institute of Biomaterials and Biomedical Engineering
IgH – Immunoglobulin Heavy Chain
LCIS – Lobular Carcinoma in Situ
LED – Light-Emitting Diode
LP – Long Pass
MOSFET – Metal-Oxide-Semiconductor Field Effect Transistor
NIR – Near-Infrared
PA – Pathology Assistant
PpIX – Protoporphyrin IX
PRODIGI™ - Portable Real-time Optical Detection, Identification and Guide for
Interventions
REB – Research Ethics Board
RPMI – Roswell Park Memorial Institute (the institute that developed this media)
TS – Tumour Signal
UHN – University Health Network
USAF – United States Air Force
WL – White Light
xiv
Publications
Publications Appearing in this Thesis
• Poster Presentation – Canadian Cancer Research Conference
o Clinical Validation of PRODIGI™ as a Real-Time Intra-Operative Margin
Assessment Tool
o Toronto, Canada
o November, 2011
• Poster Presentation – Image Network Ontario – Third Place Winner
o Clinical Validation of PRODIGI™ as a Real-Time Intra-Operative Margin
Assessment Tool
o Toronto, Canada
o February, 2012
• Poster Presentation – Photonics North 2012
o Clinical Validation of PRODIGI™ as a Real-Time Intra-Operative Margin
Assessment Tool
o Montreal, Canada
o June, 2012
xv
Thesis Preamble
This is a Master’s thesis for clinical engineering program at IBBME. The clinical
engineer program consists of 1 year of course work and 1 year of thesis work. At the end
of the program, the student is expected to produce a full-length thesis paper and to defend
his or her thesis.
This thesis project began in January 2011 and ended July 2012. It became the
pilot for a series of projects which aimed to adapt a previously conceived prototype
wound imaging device to be used to visualize breast cancer cells in vivo. The main focus
of the project was to ascertain the clinical engineering issues involved in modifying the
original imaging device. The necessary involvement of biology in the process was
primarily used to illustrate the technical issues associated with adapting the device.
- 1 -
1. Breast Cancer and Surgical Treatment
1.1. Introduction
Breast cancer is the most common cancer among women worldwide with over
20,000 cases diagnosed in Canada every year. On average, 1 in 9 women is expected to
develop breast cancer in her lifetime and one in 29 is expected to die from it [1]. Breast
cancer is a clinically heterogeneous disease with variable origins and outcomes.
1.1.1. Basic Breast Anatomy
The human breast is composed of adipose tissue and glandular, milk-producing
lobules, supported by a network of connective tissue and blood vessels [2]. (Figure 1[3])
- 2 -
Figure 1 – Anatomy of a typical human breast. 1. Chest Wall. 2. Pectoralis Muscles. 3. Lobules. 4. Nipple Surface. 5. Areola. 6. Laciferous Duct. 7. Fatty Tissue. 8. Skin. Reproduced from [3] (with copyright permission).
Glandular tissue exists in the form lobes and lobules [4]. Up to 20 lobes can be
found within each breast, each further divided into lobules where multiple milk-
producing glands reside. These milk-producing alveoli are connected through small
ducts, which then join to form larger milk ducts called lactiferous ducts and exit through
- 3 -
the nipple. Adipose tissue is divided into subcutaneous fat, intraglandular fat and
retromammary fat based on its location: subcutaneous fat is the thin layer of fat
immediately underneath the anterior surface skin; intraglandular fat is the fatty tissue that
surrounds milk glands; retromammary fat is the fat positioned at the back of the breast
next to the chest wall. Percentage of fat within the breast varies between individuals, but
generally increases over time as granular tissue disappears in the course of aging [5].
Connective tissue consists of mainly collagen, but also some elastin. The base of
the breast is aligned with suspensory ligaments called Cooper’s ligaments. Connective
tissue determines the firmness of the breast and provides support, particularly around
milk ducts. Over the course of aging, the percentage of connective tissue in the breast
increases as the breast becomes more fibrous. Literature shows that the arrangement of
connective tissue can be indicative of breast cancer [6]. Specifically, a network of
connective fibres perpendicular to the tumours is indicative of a poor patient prognosis.
1.1.2. Cancer Types and Stages
The two most common sites of origin of breast cancer are duct and lobule. A
ductal carcinoma is a type of tumor that develops from the walls of milk ducts. It usually
starts as a non-invasive cancer called ductal carcinoma in situ (DCIS), but may invade
duct walls and proceed to metastasize into surrounding or distant tissues. A lobule
carcinoma is a type of tumor that develops from with lobules where milk is produced.
The non-invasive form of this tumor is called lobule carcinoma in situ (LCIS). In some
- 4 -
cases, LCIS is not considered a true cancer, but an indication of increased risk in
developing breast cancer [7].
In clinical diagnosis, tumours are divided into two subtypes: non-invasive, which
are confined within ducts or lobules in the breast and do not invade normal tissues, and
invasive, which have spread into normal tissues [8]. The two major non-invasive
subtypes are DCIS and LCIS. The two most common invasive cancers are invasive ductal
carcinoma and invasive lobular carcinoma. In addition to histological categorization,
breast cancer can be further divided according to its biomarker expression. Biomarkers
typically tested in immunohistochemistry include estrogen receptor (ER), progesterone
receptor (PgR), human epidermal growth factor receptor 2 (HER2) and epidermal growth
factor receptor (EGFR). Hormone-receptor-positive (ER/PgR+) tumours are divided into
two subgroups: luminal A (HER2-) and luminal B (HER2+), although most luminal B are
typically ER+ with high proliferation. Hormone-receptor-negative (ER/PgR-) tumours
are divided into three subgroups: HER2 over-expressing (HER2+), basal-like (HER2-,
EGFR+) and triple negative (HER2-, EGFR-).Different combinations of biomarker
expression are indicative of different prognosis. In general, ER/PgR+ tumours have better
prognosis than ER/PgR- tumours, with triple negative tumours having the worst
prognosis. Prognosis is also dependent on the stage of the tumour [9]. Cancer stage
describes the size of the tumour and its tendency to metastasize. In Canada, breast cancer
is divided into five stages. Stage 0 cancer includes non-invasive tumours, namely DCIS
and LCIS. Stage 1 cancer includes invasive cancers that are 2cm or smaller, and has not
spread outside of the breast. Stage 2 cancer includes tumours 2-5cm in size, and/or
- 5 -
minimal lymph node involvement. Stage 3 cancer includes tumours greater than 5cm in
the largest dimension, and/or with extensive lymph node involvement. Stage 4 cancer
indicates primary tumours with distant metastasis, most commonly in bones [10][11], but
also in lungs, liver [12], the brain [13] or even intestines [14].
Cancer survival rate varies based on tumour stage, type and treatment. However,
the 5-year survival rate is generally higher when cancer is detected during an earlier stage
(0-1) [15]. This correlation leads to the implementation of screening mammograms for
women over 50 years of age. Screening for breast cancer attempts to detect tumours
earlier than stage 1, when they are smaller than 1cm and have no adjacent or distant
metastasis [16], and such tumours are usually detected using mammography by imaging
for non-palpable microcalcifications which appear as dense micromasses the image.
However only 35-45% of such tumours can be detected [17]. A contributing factor to
false-negative results in screening mammogram is breast density [18], which decreases
with age and puts elderly women at an increasing risk for false-negative results.
Incidentally, breast density is an independent risk factor for breast cancer development
[19][20]. To address this concern and improve detection rate, current research encourages
the addition of ultrasound (to MRI) imaging for women with mammographically dense
breast to improve the detection of microcalcifications in asymptomatic women with this
type of breast tissue [16].
1.1.3. Mainstream Treatment
- 6 -
Once a tumour has been identified and confirmed cancerous, several different
pathways for treatment may be considered. For early stages (stage 1-2), surgical
intervention is the mainstay of the treatments. Small tumours tend to be removed through
a breast conservation surgery (BCS, or lumpectomy) while maximally sparing
surrounding healthy tissue; larger tumours are removed through a mastectomy where the
affected breast is entirely removed. A diagnostic ultrasound and mammography are
conducted prior to an operation in order to determine the tumour size and its anatomical
relationship with the skin, nipple, and chest wall [21]. Axillary lymph nodes are also
screened for possible involvement, especially for invasive breast cancer [22]. Since it is
difficult to directly distinguish tumour cells from normal cells under white light
environment, the surgical margins are assessed using standard histopathology analysis as
a surrogate for the completeness of tumour removal. A negative margin indicates that no
tumour cells are found on the surface of the lumpectomy specimen. A positive margin
indicates tumour cells found on the surface of the specimen, and requires the patient to
undergo more surgeries until a negative margin is achieved. The rate of re-operation for
positive margins is between 15% - 40% for breast cancer, with occasional reports
indicating a rate as high as 70% [23][24].
Patients may undergo further radiation and chemotherapy post-surgery to reduce
the chance of local recurrence. Radiation therapy delivers high-energy beams (such as x-
rays) at the suspected tumour area to damage the DNA of the cell and thereby prevent the
cell from further division. Whole-breast radiation therapy (50 Gy over 25 dosages [25]) is
- 7 -
recommended following an early-stage lumpectomy procedure in an attempt to
completely destroy remaining individual cancer cells. Radiation may also be
recommended following a mastectomy procedure to reduce the rate of recurrence by up
to 70% [26]. Chemotherapy uses intravenously or orally administrated drugs that target
rapidly-dividing cells [27]. As a side effect, other rapidly-dividing healthy cells in the
body (e.g., hair, bone marrow, intestine lining) may also be affected by the drugs.
Systemic chemotherapy is typically recommended to patients with early-stage breast
cancer following an operation (called adjuvant chemotherapy) to completely remove any
residual cancer cells. The difference between targeted radiotherapy and adjuvant
chemotherapy is that radiotherapy is localized to the breast and regional nodes and targets
cells that may develop at the site of the surgery. Chemotherapy targets cells that may
have metastasized from the primary site (breast) to other parts of the body. Chemotherapy
is also prescribed to patients with later-stage breast cancer to destroy cancer cells that
have already metastasized.
If there is any likelihood for the breast cancer to metastasize into adjacent tissues
at the time of the surgery, sentinel lymph nodes are removed in addition to the standard
lumpectomy or mastectomy procedure. In this case, a radiotracer (technetium-99m sulfa
colloid) and a blue dye (e.g. Patent Blue, Isosulfan blue) are injected into the breast
during the surgery to visually trace lymph node drainage, and those lymph nodes closest
to the tumour are removed to be histopathologically inspected for the presence of tumour
cells. Tumour cells found within sentinel lymph nodes are prognostically significant and
may lead to recommendation of additional treatments or operations.
- 8 -
For larger tumours that are not immediately operable, pre-operative systemic
treatments are prescribed to shrink tumours into an operable size as judged by the
surgeons. Pre-operative treatments are typically systemic chemotherapy where necessary.
Once the tumour shrinks to an operable size, a surgery is again recommended depending
on the size and location of the tumour.
When all other procedures have been completed, patients with hormone receptor
positive breast cancers may be prescribed an endocrine therapy [19]. The most common
receptor tested in breast cancer is estrogen receptor. Estrogen is the main female sex
hormone primarily produced in ovaries. It promotes the development of female secondary
sex characteristics, including breasts. In a cancerous environment, estrogen tends to
promote the growth of breast cancer. Patients with estrogen receptor (ER)-positive
tumours are prescribed aromatase inhibitors, selective estrogen receptor modulators
(SERMs) or estrogen-receptor downregulators (ERDs) to stop estrogen production (for
post-menopausal women) and/or reduce estrogen-cell binding. Common drugs in each
category include Arimidex (anastrozole), Aromasin (exemestane) and Femara (letrozole)
for aromatase inhibitors, tamoxifen, Evista (raloxifene) and Fareston (toremifene) for
SERM and Faslodex (fulvestrant) for ERD. The main purpose of estrogen-inhibiting
drugs is to reduce cancer cell division promoted by estrogen. Endocrine therapy is taken
orally usually for five years. In some occasions, hormone therapy may also be prescribed
before surgery [28].
- 9 -
Patients with human epidermal growth factor receptor 2 (HER2)-positive tumours
have the option to take HER2-inhibition drugs such as HerceptinTM (trastuzumab).
Herceptin™ is an antibody that specifically binds to HER2 and reduces substrate binding
by competitive binding inhibition. Herceptin™ exerts anti-proliferative effects on HER2+
breast cancers by blocking the activation of the HER2 pathway which accounts for the
aggressive behaviour of these cancers.
1.2. Best Practice
Breast cancer surgery margin assessment is typically done by the surgeon by
inspection and palpation of the surgical specimens during the surgery. For non-palpable
tumours, the surgical specimen is imaged by the radiologist using x-ray (mammogram) or
ultrasound. In some cases, Intra-operative examination is completed either through touch-
prep, frozen section, or some forms of imaging assessment.
1.2.1. Touch Prep
Immediately following the removal of the tumour, the specimen is sent to the
surgical pathology lab for quick analysis, which typically includes gross inspection in
conjunction with touch-preparation cytology [29]. In short, the suspected sample margin
is quickly pressed against a glass slide such that some sample cells become attached to
the glass slide. These cells are then fixed, stained and analyzed under standard
microscopy to determine the presence of cancer cells. The principle of touch preparation
- 10 -
is that cancer cells tend to stick to the glass slide, whereas normal adipose cells do not
stick. Alternatively, small resected samples can be fixed and sectioned for assessment of
tumour cells.
Once a result is obtained, it is communicated to the surgeons. If the margin is
determined to be negative, the patient is closed up and the surgery concludes. If the
margin is determined to be positive, additional tissue is removed on the positive side and
a second touch preparation is done on the new sample. This process is repeated until a
negative margin is achieved. During the entire process (20 minutes per iteration), the
surgical team as well as the patient are on standby in the operation room.
1.2.2. Frozen Section
Frozen section operates on the same principle as standard histopathology, but at a
much faster pace in exchange for a loss in the quality of the slides produced. A small
sample is quickly placed in gel (e.g. OCT gel, Tissue-Tek, Sakura Finetek USA) and
frozen at -20 °C until the gel becomes hardened. The gel block along with the tissue
frozen inside is sliced in the cryostat, and the resultant tissue slice is transferred into a
room-temperature microscope slide by touching the slide to the tissue slice. The fixed
slice is then stained (e.g. hematoxylin and eosin, H&E) for analysis [30]. This entire
process takes approximately 10 minutes.
- 11 -
Lastly, intra-operative mammography and/or ultrasound can be performed in a
separate radiology room to detect tumour margins. This process takes 20-30 minutes
depending on how long it takes to physically move the sample to the radiology room [31].
Following the surgery, the surgical specimen is processed by standard
histopathology. Entire samples are thoroughly inspected visually following breadloaf-
slicing. Tumours are located by palpation – that is, by touch of finger on the suspected
sample slice – and any suspicious areas are sliced and stored in a cassette for H&E
staining. Isolated samples (from cassettes) are fixed, sliced, stained and read by a trained
pathologist to determine the exact tissue composition in the sample. This process is
considered the gold-standard in margin-assessment. Any margins with cancer cells
identified from H&E is considered a positive margin, and the patient is commonly
recalled to have additional surgeries. A standard pathology report takes 1-2 weeks to
produce, after which positive-margin patients are re-assessed and prepared for additional
surgeries.
1.3. Current Clinical Need
In general, 15-40% of all surgical patients have will positive surgical margins
who will require additional operations [23][24][32]. Re-operation is associated with an
increase in cost, surgery wait-time, patient discomfort and potential for future
complications such as surgical wound infection. This high rate of reoperation is a clear
indication that our current best practices in tumour margin detection are inadequate.
- 12 -
Detection by simple visual inspection and palpation is highly inaccurate, and detection by
touch-prep cytology and tissue section analysis are time-consuming, often with a high
rate of false negatives. This is particularly true for tissue section analysis, which can have
inadequate margin assessment for large tumour samples and loss of diagnostic samples
for small tissue samples. Inking and slicing tissue also alters conformation of the resected
tissue, making specimen re-orientation difficult [33]. It becomes a challenge to remap a
positive margin to the correct location within the patient.
There is real need for an affordable margin-assessment tool that provides visual
feedback intra-operatively in real time. An ideal margin-assessment tool must:
• have sufficient tumour-specificity to clearly distinguish tumour from normal cells
in a breast tissue,
• have sufficient signal contrast between cancer and normal cells such that tumour
cells are readily visible (highly sensitive)
• provide real-time visual guidance such that cancer cells can be removed
immediately in the initial surgery.
• be affordable and accessible to surgeons, and available for commercial production
2. Current Imaging Approaches Applied to Breast Cancer
Surgery
2.1. Treatment Planning and Margin Assessment
- 13 -
There are several mature imaging modalities currently being used for treatment
planning and margin assessment, as well as some emerging technologies that attempt to
fill the gap particularly in margin assessment. As described in section 1.2, intra-operative
examination is completed either through touch-prep, quick frozen sections or intra-
operative optical assessment. Mature modalities used for intra-operative optical
assessment include specimen mammography (SM) and intra-operative ultrasound.
Emerging technologies such as Raman spectroscopy and near-infrared fluorescence are
designed to address the weakness of existing imaging modalities.
2.1.1. Specimen Radiograph
Specimen mammography (SM) is a quick mammography performed intra-
operatively to assess tumour margin. The imaging principle is identical to a typical
screening mammography – using low-energy X-rays to screen the sample for lesions or
calcifications. Lesions, which are generally denser than the surrounding tissue, appear
brighter on the image compared to normal breast tissue. For specimen mammography,
orientation is maintained using metallic clips, which appear white on the image [34]. If
SM returns with a close lesion, a partial lesion or no lesions at all, additional tissue may
be removed at the discretion of the surgeon.
There are four steps to the entire SM procedure: transferring the sample to the
radiology lab, acquiring an image, developing film, and reporting results. The entire
process may take up to 30 minutes, during which time the patient and the surgical team
- 14 -
are waiting in the operating room with the patient still under general aneasthesia. Aside
from the obvious time-consuming aspect of SM, it is also highly inaccurate, with a false-
positive or false-negative rate of up to 44% [35] and a sensitivity of 58-80% [36]. A
retrospective study has shown that only 1.8% of patients out of 165 clearly benefitted
from SM for image-guided surgeries [35]. This incurs an additional operation time of 45
minutes and an additional operation cost of over $16,000 for the remaining 98.2% of
patients with no real benefits. Although this is one of the most standard procedures
performed intra-operatively right now for margin assessment, its high cost, long
performance time and low accuracy makes it a poor determinant of margin status.
2.1.2. Intra-operative Ultrasound
Ultrasound is used in breast-cancer screening in conjunction with mammography
[36]. In most cases, ultrasound is used as a supplemental modality for women with dense
breast tissue where mammography becomes ineffective [37]. Literature has shown that
ultrasound cannot always visualize non-palpable lesions detected by mammography with
a delineation rate of 50-60% [38], suggesting that ultrasound is not an effective stand-
alone screening modality. It is however able to differentiate between solid and cystic
tumours. Intra-operative ultrasound is used to detect tumour margins similarly to
screening ultrasound. A sterile transducer is brought into the operating room and placed
on the operable breast. The breast tissue is oriented in a way such that the lesion is closest
to the skin surface, and is held in position during scanning. During the surgery,
ultrasound is repeatedly used to determine the status of the tumour margin.
- 15 -
Due to the physics of ultrasonography, images produced by ultrasound have poor
resolution. This translates to inaccurate representation of the tumour region, where the
predicted tumour region was larger or smaller than the actual tumour. Graham et al [39]
showed that from 56 ex vivo tumour specimens whose margins were determined to be
negative using radiographs, only 18 were histologically confirmed negative. This high
false-negative rate of result makes intra-operative ultrasound another poor candidate for
margin assessment.
2.1.3. Optical Coherence Tomography
Optical coherence tomography (OCT) is an imaging modality that collects
backscattered near-infrared light from tissue and produces an image based on the
coherence interferometry [40]. OCT is capable of a resolution on the scale of
micrometers with a tissue penetration of 1-2 mm. Typically, a series of 2D image “slices”
is acquired over a 3D sample, and the images are compiled in 3D modeling software to
reconstruct a 3D image of the original sample.
OCT is a speedy imaging modality that can acquire images in real time, but
requires additional image processing to yield useful information about the images.
Certain platforms are capable of video-imaging, providing visual information in real
time, but with very limited field of view (~2cm) [40]. OCT is also sensitive to motion
factors, which may become an issue in vivo in breast cancer surgeries since the field of
- 16 -
view is directly above the lungs. Currently OCT as an ex vivo or in vivo imaging modality
is still in pre-clinical stage, and is not routinely used in practice.
2.1.4. SpectroPen
SpectroPen (University of Pennsylvania, USA) is a newly developed handheld
margin-detection tool that works on the principle of Raman scattering. A sampling tip is
attached to a spectrometer that can record near-infrared (NIR) fluorescence and Raman
signals [41]. SpectroPen excites at 785 nm, collects signal from the same head, passes the
signal through a long-pass (LP) filter and analyzes it in the spectrometer. It has a minimal
resolvable concentration of 2-5 x 10-11 M for indocyanine green and 0.5-1 x 10-13 M for
surface-enhanced Raman scattering agent (e.g. nanoparticles). SpectroPen has a tissue-
penetration depth of 5-10 mm.
Despite its many appealing features, SpectroPen is a point margin detection tool
that does not provide visual feedback. A point margin detection tool is prone to
inadequately assessing large tumours due to human error, where certain points along the
margin are missed because it is unreasonable to expect the surgeon to perfectly sample
each and every point along the margin. Additionally, the lack of visual feedback makes
this task even more daunting. While this tool may have the sensitivity and tissue-
penetration as a margin assessment tool for specific points of interest, it is not ideal for
image-guided surgeries.
- 17 -
2.1.5. Near-Infrared Fluorescence
A particular pre-clinical technology based on near-infrared (NIR) fluorescence
visualizes tumour using 780 nm excitation light and captures emission light at 820 nm
[42]. In this application, NIR excites fluorophores that are pre-tagged onto the tumour
with antibodies. Captured fluorescence signal must first be processed on the computer,
then digitally overlaid on top of pre-captured tumour 3D image in order to achieve
tumour co-localization. Tumours are highlighted in fluorescence, and a negative margin
is achieved if no residual tumour cells are visible in the surgical bed. Real-time image-
guided surgery using NIR fluorescence has had limited application in humans due to the
lack of available clinically approved NIR fluorophores. Most current clinical trials use
indocyanine green (ICG), a clinically approved dye that is non-specific. In breast
applications, ICG is used in conjunction with methylene blue to identify lymph nodes /
sentinel lymph nodes. To move into trials where breast cancer cells can be visualized, a
dye that demonstrates specificity to breast cancer needs to be developed first.
2.1.6. Point Fluorescence Spectroscopy
Point fluorescence spectroscopy is a type of fibre-optic spectroscopy. A small
region on the tissue (~5mm in diameter) is excited using an excitation light through the
fibre, and the emission response is collected through the same fibre and passed through a
filter. The filter attempts to block all excitation signals, allowing only the emission signal
to be analyzed in a spectrometer. The results are displayed in commercial software on a
computer, and can be exported either as a graph or as a table of values.
- 18 -
Point spectroscopy is capable of collecting true emission response from a small
tissue area (5 mm in diameter) with little scattering interference. The field of view is very
limited, but the sensitivity within the field is high. In this project, point spectroscopy is
selected to be a complimentary technology that can provide quantitative emission profiles
of small areas of interest to supplement the wide-field fluorescence imaging performed
by the prototype imaging device discussed in the next chapter.
2.1.7. Problems with Existing Technologies
Technologies that are clinically practiced have two drawbacks: they are non-
specific and time-consuming. Specimen mammography has limited detection specificity
and requires up to 30 additional minutes to complete. Intra-operative ultrasound does not
have sufficient resolution to provide useful margin-detection capabilities, and need to be
repeatedly assessed during the surgery. OCT either requires additional image processing,
or provides very limited field of view in real-time applications. None of technologies
address all the clinical needs as outlined in section 1.3. Emerging technologies attempt to
address these shortcomings and the two described in this section have so far increased
detection sensitivity and specificity. However, both are still in pre-clinical stage (tested
on phantoms) and have their own flaws. SpectroPen is a point-detection tool. The lack of
an overall visual guide and the small sampling area makes this tool prone to incomplete
margin assessment due to human errors. This tool also requires tissue contact to function
properly, adding the need for sterility and the risk for infection during use. Near-infrared
fluorescence can provide a visual feedback after signal processing (not in real time). A
- 19 -
phantom depicting the breast morphology must be constructed using additional
technologies before the fluorescence signal can be overlaid onto the sampling area. It is
also currently unfeasible clinically because there is no available NIR dye specifically for
breast cancer cells.
3. Objectives and Aims
3.1. Introduction to PRODIGI™ Platform
The Portable Real-time Optical Detection, Identification and Guidance for
Intervention (PRODIGI™) is a prototype device developed by the DaCosta Laboratory at
the University Health Network (UHN). It is a handheld optical imaging platform that
meets the clinical needs outlined in section 1.3. This prototype device is capable of
imaging both biological and non-biological samples using white-light or multispectral
fluorescence. Different biological components (such as tumour cells) can be visualized by
tuning the filter settings on the device to the fluorescence signatures of the components.
Using a commercial charge coupled device (CCD) digital camera, fluorescence signals
can be relayed back to the user in real-time with sub-millimetre spatial resolution and 3-4
mm tissue penetration. Two to four high-power light-emitting diodes (LED) arrays (0.005
mW/cm2) are attached around the lens to provide excitation light source for fluorescence
imaging. The field of view is determined by the digital camera (approx 100 cm2) and
images can be taken without contacting the sample. Additional accessories such as rigid
or flexible endoscopes can be attached to the device construct to provide more
- 20 -
functionality. Figure 2 shows a typical PRODIGI™ unit and figure 3 shows the carrying
case along with the rigid endoscope accessory.
Figure 2 – A PRODIGI™ prototype showing the back view (with the camera screen (A)). The protruding LED (B) with supporting heat sinks and fans are visible above and on the right side of the main camera body.
Figure 3 – PRODIGI™ (A) and associated accessories in a carrying case. PRODIGI™ can be combined with an endoscope (B) and different filter sets for applications in colon cancer, oral cancer,
A
B
A
B
C
D
3cm
3cm
- 21 -
chronic wound care and additional accessories can be developed for new applications. The package also includes additional batteries (C), a charger (not shown in image) and a USB connection (D) for the commercial camera.
PRODIGI™ is a low-cost, handheld mobile prototype device that can be easily
deployed in the operating room for intra-operative use. It provides real-time visual
feedback through a commercial digital camera while providing images comparable to
commercial high-cost image devices, and its intuitive interface requires minimal training.
For PRODIGI™ to become a viable margin detection tool for breast cancer surgeries, it
needs to demonstrate tumour-specificity in its imaging capabilities.
3.2. Principle of Operation
This prototype device operates on a commercial CCD camera. Each pixel on the
acquisition screen is represented by a metal-oxide-semiconductor field-effect transistor
(MOSFET) capacitor which allows the conversion of incoming photons (light) into
electric charges proportional to the intensity of the incoming light. Each pixel is
converted into a corresponding voltage, and the entire image is represented by an array of
voltages that is analyzed digitally to reproduce the captured image.
In fluorescence imaging, LEDs on the device act as the light sources. A wide-field
excitation light is applied to the tissue surface, and a portion of the excitation signal is
attenuated as it travels down from the surface due to photon absorption or scattering
within the tissue. This attenuation is a function of the excitation wavelength, and a typical
- 22 -
penetration depth of 1-2 mm can be achieved with the excitation wavelength (405 nm)
used on the prototype device. The absorbed photons are capable of elevating electrons to
a higher energy state. As this excited electron returns to ground state, a photon matching
the energy difference between excited and ground states is emitted, producing an
emission fluorescence signal that can be captured by the CCD camera. Different tissue
components absorb and emit signals at different wavelengths and intensities, allowing for
the delineation of tissue components under the AF mode. Multispectral fluorescence
imaging can be achieved by adjusting the excitation and emission filter that is placed in
front of the camera lens. Depending on the target that we want to visualize, different filter
settings can be developed to collect signals specifically from our targets of interest,
including both AF signals and signals from contrast agents.
3.3. Application of PRODIGI™ in Breast Cancer Surgical Guidance –
Related Clinical Engineering Issues
Before the start of the breast cancer project, PRODIGI™, the prototype imaging
device, was primarily used in wound care trials. In wound care, PRODIGI™ was used to
fluorescently screen advanced chronic wounds for regions of bacterial populations (e. g.
Pseudomonas) invisible under white light. Chronic wounds are often prevalent in diabetic
patients, whose wounds take much longer to heal due to compromised immune system
and narrow arteries. This application used 405 nm excitation to visualize bacteria using
autofluorescence, where healthy, clean skin appeared green (signal emitted by connective
tissues such as collagen and elastin) and bacteria population appeared red (600 nm). By
- 23 -
adopting PRODIGI™ screening, the accuracy of bacterial detection increased from 35%
to 74%. This increase in accuracy led to the brainstorming for other applications with
which this device might be compatible.
Breast cancer was identified as a field that would benefit from additional visual
information in real time. Breast tissue is known to be heterogeneous in nature. Although
the structure of a human breast is relatively defined, the tissue composition changes over
the lifespan of the patient towards a more fibrous nature. Additionally, in breast cancer,
the tumour is often similar in appearance to its surrounding benign tissue when viewed
under white light. Our goal is to tailor the excitation and emission wavelength settings on
the prototype device from a wound-specific application to a breast-specific application.
Under ideal settings, our device would be able to differentiate between different local
tissue components (adipose, connective, ducts) and assign a unique signal to tumours.
This unique signal could be used to identify tumours in an unknown sample in real time,
and guide the surgeon to the complete removal all tumour cells in a breast cancer surgery.
Aside from the technical aspects of the device, human factors may also have to be
adjusted for breast cancer application. In wound care, a nurse is able to hold the camera
in one hand and swab the bacterial region with the other hand; this is unlikely to be
feasible in breast cancer surgeries, where the surgeon is expected to perform his tasks
with both hands. A hands-free design would be ideal in this situation to eliminate the
need for additional personnel in the operating room solely to operate on the camera. This
could be achieved either by mounting the entire system on a surgical arm that can be
- 24 -
swung in and out of the surgical field, or by mounting the system on a head-mount and
projecting the visual feedback onto the surgical field. Both designs may have to be tested
in practice to determine the more user-friendly design.
Sterilization is also an issue that must be solved for the translation of application
from wound care to breast cancer. In the wound care application, the device operates
under a non-sterile environment. In the breast cancer applications, the device is expected
to operate within the sterile surgical field and is expected to be sterilized to institutional
standards. There are several approved methods of sterilization used by medical
institutions and each method needs to be explored for compatibility with our device.
More details about sterilization will be covered in the next chapter.
However, before any experiments can be conducted, the device must first pass
regulatory approvals. There are several governing bodies associated with UHN: Health
Canada, Institutional Research Ethics Board (REB), and any procedural departments (in
the case of University Health Network, the procedural department is the Central Process
Department, CPD). The first step is to demonstrate that the device complies with Health
Canada’s medical device regulations. To obtain approval from Health Canada, an
application has to be filed to the Medical Devices department, sometimes as well as the
Canadian Nuclear Safety Commission [43]. In the approval process, the device is
considered for the risks it may impose on the patient based on its usage, and is classified
according to its risks [43]. The manufacturer is then required to provide documentation of
standard procedures regarding the following areas:
- 25 -
• Handling, storage and delivery
• Installation
• Corrective action
• Servicing
The device is also required to meet the safety and effectiveness requirements
listed in Health Canada regulations. These requirements state that: all risks relating to the
device must be identified. Out of these, any risks that could be eliminated must be
eliminated; risks that cannot be eliminated should be reduced to the extent possible, and
information needs to be provided about remaining risks and safety procedures (such as
alarms) [44]. When all related certificates and fact sheets are prepared, the manufacturer
can submit an application to Health Canada for a license. The application review process
can take up to 90 days depending on the classification of the device. PRODIGI™ is
already approved by Health Canada as a class II medical device for wound care
application, and does not require re-approval for breast cancer application.
Following device approval, the actual clinical trial must also be approved. This
involves submitting a Clinical Trial Application (CTA) to Health Canada outlining the
phase of the trial, the protocol of the trial and the devices and drugs involved in the trial.
In my clinical trial involving the prototype device, there are two stages. Stage 1 includes
the imaging of human patient breast cancer samples using intrinsic fluorescence. This
stage does not involve the administration of any drugs. Stage 2 includes the imaging of
human samples after the administration of 5-aminolevulenic acid (5-ALA), a pro-drug
- 26 -
that is converted into protoporphyrin IX (PpIX) in the body. This drug has been
previously approved for clinical application in photodynamic therapy [45], but has not
been approved for application in fluorescence imaging in breast cancer. To continue with
stage 2 of my clinical trial, a CTA needs to be submitted to have 5-ALA approved for
breast cancer application. Details about the drug and its manufacturing process must be
provided in the CTA. However, this is proprietary information and the German
manufacturer, Photonamic, is still in the process of negotiating terms of usage with
University Health Network before they will be willing to release this information to
Health Canada. Thus, my CTA application cannot proceed until the negotiation reaches a
consensus. In the scope of this thesis, only applications pertaining to stage 1 were
explored.
Following these governmental approvals, this device also needs to receive
Institutional approvals. First, the entire protocol needs to be drafted and approved by
REB. This marks the initiation of the clinical trial. Since part of the trial requires the
device to operate in the Operating Room (OR), the CPD needs to validate that the device
can be properly sterilized for the sterile surgical field. Then, any participating staff needs
to obtain permission to enter the OR during the operation to use the device.
3.4. Objectives
This is a thesis project that does not aim to test a hypothesis, but to achieve
several objectives. Overall, this project attempts to determine if it is feasible to bring this
prototype device into breast cancer imaging, and, if so, what aspects of the device need to
- 27 -
be changed to make the transition from its original application (wound care) to breast
cancer. Ultimately, this project seeks to bring this device into the operating room to be
used as a margin detection tool in breast cancer surgeries.
3.5. Specific Aims
There are 4 specific aims within the scope of this project:
• First, the device needs to be approved by various administrative boards before any
trials can be conducted.
• Next, the device needs to be used on breast cancer samples to gather pilot data on
the autofluorescence properties of human breast samples. This result would help
reveal if the current filter settings on the camera are adequate for breast tissues.
• Other parameters of the device – sensitivity, resolution – also need to be
determined, more reasonably with phantom materials instead of real tissues (due
to the unpredictability of real tissues).
• In the event that autofluorescence alone is insufficient in delineating tumour from
healthy tissue, other contrast agents need to be explored to enhance the contrast
between tumour and healthy tissue.
4. Aim 1 – Obtaining Administrative Approval for Intra-
Operative Use
4.1. REB Approval
- 28 -
As mentioned in section 3.2, this prototype device is already undergoing several
human clinical trials at the University Health Network (UHN) and partnering facilities.
Its clinical applications include predicting bacterial infection in chronic wounds,
identifying sarcoma (on the leg), visualizing pre-cancerous polyps in colon cancer and
differentiating oral cancer cells from normal tissue. To translate the device from one of
those applications to my new breast cancer application, a new REB approval needs to be
obtained. This process starts by drafting a detailed protocol about the entire trial, which
includes:
• the objectives of the trial
• the number of patients to be recruited, and the inclusion and exclusion criteria
• the possible endpoints of the trial
• the specific procedures that will be taken during the trial
• the data to be collected
o any use of human samples or animal models must be outlined and reasons
justified
• if a drug is to be used (in this case, clinical-grade 5-ALA), the protocol must also
justify the need to use drugs, and provide evidence that the drugs are safe for the
patients
• a patient consent form detailing the potential risks for participating in the trial
• the participating staff and their contact information
This protocol was drafted and reviewed by all participating members on the team,
then sent for review to the REB. Since this device has already obtained approval from
- 29 -
Health Canada, the REB approved our application with few questions, and I commenced
my clinical trial.
As I moved forward in the trial, I collected data that led to the modification of the
protocol. Per institutional requirements, any time there has been a deviation from the
original protocol as submitted to REB, an amendment must be made to identify the
change and justify the change. Over the course of my trial, 6 such amendments have been
submitted. Each amendment required an amendment application form which highlighted
the change, and the reason for the change. In a supplement document, background
information relating to the change was provided, and a complete protocol reflecting the
latest procedures was also included. This package was typically drafted by me, and sent
for review amongst the participating members. Once approved by the group, the Principle
Investigator signed the application, and it was submitted for expedited review at the next
monthly REB meeting.
• Amendment 1 was a request for additional hematoxylin and eosin (H&E) staining
reports. This allowed me to make a correlation between the AF signal and the
underlying tissue composition.
• Amendment 2 requested permission to use point spectroscopy to quantify the
wavelengths that were being detected.
• Amendment 3 requested permission to use a contrast agent 5-ALA to improve the
tumour-to-normal contrast.
- 30 -
o 5-ALA has been previously approved for human clinical use, albeit for
different applications (brain cancer, skin cancer, Barrett’s esophagus). 5-
ALA produced PpIX in cells, generating a red fluorescence signal that was
preferentially sequestered in tumour cells (e.g. brain cancer). If the same
effect could be observed in breast cancer, tumour cells could be easily
delineated from normal cells in an unknown sample. The impact of adding
5-ALA to patients was described in the application. Briefly, since this is a
naturally-occurring molecule in the human body, there was no expected
negative impact from administrating additional 5-ALA to the patient.
However, at low concentrations (10mg/kg), 5-ALA has displayed
embryotoxicity in chicken embryos [46]. For precautions, it should be
avoided if the patient is pregnant. This risk was also reflected in the
updated patient consent form used to recruit patients for the trial.
• Amendment 4 requested the inclusion of additional imaging staff onto the
protocol, in the event that I was not able to take the images.
• Amendment 5 was to update the patient consent form that would reflect the new
protocol after the implementation of all the changes from the amendments to date.
• Amendment 6 asked for continuation of recruitment until there were enough
patients to guarantee 30 sets of individual patient data.
o This amendment was added after the trial had been continuing for over a
year. Due to the inclusion and exclusion criteria set by the pathology team,
and a delay in getting the point spectroscopy device ready, I was not able
to collect full data sets at the end of my original patient recruitment limit.
- 31 -
All 6 amendments were submitted and accepted by the REB at its monthly meeting.
In addition to the initial REB approval application and the amendments, each
application only received approval for a maximum of one year. After this period, the
application needed to be renewed for an additional one year approval. In the renewal
application, any deviations from the protocol since it was approved a year ago needed to
be listed, as well as any changes to be made to the patient cohort. Renewal applications
were expected to be submitted 2 months in advance, and the new approval must be
received before the original expiration date. Two renewals were submitted during the
course of my thesis work. The current application was approved for Sep 15, 2013.
4.2. Equipment Sterilization
In addition to REB approval, the device also needed to receive approval from the
CPD in order to enter the OR. The CPD is the department responsible for overseeing all
devices to be used in the OR and how the devices must be processed to be compliant to
the safety standards. In the case of our device, it needed to be sterilized for the sterile
surgical field. This sterilization method must not compromise the integrity of the device,
or significantly impact the effectiveness of the device.
4.2.1. Available Sterilization Methods
- 32 -
At UHN, approved sterilization process are autoclaving, plasma sterilization and
draping. The first two methods attempt to sterilize the device by deactivating
microorganisms through irreversible metabolic inactivation or breaking down of
structural components [47] . Autoclaving sterilizes by subjecting the device to high
pressure saturated steam at 120 degrees Celsius [48]. This process is capable of
inactivating bacteria, fungi, viruses and spores. Unfortunately, the outer casing of the
device is made of a polymer that will disintegrate under such a high temperature, and
cannot be subjected to sterilization by autoclaving. Plasma sterilization is a process where
ultra-violet light is used to generate chemically reactive agents. These chemical agents
react with microorganisms and destroys them to achieve sterilization. Since this is a
reactive process, UHN requires that all devices undergoing plasma sterilization need to
obtain an approval license from their vendors. Unfortunately, the commercial vendor for
the digital camera within the device was unable to provide us with an approval for plasma
sterilization, so the device could not be sterilized this way. This left draping as our only
option. Draping the device means to cover the device with a standard surgical drape (a
sterile plastic bag). In this case, Laser Arm Drape (Cardinal Health) was selected because
it suited the dimensions of our device and was commercially available. This drape would
cover the device entirely with the exception of the power cord, which would be covered
for the part that remains within the sterile surgical field.
After the appropriate method of sterilization was selected, this device along with its
sterilization protocol was presented to CPD at UHN. CPD voted on if the device was
deemed in compliance with regulations. If the device was rejected, a new presentation
- 33 -
must be made at the following monthly meeting. PRODIGI™ was presented at three
consecutive meetings, with suggestions made at each of the first two meetings on the
appropriate type of sterilization method, and the required compliance. During the third
meeting, we presented the device with a demonstration of how it would be draped in the
OR, and it finally received approval from the CPD to move into the OR. This entire
process took place over approximately 6 months.
4.2.2. Interference of Sterilization Process with PRODIGI™ Imaging
The drape covers the device entirely, including the camera lens. This added layer of
plastic in front of the lens is expected to impact the imaging quality. To measure the
degree of impact, two experiments were conducted to check both the reduction in
resolution and the amount of colour shift.
In experiment 1, a phantom target was used to qualitatively determine the impact of
the drape on the image quality. White-light and autofluorescence images were taken with
and without the drape using the phantom target. (Figure 4) The pattern was printed on a
piece of regular commercial photocopy paper. The fluorescent dye was from a standard
high-lighter. The purpose of this phantom was to be able to produce signals under both
WL and FL.
- 34 -
Images taken under the drape had a slight blurring artefact and no colour-shift
under white-light or autofluorescence with the drape.
In experiment two, the degree of loss in resolution was explored in detail. A
United States Air Force (USAF) 1951 resolution test chart was used as the target. The
device was fixed at a height of 11 cm above the table for all images. Since the purpose of
this experiment was to explore resolution loss, only white light images were taken.
(Figure 5)
a) b)
c) d)
Figure 4 – Comparison between undraped WL (a) and FL (b) images and draped WL (c) and FL (d) images taken by the prototype device. Normally, WL images were taken without the filter in front of the lens. However c) was taken with the filter and resulted in the red tint in the image. Otherwise, c) and d) showed a slight blurring artifact compared to a) and b), with no visible colour-shift.
- 35 -
Figure 5 – Spatial resolution comparison between images taken under a) White light, without filter, undraped. b) white light, with filter, undraped. c) white light, without filter, draped. d) white light, with filter, draped. A ~10% resolution loss occurred with the addition of the camera filter, and a ~50% resolution loss occurred with draping.
Table below summarized the results from this test.
Table 1 – Numerical measures from resolution test with drape. Between images with and without filter, a 10% loss in spatial resolution was observed. Draping added another 50% loss in resolution.
Image Reading Resolution (µm)
a) Group 4, element 6 175.4
b) Group 4, element 5 196.9
c) Group 4, element 1 312.5
- 36 -
d) Group 3, element 6 349.7
From the experiment, adding a filter in front of the lens (to remove excitation
light) reduced the resolution by roughly 10%, while draping reduced the
resolution by roughly 50%. While the results seemed sub-optimal, they were
deemed acceptable by the surgeons on the team. Thus draping was accepted as the
method of sterilization for this prototype device.
4.3. Equipment Safety
Health Canada requires all medical devices to comply with a list of standards, which
cover several topics, including [49]:
• methods of sterilization: which parts are provided sterile, the methods used to
achieve sterility, the sterility assurance level intended, and the residues expected
from the method of sterilization
• performance specifications: any claims need to be backed up by data
• output (acoustic limit) and display: demonstrate that the output does not exceed
the acoustic output limits
• thermal index: justification needs to be provided for any thermal index exceeding
6.0
• thermal, mechanical and electrical safety: outlined in standard IEC 60601-1
- 37 -
• patient contact material: identify the composition of the material and its
biocompatibilities
• software: outlined in standard IEC 60601-1-4
Our device is a non-contact device that runs on commercial software. Thus the only
two related standards are sterilization (discussed in 4.2) and thermal, mechanical and
electrical safety. Out of these, thermal safety is the most relevant and immediately
testable.
4.3.1. Thermal Analysis
The drape was tested for thermal disintegration under continuous use. The device
was turned on and kept running for 60 minutes while placed inside the drape.
Temperatures of 5 points of interest were measured at designated time intervals (Figure
6) by pointing a handheld infrared thermometer (Mastercraft, Canada) at the point of
interest from outside of the drape. At each time point, 5 points were measured and
recorded from point 1 to point 5. At the end of 60 minutes, the device was turned off and
the drape was visually inspected for disintegration and damage.
- 38 -
Figure 6 – Points measured in drape safety test. Point 1 and 3 were the heat sinks attached to the LEDs. Point 2 and 4 were the filters directly in front of the LEDs. Point 5 was the filter directly in front of the camera lens and was the point that would be closest to the patient during actual use.
There was no visible tear or deformation of the drape after 60 minutes of
continuous use. The temperatures of all 5 points of interests were below 70 degrees
Celsius for the entire duration of the test (Figure 7), which was well below the 120
degrees Celsius limit outlined in IEC 60601-1 and was considered safe by this standard.
There were concerns that the increase in temperature may affect the performance of the
CCD within the camera. However, the points measured were on the peripherals of the
device and were not indicative of the temperature at the core of the camera where the
CCD was. Furthermore, in real applications, the device was not expected to run
continuously for longer than 1 minute. Within the first minute, point 5 (the point closest
to the CCD within the camera) only showed a 2-degree increase in temperature (26.1 to
28.2), and would not significantly impact the performance of the CCD.
- 39 -
Figure 7 – Results from the drape safety test. Each point was measured using an infrared thermometer at the following time points: every minute between 0-15 minutes, and every 10 minutes between 20-60 minutes. All points were below 70 degrees Celsius, which was 50 degrees below the Health Canada safety limit of 120 degrees Celsius, and was considered safe. The decrease in temperature as observed at time points beyond 40 minutes may have resulted from increased heat transfer between the drape and the surrounding due to a greater temperature gradient created by the elevated temperature within the drape.
4.4. Conclusion
The device was able to be used sterilely with a standard surgical drape (Cardinal
Health, Ltd), which was tested to be thermally safe, with some, but acceptable
compromises in image quality. The device was ready to be used in the clinical trial with
administrative approvals.
- 40 -
5. Aim 2 – Determining Device Sensitivity
5.1. Introduction
The results in the in vitro and in vivo experiments indicated that PpIX signal was
present in tumour, but at such an intensity that was not visually detectable by our
prototype device. Yet point spectroscopy was able to detect the same signal,
demonstrating a difference in signal sensitivity between the two imaging platforms. To
better quantify the difference in sensitivity, an experiment was conducted to compare the
sensitivity of the two platforms.
5.2. Apparatus and Methods
5.2.1. Devices
Two devices were used for this sensitivity test: PRODIGI™ prototype device and
point spectroscopy.
Point spectroscopy was performed in a dark room. The test solution was placed in
a clear cuvette. The detection probe of the spectroscopy device was placed directly into
the solution, submerged at about 0.5 cm below the surface. An excitation wavelength of
405 nm was used. 5 measurements of 100 ns were taken consecutively and averaged to
form a single data point. This step was performed automatically by OceanOptics software.
After each concentration was measured, the cuvette was completely emptied and washed
twice with deionized water, and aired dry to remove residual fluorophore.
- 41 -
Prototype device imaging was also performed in a dark room. Each test tube
containing the test solution was placed against a black mat. The device was set in
fluorescence mode and the entire test tube was imaged. A blank solution (one which
contained no solute) was also imaged in the same test tube as control.
5.2.2. PpIX Solution
PpIX solution was prepared by dissolving 2.6 µg of PpIX powder in 100 µL of
dimethyl sulfoxide (DMSO). The mixture was vortexed thoroughly and centrifuged to
remove all suspending, undissolved PpIX powder particles. The stock solution was
prepared by dissolving the supernatant from above in 1 mL of 1.5N HCl. The true
concentration of the stock solution was determined by PpIX absorbance. An average
absorbance value of 0.374 was obtained at 408 nm from the absorption spectrometer.
Using the extinction coefficient 293.1 [50], the absolute concentration of PpIX was
determined to be 3.2 mM. From here, a serial dilution by a factor of 2 was performed.
Briefly, 1 mL of stock solution at a concentration of 160 µM was prepared. 0.5 mL of
this stock solution was added to 0.5 mL of HCl in a tube labeled A. The resultant mixture
was vortexed and centrifuged briefly. Tube A was set aside to be imaged. 0.5 mL of HCl
was then added to the stock solution to produce the “new” stock solution at half the
concentration. This process was repeated for each subsequent dilution, producing a new
labeled tube (B, C, D… R) in each iteration, and a new “stock” solution with the same
concentration as its labeled counterpart. The concentrations produced in this serial
dilution ranged from 160 µM to 0.61 nM.
- 42 -
On a separate occasion, PpIX solutions were prepared for point spectroscopy in a
similar protocol. An initial concentration of 1.2 mM was determined for the stock
solution, which was subsequent serially diluted by a factor of 10 for 5 iterations. The
emission spectra from the resultant 6 test solutions were measured with point
spectroscopy.
5.2.3. Image Analysis
Point spectroscopy produced numerical data that did not require image analysis.
Device images were analyzed using ImageJ. Briefly, each image was split into colour
channels. Under the red channel the solution region was selected by freehand. The
average pixel intensity was calculated in the red channel, and subtracted from the average
pixel intensity in the red channel of the control image. The sensitivity/linearity plot for
each device was plotted separately.
5.3. Results
The prototype device was able to detect a signal from concentrations 2.4 nM and
above. For concentrations 625 nM and above, the PpIX signal saturated the red channel
(pixel value 255) and the average pixel value in the image was not representative of the
concentration of the PpIX solution measured. The data points were graphed as shown in
figure 34.
- 43 -
Figure 8 – Fluorescence signal intensity of PpIX solution at various concentrations as detected by the prototype device. A detection limit of 2.4 nM was observed, and a display limit of 625 nM was measured, above which the pixels in the red channel were saturated and were no longer representative of the PpIX concentration in the solution. Within these limits, there was an unexpected jump in FL intensity for concentrations below 4.9 nM. Each error bar indicates one standard deviation (N=3).
An unexpected increase in FL intensity was observed at the 4.9 nM data point,
and this increase was proportionally propagated to all concentrations below 4.9 nM. This
lead me to believe that a protocol error has been made during serial dilution at this
particular data point. For the purpose of this analysis, all data points from 4.9 nM and
below were discarded. All data points from 625 nM and above were also truncated to
remove the saturated region. The result graph is shown in figure 35.
- 44 -
Figure 9 – A re-plot of the truncated data collected in figure 34. The saturation region and concentrations below 4.9 nM were removed. A logarithm fit was introduced to the plot with a coefficient of correlation of 0.98. Each error bar indicates one standard deviation (N=3).
The “valid” data region shown in figure 35 had a logarithm relationship. The log
trend line had an R2 value of 0.98, indicating that the camera in the prototype device had
a strong logarithm signal-to-concentration correlation within the PpIX emission region.
Point spectroscopy analysis consisted of 5 data points plotted in figure 36.
Concentrations from 0.12 mM to 12 nM were tested. A signal was detected for all 5
concentrations. A linear fit was introduced to the plot with a coefficient of correlation of
0.99.
- 45 -
Figure 10 – Point spectroscopy data collected from PpIX solutions at five different concentrations. Each data point was the average fluorescence intensity of 5 points between 635.14 nm and 635.93 nm. The six concentrations used were 0.12 mM to 12 nM, diluted by a factor of 10 in each subsequent dilution. A linear fit introduced to the data had an R2 value of 0.99. Each error bar indicated one standard deviation (N=5).
5.4. Discussion
This experiment aimed to determine the lower detection limit of the prototype
imaging device, as well as its linearity in signal detection. Results from experiments in
section 6 demonstrated that point spectroscopy had a higher sensitivity than the prototype
device, as illustrated in the example where the prototype device was unable to visualize a
PpIX signal in the tumour, but a signal was detected using point spectroscopy. In this
experiment, using a series of PpIX solutions at varying concentrations, the approximate
lower detection limit of the prototype device was determined to be around 2.4 nM. The
exact value could be determined by repeating this experiment with concentration
differences smaller than those used in this experiment. At 4.9 nM, an unexpected increase
- 46 -
in FL signal was observed. This error was systematically propagated down all
concentrations below 4.9 nM. This was most likely due to an error in my serial dilution at
this particular concentration. All points below 4.9 nM were removed from the analysis.
Within the measurable, valid concentrations, the prototype device had a
logarithmic detection relationship in the emission wavelength of PpIX. This is expected
from a commercial digital camera. Although CCD in the camera exhibits a linear
photometric response [51], the digital processing unit generally adopts a nonlinear
transformation to make the final image more pleasing to the human eye [52]. Since the
human eyes detect light intensity on a logarithmic scale, digital cameras try to mimic the
same effect by applying a logarithmic tonal curve to the absolute signal.
No conclusions can be made about the linearity of the device at concentrations
exceeding the saturation concentration (under the current acquisition conditions),
however the experiment can be repeated with a different set of acquisition conditions (e.g.
lower exposure) increase the maximum measurable concentration below pixel saturation
point in the red channel.
In my in vivo experiments, the administrated 5-ALA concentration was
approximately 0.5mM. At an expected conversion efficiency of 5% [53], the expected
PpIX concentration at 3 hours post-injection was 25 µM, which should have been
detected by the camera if all PpIX signal was concentrated within the tumor. The lack of
- 47 -
signal in the observed results could indicate an inefficiency in the conversion of 5-ALA
to PpIX within the tumour, most likely due to low tumour-uptake of 5-ALA.
The point spectroscopy device also displayed linearity for the range of
concentrations tested. Furthermore, point spectroscopy was able to detect a signal at all
tested concentrations, signifying a sensitivity greater than the minimum concentration
tested. The confirmed sensitivity for PpIX fluorescence in this set of experiments was 1.2
nM.
6. Aim 3 – Evaluating Existing Filter Settings on Breast
Cancer Samples Using AF Alone
6.1. Autofluorescence
Autofluorescence (AF) uses endogenous tissue fluorophores to produce light signals
[54]. Certain tissues in the body have natural emission when excited under certain
wavelengths. When cells undergo physiological or pathological changes, the amount of
endogenous fluorophores in the tissue may change, resulting in a change in
autofluorescence signature. For tissues with a wide emission spectrum, different
excitation wavelengths may lead to different emission (response) wavelengths. This
property, while problematic in fluorescence microscopy, can be captured and analyzed
for diagnostic and intervention purposes when paired with the correct filter settings.
Information about the underlying tissue structure and morphology may be inferred from
- 48 -
its AF signature. Since the origin of AF is the organelles or cells within the tissue, AF can
easily achieve single-cell resolution when combined with microscopy. AF has the
advantage of processing in real time because it is a natural biophysical phenomenon that
requires minimal pre-imaging treatment or fixation, making it an attractive choice for
providing visual feedback in image-guided medical functions.
Considerable research has been conducted on intra-operative image-guided surgeries
for oral, gastrointestinal and brain cancer, but little research has been reported for breast
cancer.
Autofluorescence (AF) for tissue differentiation relies on different AF signatures
produced by each tissue type. Normal human breasts contain several tissue types that
produce distinctive AF signatures. The main AF peak in breast tissue is attributed to
flavin adenine dinucleotide (FAD), a redox cofactor. Its emission peaks at 520 nm when
excited with 488 nm light. Hemoglobin causes a characteristic absorption at 540-580 nm,
[55] and may attenuate the signal if the excitation light is within this range. Collagen and
adipose tissue produce very different peak emissions depending on the excitation
wavelength. Under 405 nm excitation, collagen emission peaks at 475 nm [56]. Adipose
tissue emission at 405 nm excitation has not been reported in the literature, but is
expected to be beyond 600 nm. PpIX, where present, has a signature AF emission at 635
nm.
- 49 -
Previous research has demonstrated through AF ductoscopy that it is technically
possible to isolate intraductal lesions in the human breast both ex vivo [57] and in vivo
[58]. Tumour-to-normal contrast may be enhanced with additional contrasting agents, but
there is sufficient contrast using AF alone. In particular, tumour-to-normal AF contrast is
the most prominent at 390 nm excitation. However, Health Canada placed a maximum
exposure limit on wavelengths below 400 nm (UV region). As a compromise,
PRODIGI™ uses 405 nm excitation light for breast cancer AF imaging.
6.2. Materials and Methods
6.2.1. Patient Sample Preparation
Human patient samples were obtained through collaboration with in-house
clinicians. Eligible female patients 18 years or older who are scheduled for standard
surgery for primary invasive cancer were recruited for the clinical trial. Exclusion criteria
included any pre-operative treatments, lack of a in-house tumour core biopsy, a prior
history of photosensitivity, liver diseases and recurrent diseases, as well as any pregnancy
or general inability to consent. Eligible patients were recruited and consented on the day
of and prior to their surgery. Patients with liver diseases were excluded from the study in
anticipation for the inclusion of contrast agents, which would be flushed out of the body
through liver metabolism and would require that the patients have normal liver functions.
Samples were obtained directly in the surgical pathology lab immediately following
lumpectomy or mastectomy procedures. In total, 21 lumpectomy and 15 mastectomy
- 50 -
samples were received from consenting patients. Samples were processed by the
pathologist in the surgical pathology lab. Briefly, each sample was dried of surface blood
and surface-coated with silver nitrate (AgNO3). AgNO3 allowed the pathologist to detect
the surgical surface of the sample once the sample had been cut up for analysis. The
anterior surface of the sample was painted with two dyes: Blue and Indian dye. Blue dye
was coated on the superior half of the sample on the anterior surface and Indian dye was
coated on the inferior half of the sample. These dyes served a similar purpose as AgNO3
– to allow the pathologist to orientate the sample once it had been cut up. After surface
treatment, the sample dimensions were measured, and any surface markings were
recorded as a sketch for co-localization purposes. The sample was then sliced in a bread-
loaf style. The location of the tumour was determined by palpation and the dimension of
the tumour was measured.
6.2.2. Autofluorescence Imaging with PRODIGI™
Once all measurements were taken, the sample was placed on a non-reflective black
mat in a dark room. Each slice of the tumour was imaged with PRODIGI™ under white
light (WL) mode (images taken under room light), and fluorescence (FL) mode (images
taken under fluorescent LED in the absence of room light). In FL mode, a filter was
added in front of the camera lens to block excitation light reflected off the surface of the
sample. For each slice, multiple images were taken under both WL and FL mode, and in
areas where tumours were identified, additional zoomed-in images were taken.
- 51 -
Sample orientation was recorded using sutures: a long suture indicated the lateral
surface, and a short suture indicated the superior surface.
6.2.3. Data Consolidation and Documentation
To organize and centralize all the images taken for the clinical trial, an image
catalogue was constructed to provide a quick overview of each eligible sample. Within
the catalogue, information about the surgery date, surgery type, tumour type, surface
marker combination and any other comments made by the pathologist during the image
session was recorded. A set of WL and FL image for each sample imaged was also
provided. This catalogue was stored and backed up on the internal network at UHN.
Following each image session, all images taken from the session were stored onto
an encrypted computer within UHN. Each session was dated, and each image was
labelled by the slice number. Then, the most representative set of WL and FL images
from each slice was selected and added into the image catalogue. Additional comments
were then included in the catalogue where applicable.
6.2.4. Tumour Fluorescence Analysis
Each image was analyzed using the image-processing software ImageJ (National
Institutes of Health, USA). To determine tumour fluorescence intensity, 3 measurements
were taken from each selected image: tumour, surrounding tissue, and background
threshold. A freehand tool was used to select the entire tumour based on the margins
- 52 -
determined by the pathologist. An average pixel intensity of the selected area was
measured. This value was assigned as the tumour signal (Ts). An area within the 1 cm
surrounding the tumour (as selected) was then selected and the average pixel intensity
was measured as background signal (Bs). Thirdly, the darkest area on the image was
selected and measured as background threshold (Bt). The background threshold was used
to compare different shots of the same tumour. While the distance between the camera
and the tumour was kept as constant as possible, a slight change in height between the
lens and the surface of the tumour would affect the average brightness of the shot. To
make subsequent shots comparable, Bt was subtracted from all signals to calibrate the
systemic difference in brightness between these shots.
The ratio of relative fluorescence between tumour and background was determined
using the formulae:
Ts – Bs – Bt
Bs – Bt
If the resultant value was positive, the tumour was categorized as “bright”; if it was
negative, then the tumour was categorized as “dark”.
6.2.5. Point Spectroscopy with OceanOptics Spectrometer
After each slice was imaged, point spectra of areas of interest were taken if the
tumour was large enough for punch biopsy (diameter > 2 cm). To take a point spectrum,
PRODIGI™ was first used to screen the surface of the tumour slice. An area displaying
characteristic signals of “green” or “red” was selected as an area of interest. A point
Eq 1
- 53 -
spectroscopy probe (5mm diam.) was place directly on top of the area of interest, and
after the room lights as well as PRODIGI™ were turned off, a point spectrum was taken.
Each data point was calculated by averaging 5 consecutive 500-ms measurements. (This
was handled by the SpectraSuite software). This would reduce the probabilities of
capturing an atypical signal.
6.2.6. Punch Biopsy
Immediately after the sample had been imaged, punch biopsies were taken if the
tumour met the sampling criteria of having a diameter larger than 2 cm. Tumours smaller
than 2cm in diameter were excluded to retain tumour tissue for clinical diagnoses. Punch
biopsy cores were to be used to correlate with point spectroscopy data and uncover any
relationships between the AF signals and the underlying biological components within
the tumour. The entire procedure was performed by a trained pathology assistant (PA)
and was done in conjunction with point spectroscopy. Upon identifying an area of
interest, a 2 mm biopsy core was taken using a biopsy punch at the same location as the
point spectra, after the point spectra had been taken. In each sample, 4 cores were taken
from the same slice: 2 from tumour region as determined by the PA, and 2 from non-
tumour region. All samples were dyed on the un-imaged surface using Indian dye to mark
the orientation of the samples and delivered to the pathologist for reading in a cassette.
6.3. Results
- 54 -
I recruited 36 patients for this arm of the study. Out of these 36 patients, 1 patient
received pre-op therapy and was excluded from the data set. 2 patient samples were very
small (less than 4cm in diameter) or irregularly shaped and were not sliced, so those
samples could not be imagined. The remaining 33 patient samples were imaged under
both WL and FL mode.
Under WL mode, all sample surfaces appeared to be brown to dark red in colour and
crisp in texture. The red tint in colour resulted from bleeding during surgery, and crisp
texture resulted from burning of tissue in the resection process. Once sliced, the internal
structure of samples appeared to be a network of soft tissue, varying from pale white to
light orange in colour. In most cases, blood was inevitably sealed in the sample when it
was initially electrocautered. During slicing, blood would ooze out of the vessels within
the sample. If sufficient blood was sealed in the sample, the oozing of blood post-slicing
would turn tissue sample from pale orange to deep orange in colour. This would
negatively affect the subsequent imaging because blood would absorb some AF signal. A
deep-blue spot was often observed near the inferior surface of the tumour where
methylene blue was injected during the surgery (to help locate adjacent lymph nodes).
Methylene blue did not interfere with imaging due to its unique absorption spectrum
(Figure 8). Tumours generally did not appear significantly different from their
surrounding tissue in colour, but much firmer in texture. Most ductal carcinoma had well-
defined edges and appeared round in shape, while most lobular carcinoma had undefined
edges and were irregularly shaped. Occasionally, if the tumour was necrotic (by visual
inspection), the centre of the tumour was found to be indented with soft pus in some
- 55 -
cases. In both tumour and normal tissue, while each lobule was clearly visible, there was
little visual contrast between different components (lobules, adipose tissue, and
connective tissue) of breast tissue.
Figure 11 – Absorption spectrum of methylene blue. Absorption was lowest at around 405 nm, the excitation wavelength used by our device. Thus it did not attenuate any signal by absorbing the excitation light. Reproduced from [59].
Under FL mode, there was a mixture of different autofluorescence signatures,
presumably produced by different components of the sample. Bright green fluorescence
appeared in the form of a network throughout the breast and displayed characteristic AF
signatures of connective tissue. Pale pink fluorescence, characteristic of adipose tissue,
covered most of the remainder portion of the tissue. Lobules could be inferred from the
surrounding connective tissue network. Methylene blue did not appear to have any effect
on the AF signature, while blood generally reduced AF signal on the whole. Tumours had
a wide range of AF signals, from much darker than the background adipose tissue, to as
- 56 -
bright as the bright green connective tissue. Bloody tumours tend to appear much darker
than bloodless tumours under FL mode.
To better describe tumour AF signatures, each image was processed as outlined in
section 5.2.4. Each tumour was categorized as either “bright” or “dark. The tumour
margin was selected based on visual inspection and palpation, without histopathology
verification (Figure 9).
Out of 33 tumours imaged, only 29 were available for analysis. The remaining
images were excluded because they were acquired when the camera was in auto-
adjustment mode and processed the raw signals. Out of these 29, 17 were classified as
“bright” and 12 were classified as “dark”. (Figure 10) These classifications were further
divided based on the histopathologic diagnosis of the tumours. (Figure 11)
Figure 12 – An example of WL and FL human breast cancer image taken by PRODIGI™. Under the WL image, tissue components could not be distingushed, and the tumours (identified by palpation, circled in red) did not appear any different from their surrounding tissue. Under FL, an array of components can be seen based on their unique AF signatures. In this image, green FL was associated with connective tissue, pale pink FL was associated with adipose tissue, and tumours (identified by palpation, circled in red) appeared darker than their immediate surrounding tissues.
- 57 -
Figure 13 – Signal-to-background ratio of every ex vivo sample analyzed. Tumour signals were normalized to background in each image. A positive ratio was classified as “bright”, a negative ratio was classified as “dark”. Out of 29 patients, 17 patients were classified as “bright”, 12 were classified as “dark”.
Figure 14 – Tumour breakdown by its pathological diagnosis. 4 different types of tumours were present amongst all the patient samples - invasive lobular, invasive intraductal, adenoid cystic, invasive ductal. There was no statistical difference between “bright” and “dark” tumour under each type, suggesting that tumour signal as observed under AF was not indicative of its underlying pathology.
- 58 -
Histopathology diagnosis included the origin of the cancer (ductal or lobular) and
the invasiveness of the cancer (invasive or in situ). Out of the 4 categories presented, both
invasive intraductal and adenoid cystic had only 1 case and this was not statistically
significant. Invasive lobular had 4 cases. While all 4 cases had bright tumours, they were
still not statistically significant when analyzed using student’s T-test (P >0.05). The
remaining category invasive ductal had 10 bright tumours and 8 dark tumours and did not
show a statistical pattern between tumour brightness and underlying pathology (P >
0.05), which was in agreement with the literature [30].
Due to patient recruitment constraints, we did not obtain any useful patient data
for point spectroscopy. Only 2 patients were able to provide point spectroscopy data, of
which 1 was collected before the point spectroscopy device was fully optimized (thus the
data was not usable). Thus so far, there was only 1 complete set of patient data that
composed of AF imaging, point spectroscopy and punch biopsy. From this one patient,
punch biopsy results are shown in table 2.
Table 2 – Histopathology analysis of the biopsy cores obtained from patient # 35 indicated that the peak emission wavelengths as measured by point spectroscopy were indicative of the % of connective and adipose tissue in each core. Higher % of connective tissue corresponded to a lower peak emission wavelength (closer to 475 nm, the AF wavelength of collagen). T1 – tumour core 1, T2 – tumour core 2. N1 – normal core 1 (green). N2 – normal core 2 (red).
% connective tissue
% adipose tissue
% invasive tumour
Comments
T1 100 0 40 But a mix of cell types with connective tissue and calcification
T2 95 5 35 But a mix of cell types with connective tissue
N1 65 35 0
N2 5 95 0
- 59 -
The corresponding point spectra are shown in figure 12.
Figure 15 – Point spectroscopy for patient #35 showing the peak emission wavelength for each core. T1, T2, N1, N2 correspond to the 4 cores in punch biopsy. T1 and T2 with 100% and 95% connective tissue had lower peak wavelengths at 499 nm and 500 nm, respectively. N1 and N2 had higher peak wavelengths at 508 nm and 524 nm, respectively.
6.4. Discussion and Conclusion
The purpose of this set of experiments was to determine if the device was able
detect tumour margins in a human breast cancer sample with AF signals alone using the
current configuration. These results demonstrated that the device was able to detect
different signals from different tissue types. Particularly, the device was able to isolate a
red signal associated with fatty tissue, and a green signal associated with connective
tissue. While the commercial camera embedded in the device was not equipped with the
- 60 -
proper imaging processing software that would give quantitative feedback on the
wavelengths associated with the red and green signals, this information can be obtained if
the device was used in conjunction with point spectroscopy. As demonstrated in the
complete patient data set (patient #35), N1 was taken from a “green” spot on the tissue,
and had a corresponding peak wavelength of 508 nm.N2 was taken from a “red” spot on
the tissue, and had a longer wavelength at 525 nm. Histopathology analysis revealed that
N1 consisted of 65% connective tissue, whereas N2 consisted of only 5% connective
tissue and 95% adipose tissue. This is in agreement with the literature stating that a green
AF signature is associated with connective tissue, whereas a red AF signature is
associated with adipose tissue. From this set of data alone, this device already
demonstrated its capability to differentiate between different AF signatures. When used
together with point spectroscopy, we can extract even more information about the tissue
composition. For example, T1, T2, N1 and N2 had a decreasing percentage of connective
tissue. Correspondingly, the peak wavelengths of the 4 cores increased in order, showing
that a lower peak wavelength corresponds to a higher percentage of connective tissue in
the composition. More data sets could further explore the validity of this trend. The
presence of malignant cells did not seem to impact the peak wavelength of the core; more
patient data may help reveal more about the impact of the presence of these malignant
cells. However with the existing patient data, no definitive tumour margins can be
delineated using AF alone, simply because tumour signals were very heterogeneous, with
no point spectra or biopsy data to further explain any observable trends. This means that
PRODIGI™ in its current settings has limited ability to predict tumour locations (and
subsequently, tumour margins) on an unknown sample. Specifically, if the tumour
- 61 -
displayed a signal distinct from either the characteristic red associated with adipose tissue
or the characteristic green associated with connective tissue, then the tumour would be
delineated from its surrounding tissue. However, if the tumour displayed a signal very
similar to one of the two “normal” signals, such as in the event of a very fibrous tumour
with high connective tissue content, then it would not be distinct enough to be delineated
with confidence. Furthermore, image analysis confirmed that there was no distinctive
pattern between apparent tumour signal and its underlying diagnostic histopathology.
Thus I could not construct a look-up table that would aid in identifying a tumour based on
its known diagnosis. On a side note, the camera in the prototype device is non-linear, so
the true variability in light intensity across different samples is likely much greater.
However, this does not fundamentally impact the result from this analysis.
One way to overcome this limitation is to introduce a contrast agent that can
enhance the contrast between tumour and normal cells. One such contrast agent is 5-ALA,
which can be added to provide a uniform signal to all tumour types, thus eliminating the
need to characterize tumour signal. 5-ALA has been in use for other cancer applications,
such as skin cancer (for treatment with photodynamic therapy) and brain cancer (for
tumour cell detection/visualization). If it is shown that it is also compatible with breast
cancer, then it could be used to enhance tumour fluorescence signal and to delineate
tumour from surrounding healthy tissue. Furthermore, the signal obtained using 5-ALA is
expected to be stronger than autofluorescence. This could help alleviate the signal
attenuation in vivo from photon absorption by tissue, and wide-field scattering (which
- 62 -
contributes to background autofluorescence) [61]. More background information about 5-
ALA is outlined in section 6.
Another limitation in the technology arose from the limitation in the commercial
camera embedded in the device. This digital camera was intended for macroscopic
imaging and did not have the resolving power (200+ µm as tested in the resolution
analysis) to distinguish pixels at the single-cell level (10-100 µm). This means that the
device was not capable of visualizing individual tumour cells that may be hidden within
the adjacent connective tissue. However, this limitation could easily be overcome by
replacing the camera with one that had a higher resolution as needed.
Through this experiment, the device demonstrated its ability to excite the tissue
with a single light source and detect a range of AF signals from different tissue
components within a human breast sample. Using AF alone, the device was able to
differentiate between adipose and connective tissue, and visualize tumour signals that are
distinctively different from normal adipose and connective tissue signals. When used in
conjunction with point spectroscopy, a correlation could be observed between the peak
emission wavelength and the underlying tissue composition – a high percentage of
connective tissue in the illuminated area translated to a peak emission wavelength closer
to the green fluorescence wavelength of collagen (450 nm). Tumour contrast was
suboptimal due to the intrinsic heterogeneity of the signals, but this issue may be resolved
by administrating a contrast agent. Details about the contrast agent 5-ALA is explored in
the next two sections.
- 63 -
7. Aim 4 – Feasibility Testing of Using a Contrast Agent (5-
ALA)
7.1. Contrast Agents and Image Probes
7.1.1. Image Probes
Fluorescence signals can be generated from either endogenous or exogenous
fluorophores. Not all body tissues have endogenous fluorophores (autofluorescence), so
exogenous fluorophores (image probes) are sometimes added to provide contrast for
structures of interest. Fluorescent molecular probes have been extensively used to label
biological markers ex vivo and in vivo. In general, image probes designed for in vivo use
can be divided into three categories: untargeted, targeted, and activatable. Untargeted
probes do not have tissue- or cell-specific recognition sites and are used to provide
transient, local or systemic fluorescence signals based on probe diffusion (typically
intravenously) [61]. These are frequently used to trace circulatory or lymphatic pathways,
or track changes in blood vessels. Targeted probes have recognition sites for specific cell
surface markers and are intended to only bind to cells of interest. Targeting is achieved
by conjugating fluorophores to target-specific antibodies, peptide sequences,
nanoparticles, quantum dots or molecular beacons. In certain cases, targeting is achieved
by taking advantage of specific cellular metabolic pathways that preferentially break
down or produce fluorophores. If these pathways are only present in certain cell types,
then by tuning imaging filters to the corresponding wavelengths, only these cells will be
- 64 -
visible under fluorescent conditions. Activatable probes are molecules where multiple
fluorophores are attached but positioned in such proximity that self-quenching is
achieved, and the molecule is not visible under fluorescence. Upon entering the body or
attaching to sites of interest, the molecule is structurally altered (typically by an
enzymatic process specific to the target ) and “activated”. Fluorophores on the molecule
are restored to an unquenched position, and the molecule is visible under excitation of
fluorescence. This type of probe requires molecular engineering, but produces a high
signal-to-noise ratio.
In breast cancer margin detection, targeted probes are most applicable. Ideally,
tumour-targeting molecules are attached to fluorophores. Upon injecting into the body,
the targeting molecules bind to only tumour cells, and excess molecules are removed
from the circulatory system. Tumours are then visible under fluorescent conditions,
whereas normal cells are not fluorescent.
7.1.2. Fluorescent Antibody Imaging Agent
Antibodies are macromolecules with high affinity to specific antigens. When
conjugated with a fluorophore, these molecules make excellent targeted imaging probes.
Natural antibodies have a long clearance time from the body, which results in unwanted
non-specific background fluorescence [62]. To address this problem, antibody fragments
have been used to shorten the clearance time and reduce non-specific binding. Antibodies
are protein structures with specific binding recognition sites. Artificial antibodies or
- 65 -
peptide structures can be constructed to match antigens of interest, making them highly
versatile and target-specific. As fluorescence imaging probes, antibodies allow
multispectral imaging by having multiple antibodies (or its fragments) conjugated with
different fluorophores attach to different markers on the same cell surface. For signal
processing, a third antibody can be used to non-specifically tag all cells. This background
fluorescence can then be subtracted from the other two signals. Alternatively, quenching
can be used to reduce background noise. Inactive probes with quenched fluorophores
undergo structural change upon binding with target sites, and the fluorophore moves into
an unquenched position to generate a signal. This approach allows a high signal-to-noise
ratio in vivo.
Research has reported pre-clinical application of antibody-based fluorescence
imaging in breast cancer, but with limited scoop [62] Specifically, while there are
multiple cancer-specific surface markers (HER2, ER, Progesterone receptor) that could
be used for pre-clinical proof-of-principle experiments, there is no single generic breast
cell surface marker that is present on all cancer cells for clinical diagnosis. In fact, a
fraction of breast cancer patients are “triple-negative”, indicating that they do not possess
any of the three major breast cancer surface markers. In the absence of a surface protein
marker, antibodies cannot be used to tag the cells, rendering the fluorescence agents
infeasible for specific types of breast cancer.
7.1.3. Contrast Agent 5-ALA
- 66 -
5-aminolevulinic acid (5-ALA) is a biomolecule that naturally occurs in the
human body. It is an intermediate product in the heme pathway and is used to synthesize
protoporphyrin IX (PpIX) downstream, which has fluorescence properties with a peak
excitation wavelength of 405 nm and a peak emission wavelength of 635 nm [55]. In a
cell in its natural state, 5-ALA is closely controlled to prevent PpIX accumulation
downstream. This negative-feedback loop is compromised in cancerous cells due a
decrease in activity in the enzyme ferrochelatase and the limited availability of iron to
consume PpIX downstream, resulting in an accumulation of PpIX in malignant cells [63].
5-ALA-induced-PpIX fluorescence has been extensively tested in cell lines, in vivo
animal models and human clinical trials [64] and has shown increased contrast in brain,
prostate and ovarian cancer. Further 5-ALA-PpIX fluorescence has been shown to be
increased in breast cancer cells, making it a suitable candidate for intra-operative margin
assessment probe.
In an early clinical feasibility test, Ladner et al. have shown that breast cancer is
highly heterogeneous and causes variations in metabolic contrast agents such as 5-ALA
[65]. Furthermore, 5-ALA is embryotoxic and is not suitable for pregnant women [46].
Yet previous experiments in our lab demonstrated that 5-ALA-induced PpIX
fluorescence preferentially sequesters in cancer cells regardless of the malignancy, and
the literature shows that PpIX fluorescence was significantly higher than surrounding
normal tissue in an ex vivo breast tissue study [65]. Based on these two observations, 5-
ALA/PpIX fluorescence was selected as a promising candidate for tumour-to-normal
contrast enhancement that could provide ubiquitous contrast for all breast tumours
- 67 -
regardless of its malignancy and structure composition. Many studies have been reported
for 5-ALA/PpIX contrast enhancement in various cancers [66][67], but little has been
done on 5-ALA/PpIX with breast cancer at the start of this thesis project. Thus prior to
adopting 5-ALA as a new contrasting agent to be used in conjunction with PRODIGITM,
a feasibility test needed to be conducted in in vitro and in vivo models of breast cancer,
with the intention to:
• Confirm that 5-ALA/PpIX produces fluorescence signature in existing human
breast cancer cell lines.
• Confirm that 5-ALA/PpIX fluorescence is tumour-specific
• Confirm that in vitro results are translatable to in vivo xenograft models
7.2. Clinical Implementation of 5-ALA at UHN
7.2.1. ALA and Clinical Interference – FISH Analysis
FISH, or Fluorescence In-Situ Hybridization, is a staining process used clinically
to detect surface protein expression in cells. In breast cancer applications, FISH is used to
detect the presence of HER2 receptors. FISH operates by binding fluorescently-tagged
antibodies to the surface markers of interest and counting the number of activated dye
“spots” in each cell. A scoring system is used to determine the relative protein expression
on the surface of the cells. The results from FISH analysis could determine the
subsequent hormone treatment that may be prescribed to the patient. 5-ALA-induced-
PpIX fluorescence has an emission peak of 630 nm, which is close to the emission peak
of commercial SpectrumOrange dye (588 nm) used in FISH analysis. Clinicians are
- 68 -
concerned that the fluorescence emission from PpIX may also produce red fluorescence
emission in the FISH red channel, and cause detection interference. To address this
concern, a validation analysis was included in the 5-ALA arm of the study to analyze the
degree of interference between PpIX and SpectrumOrange.
For FISH validation analysis, 3 human breast tumour cell lines were used: MCF7
(Dr. Shirley Wu, University of Toronto), MDA-MB-231(Dr. Reilly, UHN) and MDA-
MB-231-H2N (Dr. Kerbel, Sunnybrook Hospital). Two 25-mL culture flasks of cells
were prepared for each cell line. Cells were cultured until 100% confluency. 3 hours prior
to FISH analysis, the cell culture media was removed from 1 of the 2 flasks in each cell
line. 3mL of 5-ALA at a concentration of 3mM was added and incubated for 3 hours.
After 3 hours, all 6 flasks were immediately sent to the FISH technician for fixation.
Immunostaining was performed using an IgG antibody conjugated with standard
SpectrumOrange dye used in FISH analysis. After staining, FISH signal in each flask was
counted and compared between 5-ALA+ and 5-ALA- flasks in each cell line.
The three cell lines chosen for FISH testing were selected for their relevant to
clinical FISH diagnosis in breast cancer. In breast cancer application/ diagnosis, FISH is
used to determine HER2 expression. Naturally, the HER2+ transfected cell line MDA-
MB-231-H2N was selected. To compare this cell line with a HER2- line, its
untransfected counterpart MDA-MB-231 was used. MCF7 cells were also used as an
additional HER2- line. Images obtained from FISH analysis are shown in Figure 13. Each
green and orange spot in the image represents an antibody attachment, and the same cell
- 69 -
line should have the same number of attachments (in each colour). As shown, there was
no difference between the cells that were incubated with 5-ALA and those that were not,
suggesting that the administration of 5-ALA pre-analysis would not affect the FISH test
result.
The FISH staining results also demonstrated that administration of 5-ALA
into the cell culture would not interfere with the fluorescence signature obtained
in FISH diagnostic testing. In my experiment, the antibodies used were against
human B cell leutkemia 2 (Bcl2) and Immunoglobin H (IgH). Bcl2 was tagged
with SpectrumOrange and IgH was tagged with SpectrumGreen. Although
Her2/neu antibodies were not used, the results still demonstrated a proof-of-
Figure 16 – FISH images for each respective cell line. The number of antibody stains remained constant between the 5-ALA+ sample and the 5-ALA- sample. MCF7/ALA- image was re-processed by the technician who performed FISH analysis because this image did not have enough contrast for unknown reasons. However, the number of antibody stains in MCF7/ALA- image remained constant before and after processing.
- 70 -
principle concept that PpIX fluorescence was no longer observed at the time of
FISH analysis. In fact, PpIX fluorescence was absent after the fixation prior to
FISH, indicating that PpIX fluorescence would not interfere with FISH regardless
of the antibodies used.
7.3. Institutional and Governmental Approval
In order to use 5-ALA as a contrast agent in a clinical setting, the existing clinical
trial protocol needed to be revised to include 5-ALA, and the new protocol needed to be
approved by both Health Canada and by the REB at UHN. As described in section 3.2,
the clinical-grade drug for human application is manufactured in Germany, and the
manufacturer has not yet reached a consensus with UHN. Thus the approval from Health
Canada has not been obtained, and no human trials can be conducted as of yet.
REB amendments were submitted to UHN REB for approval. Details about these
amendments were described in section 4.1.
The addition of in vivo animal models involved another administrative department
in UHN – the Animal Resources Centre (ARC). ARC required every experiment
involving animals to have a properly approved protocol (Animal Use Protocol, AUP). In
this protocol, details about the experiments are outlined, including:
• The types and numbers of animals involved
• Justification for using the requested animals (instead of a cell line, or an animal of
a lower class)
- 71 -
• Specific procedures for the experiments
• The expected endpoint of the experiment, and the expected endpoint for the
animals.
• The expected contribution that this set of experiments is to bring into the field
AUPs must be drafted and submitted online by the principle investigator, and are
reviewed during monthly ARC meetings. AUPs are approved provisionally if there are
minor issues with the application. Provisional approvals are typically 2 months long, and
any issues identified are expected to be rectified by the end of the provisional approval
duration. If the application is accepted, an official approval is given. Approved AUPs are
assigned an AUP number, and any experiments involving animals are required to
reference the associated AUP number.
For my experiments, one AUP was drafted by me and submitted. Initially, it
received a provisional approval because in this feasibility test, the drug dosage to be used
was uncertain. ARC expected a preliminary test within the provisional approval period.
Once we had a definite drug dosage, the application was updated and an official approval
was received.
7.4. In Vitro Feasibility Test
The first stage in the feasibility test was to test if 5-ALA-induced-PpIX
mechanism works in vitro. This stage aimed to determine:
- 72 -
• if breast cancer cells incubated with 5-ALA would produce PpIX fluorescence
detectable by our imaging device.
• if there was a significant signal difference between 5-ALA-incubated breast
cancer cells and non-ALA-incubated breast cancer cells.
• when was the optimal time point the capture images that would produce the
highest amount of image contrast between PpIX+ and PpIX- cells.
7.4.1. Materials and Methods
7.4.1.1. Cell Culture
Three cell lines were used for in vitro test: MCF7 (HER2-, ER+, provided by Dr.
Shirley Wu, University of Toronto), SK-BR-3 (HER2+, ER-, provided by Dr
Muthuswamy, Princess Margaret Hospital) and MDA-MB-231-H2N (HER2+, ER-,
provided by Dr Kerbel, Sunnybrook Hospital). MDA-MB-231-H2N is a standard MDA-
MB-231 line transfected with HER2+ gene. MCF7 cells were cultured in DMEM
supplemented with 10% FBS. SK-Br-3 and MDA-MB-231-H2N cells were cultured in
RPMI supplemented with 10% FBS and 1% insulin. Cells were cultured in 25 mL cell
culturing flasks and passaged at 75-100% confluency.
7.4.1.2. 5-ALA Incubation
Three trials of 5-ALA in vitro imaging were performed in total. In each trial, the
preliminary preparations were completed in the same manner as outlined below:
- 73 -
48 hours prior to 5-ALA incubation, cells were trypsinated from 25 mL culture
flask and transferred onto an 8-well chambered coverglass at approximately 10%
confluency. After 48 hours, cells grew to approximately 50% confluency.
One hour prior to imaging, the culture media was aspirated from the bottom 4
wells (ALA-treated group) and 400 µL of 5-ALA dissolved in culture media was added
at a concentration of 500 µg/mL. Cells were incubated with 5-ALA for 1 hour, then
transferred to the fluorescence microscope for imaging.
In trial 1, MCF7 and SK-BR-3 cells were used. These cells were incubated for 6
and 8 hours. Microscopy images were taken prior to any incubation with ALA. After this
“pre-incubation” time point had been imaged, all 8 wells were incubated with 5-ALA. At
6 hours, 5-ALA from the top 4 wells were aspirated and replaced with cell culture media.
At 8 hours, the bottom 4 wells had their 5-ALA replaced with culture media.
Similarly in trial 2, MCF7 and SK-BR-3 cells were incubated for 3 and 6 hours.
Top 4 wells were aspirated at 3 hours and bottom 4 wells were aspirated at 6 hours.
In trial 3, MDA-MB-231-H2N cells were used. To obtain time-lapsed images, 4
wells were incubated with 5-ALA initially. At each hour post-incubation, cells were
removed from the incubator and imaged quickly, then returned to the incubator. Cells
were imaged over 6 hours, and discarded at 6 hours after imaging.
- 74 -
7.4.1.3. In Vitro Fluorescence Microscopy
An AxioObserver system (Zeiss Ltd) was used for all in vitro fluorescence
microscopy. Since the on-stage incubator for live-cell imaging was malfunctioning, cells
were kept in the incubator next to the microscope system between imaging time points.
An 5-ALA/PpIX-specific filtering cube (excitation 405 nm, emission 600 nm LP) was
used for all imaging sessions. Three channels were used for imaging: white-light (phase-
contrast microscopy), autofluorescence (excitation 488 nm, emission 500 nm LP, signal
from elastin/collagen), and 5-ALA/PpIX. The focus of the microscope was grossly
adjusted under WL, then fine-tuned under autofluorescence before imaging in all 3
channels. The 5-ALA/PpIX channel was minimally used in focusing due to rapid
photobleaching of PpIX in vitro. All images were taken under 40x objective at an
exposure time of 1000 ms and a gain of 1. For each time point, 3-5 images were taken
from the same location, under the same channel. These images were considered a set.
7.4.1.4. Data Analysis
Images were analyzed using ImageJ. The AF image in each image set was used as
calibration when comparing fluorescence intensities between different 5-ALA
fluorescence images. Average image intensity in each 5-ALA image was measured using
ImageJ, and normalized against the average image intensity in the corresponding AF
image, then compared across all 5-ALA images. In each image set (images taken from
- 75 -
the same time point), the normalized values were averaged to produce the final value for
this set.
7.4.2. Results
In trial 1 (for in vitro fluorescence microscopy), two cell lines were used: MCF7
and SK-BR-3. SK-BR-3 was incubated with 5-ALA for 6 hours prior to imaging, then
incubated for a further 2 hours for a second set of images. Qualitatively, AF signal
(natural cellular signal for calibration purposes) was observed in all 3 image sets (control,
6hr incubation, 8hr incubation), but 5-ALA-induced-PpIX signal was only observed in
the 6hr and 8hr image sets (Figure 14).
Figure 17 – SK-BR-3 imaging using AxioObserver (40x magnification). Autofluorescence channel was used to calibrate between different images. The levels of FL intensity in AF were normalized across different image sets, and the corresponding, adjusted PpIX fluorescence were compared. Cells incubated with 5-ALA produced bright red fluorescence due to 5-ALA-induced PpIX. Control cells without 5-ALA did not produce any fluorescence in the PpIX channel, suggesting that 5-ALA could
- 76 -
increase breast tumor-to-normal fluorescence contrast in vivo. 5-ALA-induced PpIX fluorescence can be detected using the existing filter settings on the prototype device.
MCF7 cells were also incubated with 5-ALA for 6 hours and 8 hours. However,
the data collected from 6 hours were inaccurate due to extreme photobleaching. The cells
used in the 8-hour time point were different, and were only exposed to 405 nm excitation
light during imaging. AF channel was used to focus the image instead. In doing so,
photobleaching in the 8-hour cells was minimized, and the data obtained were usable.
Similar to the results obtained from SK-BR-3 cells, there was uniform AF signal from
both control and 5-ALA-incubated cells, but only 5-ALA incubated cells had PpIX signal
(Figure 15).
Figure 18 –MCF7 imaging using AxioObserver (40x magnification). Similar to Figure 14, AF channel was used to calibrate between different images. In this cell line, 5-ALA-induced-PpIX fluorescence was only visible in incubated cells, and was absent in control cells. The WL image for 8 hour timepoint was lost due to image corruption during saving.
6 h 8 h
- 77 -
SK-BR-3 control cells had an average intensity of 332.4±4.7 (arbitrary
unit, AU) after normalization. At 6 hours, the 5-ALA-induced-PpIX FL had an
average intensity of 968±71 AU, and at 8 hours, the average intensity was
830±103 AU. MCF7 control cells had an average intensity of 336.5±3.8 AU. At 8
hours, the 5-ALA-induced-PpIX FL had an average intensity of 620±67 AU.
Quantitative comparisons between SK-BR-3 and MCF7 at various time points in
trial 1 are shown in Figure 16.
Figure 19 – Quantitative comparison of average PpIX FL intensity between MCF7 cells and SK-BR-3 cells at various time points. Both control cells had comparable fluorescence intensity. SK-BR-3 had a higher FL intensity than MCF7 cells at 8 hours, and SK-BR-3 had a higher FL intensity at 6 hours than at 8 hours. Each error bar indicates one standard deviation (N = 5).
Data from trial 1 showed that 5-ALA was able to induce PpIX fluorescence in
breast cancer cells SK-BR-3 and MCF7, and the fluorescence signal was visible under
device filters. However, there was a decline in fluorescence intensity between 6 hours
and 8 hours post-incubation; suggesting that the optimal intensity time point was equal or
- 78 -
prior to 6 hours. Based on these conclusions, the experiment was repeated with the same
cell lines, but at earlier time points. In trial 2, both cells were imaged at 3 hours and 6
hours post-incubation. Fluorescence microscopy images are shown in Figure 17 and
Figure 18.
Figure 20 – SK-BR-3 breast cancer cells incubated with 5-ALA produced PpIX fluorescence at both 3 hours and 6 hours. 5-ALA- cells did not produce any PpIX fluorescence. The AF channel was used similarly to trial 1, as a calibration channel to compare the 5-ALA intensity across different images. (40x magnification)
- 79 -
Figure 21 – MCF7 breast cancer cells incubated with 5-ALA produced PpIX fluorescence at both 3 hours and 6 hours. 5-ALA- cells did not produce any PpIX fluorescence. The AF channel was again used as a calibration. (40x magnification)
Following the confirmation from these two trials that 5-ALA-induced-PpIX
fluorescence was compatible with breast cancer cells, an in vivo trial in a xenograft model
was attempted. Upon initial inoculation, it was apparent that neither cell lines used in the
first two trials were tumourgenic. A third, tumourgenic cell line MDA-MB-231-H2N was
obtained with courtesy to Dr Kerbel’s lab (University of Toronto). Since this was a new
cell line, a third in vitro trial was performed to confirm that 5-ALA-induced-PpIX
fluorescence was compatible with this cell line.
- 80 -
In this trial, a systematic imaging procedure was used. Cells were imaged at 1
hour intervals for up to 6 hours post-incubation. (Figure 19)
Figure 22 – MDA-MB-231-H2N cells incubated with 5-ALA displayed an increase in PpIX fluorescence over time from 1 hour to 6 hour post-incubation. PpIX fluorescence signal could be observed starting at 1h post-incubation. (40x magnification)
- 81 -
A quantitative analysis showed that MDA-MB-231-H2N displayed an increasing
trend over time. The normalized average fluorescence at each time point was 0.38±0.05,
0.26±0.07, 0.43±0.11, 0.53±0.17, 0.50±0.01 and 0.71±0.04 AU. A linear trend was used
to fit the data, with an R-squared value of 0.75 (Figure 20)
Figure 23 – Normalized fluorescence intensity over time in MDA-MB-231-H2N. There was a general increase in fluorescence over the period of 6 hours when the 5-ALA-induced-PpIX fluorescence was normalized against the corresponding autofluorescence intensity. A linear fit had an R-square value of 0.75 (75% “fit” to the data). Each error bar indicates one standard deviation (N=3).
7.4.3. Discussion and Conclusion
All three in vitro trials demonstrated that 5-ALA-induced-PpIX fluorescence was
compatible with breast cancer cells. While all the images presented here were obtained
using the a microscope, the wavelengths used in the microscope filter (405 nm excitation
and 630 nm emission) were identical to those used in PRODIGI™, signifying that that
the signal generated by PpIX fluorescence was also detectable by PRODIGI™ .
- 82 -
In trial 1 and 2, 5-ALA produced PpIX when incubated with breast cancer cells
SK-BR-3 and MCF7. In both cell types, PpIX had a comparable increase in fluorescence
contrast compared to control cells. This validated that 5-ALA-induced-PpIX enhanced
the fluorescence contrast in breast cancer cells regardless of the cancer type, making it a
suitable candidate in vivo contrast agent. Between the 2 trials, the optimal incubation time
for SK-BR-3 and MCF7 cells was determined to be before 6 hours. This is in agreement
with the published literature [68][69][70].
Trial 3 demonstrated that 5-ALA-induced-PpIX fluorescence was also compatible
with the cell line that would be used in vivo. The PpIX signal was already visible at as
early as 1 hour post-incubation with this new cell line MDA-MB-231-H2N, and the
signal intensity increased over time up until the last time point, suggesting that the
prototype device would be able to observe a PpIX signal within the first 6 hours post
incubation. With these positive results, the device was ready to move into in vivo
validation.
7.5. In Vivo Feasibility Trial
Moving from in vitro to in vivo, a second feasibility test was completed in a murine
xenograft model to validate the results I obtained from the in vitro test.
7.5.1. Materials and Methods
- 83 -
7.5.1.1. Xenograft Model
Female nude mice (NCRNU-F, Taconic.) were used and kept in the PMH
research animal facility. Two different cell lines were used for the experiments. In trial 1
and 2, MDA-MB-231-H2N cells were used for tumourgenesis. The inoculation process
was the same both trials. Briefly, cells were cultured as described in 5.3.1.1 and harvested
at a concentration of 1 x 10-6 / mL. 100 uL of cells were injected into the right (trial 1) or
left (trial 2) flank of each mouse subcutaneously (in matrigel). Tumours were allowed to
grow until a diameter of 0.8-1 cm before images were taken. 24 hours prior to imaging,
mice were switched to a diet with chlorophyll-free food to minimize unwanted
autofluorescence.
On the day of imaging, each experimental mouse was injected through tail-vein
injection with 200 uL of 5-ALA at a concentration of 2.5 mg/mL. Each control mouse
was injected with 200 uL of saline.
In trial 3, a window chamber model was adopted. A window chamber was
surgically installed onto the dorsal skin fold of a nude mouse. 100 uL of MCF7-EGFP
(enhanced green fluorescence protein) cells were injected at a concentration of 1 x 10-6 /
mL into the centre of the window chamber. Trial 3 was conducted at 14 days post tumour
cell injection.
7.5.1.2. Maestro Imaging
- 84 -
A Maestro system (CRi, Caliper Life Science, Inc) was used to image mice in trial
1. This is a multispectral fluorescence image system used for small animal in vivo
imaging. Mice were anesthetized with isoflurane at t = 0, 1, 2, 3 hours post-injection.
Each mouse was placed in the Maestro and imaged at two zooms: full body and zoomed
at the left flank. At each zoom, WL and FL (excitation: 503 - 555 nm, emission: 580 nm
LP) images were taken. WL images were exposed for 20 ms, while FL images were
exposed for 5000 ms. Images were analyzed using Maestro’s commercial software to
isolate the PpIX signal from skin autofluorescence. Briefly, Maestro was able to
spectrally “unmix” signal data from background data using a proprietary algorithm. After
a pre-injection image was taken, the entire area of the mouse was selected and labelled as
“background”. The spectrum from this “background” was then calculated and stored in a
library. Any subsequent images (t=1, 2, 3 hours post-injection) were compared to the
library data, and the difference in signal (PpIX signal) was separated and mapped into a
new image for analysis.
7.5.1.3. PRODIGI™ Imaging
PRODIGI™ was used to image mice in trial 1 and 2. In trial 1, PRODIGI™
images were taken concurrently with Maestro images. After 3 hours post-injection, mice
were sacrificed through ketamine/xylazine overdose to stop the 5-ALA-to-PpIX
conversion, and placed on a black mat to reduce background fluorescence. Each mouse
was imaged using PRODIGI™ under both WL and FL (excitation: 405 nm, emission:
600 nm LP). Images were taken for several cases:
• Tumour with skin
- 85 -
• Tumour without skin
• Tumour bed
• Tumour in a Petri Dish
• Organs (kidney, heart, lungs, spleen, and muscle sample)
In trial 2, mice were directly sacrificed at 3 hours post-injection. The remaining
imaging procedure was the same as in trial 1 with the difference that emission filter was
changed to 550-650 nm BP. A total of 10 mice were imaged between the two trials.
7.5.1.4. AxioObserver Imaging
AxioObserver was used to image mice in trial 3. AxioObserver is a fluorescence
microscope with a mini incubator on-stage that allows live small animal imaging. This
microscope was chosen to image the window chamber mice due to its high sensitivity,
which may allow it to visualize PpIX signals that were not visible by eye. Mice were
anesthetized with ketamine/xylazine mixture at t=0, 1, 2, 3 post-injection. In each mouse,
the window chamber was fixed onto the microscope stage using a specialized stage and
the mouse was covered physically to prevent heat loss and unnecessary light exposure to
the rest of the body. 3 images were taken at each data point: phase contrast, green
fluorescence protein (GFP) FL (excitation: 470 nm, emission: 525 nm), PpIX FL
(excitation: 405 nm, emission: 610 nm LP).
7.5.1.5. Point Spectroscopy
- 86 -
Point spectroscopy was used in trial 1 and 2. The device used was a measuring
probe from OceanOptics coupled with a lab-made light source. Point spectra were taken
from a 5-mm-diameter area of interest (excitation: 405 nm, emission: adjustable band-
pass filter adjusted according to background signal). Each measurement consisted of 5
500-ms collections averaged. Prior to each measurement session, the detection probe was
calibrated using a background area to eliminate any external signal from the environment.
When taking measurements, the image probe was directly in contact with the area of
interest. An area of interest was selected by first screening the entire animal with the
prototype device. An area representing the typical signal of its immediate surroundings
was selected to represent this particular signal (i.e., the centre of an area that displayed a
“red” signal was selected to measure the point spectrum of this “red” signal). Any small
areas that displayed a significantly different signal could also be chosen to be measured.
7.5.1.6. Data Analysis
Maestro’s commercial software was able to measure fluorescence intensity of the
images it took. PRODIGI™ images were not directly quantitatively analyzed. Instead,
quantitative data were obtained using point spectroscopy. AxioObserver images were
analyzed in ImageJ. In trial 3, the tumour cells used were MCF7 transfected with the GFP
gene. Thus, the entire tumour possessed GFP fluorescence. When analyzing these
tumours in ImageJ, the tumour area was first selected using the GFP channel. The
selection was then superimposed onto the PpIX channel, and the average pixel intensity
was measured.
- 87 -
7.5.2. Results
In trial 1, mice were first imaged using Maestro, sacrificed and imaged using
PRODIGI™. Maestro was able to detect an increase in whole-body fluorescence in all 5-
ALA-injected mice compared to control at as early as 1 hour post-injection (Figure 21).
This increase was not observed in control mice. In some cases, Maestro was able to
detect strong FL signals in certain organs if these signals were intense enough to be
detected from under the skin. In both injected and control mice, there was a gradual
accumulation of fluorescence in the liver (primarily) and kidney (in certain mice).
a)
4 cm
- 88 -
Figure 24 – Maestro images of a) an 5-ALA-injected mouse and b) a control mouse. White light images are at the top, and their corresponding fluorescence images are at the bottom. Mouse was facing down, with the tumour on its right flank (indicated by an arrow in WL image). In the 5-ALA-injected mouse, there was a visible increase in FL over time, with a more dramatic increase in the liver (indicated by an arrow in FL image). In the control mouse, the whole body FL signal was similar at all time points. Liver had a higher signal than surrounding skin (indicated by arrows in FL images).
Figure 25 – Whole body mouse imaging with Maestro. 5-ALA-injected mice displayed a general increase in FL intensity over time, while control mice showed no change in FL intensity over time. Each error bar indicates one standard error (N=3).
b)
- 89 -
PRODIGI™ was not able to detect any PpIX signal when the skin was intact. The
entire body had a similar, intense AF signal over all time points. (Figure 23)
Figure 26 – Whole body mouse imaging with PRODIGI™ of a) an 5-ALA-injected mouse and b) a control mouse. Tumour (indicated by an arrow in WL) was not visibly different from skin AF in the FL channel. There was no visible increase in FL over time in either a) or b). This either indicated that 5-ALA-induced-PpIX mechanism did not work in in vivo xenograft model, or that the signal was too weak to be detected by PRODIGI™ or by eye.
When skin was removed from tumours, PRODIGI™ was able to detect a different
AF signal from the muscle underneath. However, PpIX signal was not detected
throughout the tumour as expected. Instead, only 1 of the 3 injected mice had a small dot
(1 mm in diameter) of bright FL (Figure 24). When tumours were sliced open, there was
a)
b)
4 cm
4 cm
- 90 -
no visible difference between control tumours and 5-ALA-injected tumours. There was
no visible PpIX signal on the tumour bed (Figure 25).
Figure 27 – Tumours from control and injected mice. In control mouse there was a faint AF signal throughout the tumour; in 5-ALA-injected mouse #1 (ALA+ #1), there was a spot of bright FL (indicated by arrow) on the surface of the tumour. However, when both tumours were sliced, there was no visible difference in FL signal between control and injected tumours.
1 cm
- 91 -
Figure 28 – Tumour (without skin) and tumour bed imaging using PRODIGI™. Tumour location is circled in blue in FL channel. In both cases, there was no visible signal from the tumour that was different from the surrounding muscle signal in either injected or control mouse. Skin AF was much brighter than either tumour FL or muscle AF.
There was no significant difference in FL intensity between organs from injected
mice and control mice (Figure 26). In both cases, kidney and liver showed the highest
amount of FL. There are traces of FL signal in the spleen, and no visible FL signal in
heart or muscle. Results from pixel intensity analysis with ImageJ agreed with results
from visual inspection (Figure 27). Interestingly, in all organs, control mice had a higher
FL intensity than injected mice.
4 cm
- 92 -
Figure 29 – Organs from control (left) and injected (right) mice. There was no visible difference in FL signal between the two. In both cases, kidney and liver had the strongest singles, while the remaining samples had no visible signals.
Figure 30 – Quantitative analysis of organs and muscle FL signal taken by PRODIGI™. In all cases, the controls had a stronger signal, although both were weak. Kidney and liver had significantly higher signals than heart, spleen and muscle. This was in agreement with the observations made directly from the images. Values on the y-axis indicate the average pixel intensity. Each error bar indicates one standard deviation.
2cm
- 93 -
Point spectroscopy (Figure 28) revealed a small peak (635 nm) at the expected
PpIX emission range (630 nm) in 5-ALA-injected mice. This peak was not visible in
control mice. In both cases, the dominating signal was at 500 nm, with an intensity about
25 fold that of the 635 nm peak.
Figure 31 – Point spectra of xenograft tumours. In both curves, the max peak was at around 500 nm. Control tumour had a secondary peak at 600 nm. 5-ALA-injected tumour had a secondary peak at 635 nm (arrow), which was in agreement with PpIX emission peak. This indicated that if the signal intensity reaches the detection threshold of the camera in our prototype device, PpIX signal could be picked up by the camera in its current filter settings. Peak amplitudes were normalized.
The purpose of trial 2 was to focus on PRODIGI™ imaging, thus Maestro was no
longer used. Since our prototype device was not able to detect tumour signals when the
skin was intact, mice were sacrificed directly at the 3 hour point (recommended time-
point in clinical usage [71]) and imaged with PRODIGI™. In trial 1, there was an
interference of red AF signals in the PpIX channel; thus in trial 2, the emission filter was
changed to give the AF signal a different colour (green) from PpIX (red). This helped in
visualizing PpIX signal independent of the background AF.
- 94 -
In this trial, PpIX signals could be directly visualized with the prototype device.
In 5-ALA-injected mice, there was a strong red FL signal over the entire skin area. The
same red signal was absent from control mice (Figure 29). This was a clear indication
that 5-ALA-induced-PpIX FL pathway was functional in xenograft model, producing a
visible red signal in injected mouse. This signal was present throughout the mice and did
not appear to be tumour-specific.
Figure 32 – PRODIGI™ images of 5-ALA-injected (top) and control (bottom) mice. Tumour was circled in red. There was visible skin red FL in the injected mouse, and no such signals in the control mouse. However, there was no visible FL in the tumour in either case. This indicated that while PpIX was being produced in mice, it was not produced specifically in tumours.
2cm
2cm
- 95 -
Tumour bed and organs showed a similar result as trial 1. There was no visible
signal in the tumour beds of injected or control mice. In organs, both injected and control
mice had PpIX FL signal in liver and kidney.
Point spectra were taken from tumour cross-section of all 5 mice imaged in this
trial. Similar to trial 1, all 3 injected mice showed a small peak at 635 nm, with relative
amplitudes much smaller than the dominant peak at 500 nm. In control tumours, there
was no peak at 635 nm (Figure 30). As a proof of principle I also measured several areas
of from the same mouse: skin, tumour bed, tumour cross section (bright and dark spots),
and one organ. The point spectra look very similar with varying amplitudes (Figure 31).
There was a characteristic peak at 635 nm throughout all the areas, indicating that PpIX
was present throughout the mouse, but in varying amounts.
Figure 33 – Tumour cross section fluorescence from all 5 mice in trial 2. All 3 5-ALA-injected mice had a small secondary peak at around 635 nm (arrow), coinciding with PpIX emission peak. This peak was absent in both control mice. This indicated that 5-ALA-induced-PpIX was a viable mechanism in xenograft model, producing a detectable signal (by point spectroscopy).
- 96 -
Figure 34– PpIX signal in various components in a single injected mouse. All components showed a peak at 635 nm (red dashed line), with varying intensity. This suggested that PpIX was present throughout the mouse, but at varying quantities.
Results from trial 1 and 2 confirmed the presence of PpIX signal in the tumour,
but indicated that the digital camera embedded in the device was not able to directly
visualize PpIX signal in the tumour. To visualize and image PpIX signal in real time,
AxioObserver was chosen as the imaging system because it was able to detect PpIX
signal at a much lower intensity. In this trial, window chamber model was selected to
better accommodate the imaging stage of AxioObserver.
Three mice were used for this trial. 2 were injected with 5-ALA while 1 served as
control. Under the microscope, there was no visible difference between all 4 time points
- 97 -
(pre-injection, 1, 2, 3 hours post-injection) in control or injected mice (Figure 32).
Tumour appeared bright green under GFP channel while there was very little background
noise. Tumour border was clearly visible. Under PpIX channel, tumour also had a signal
even at pre-injection, and this signal maintained its intensity over time. There was
relative low contrast between tumour signal and background noise. When analyzed
quantitatively, both control and injected tumours showed a similar level of fluorescence
intensity (after normalization) over time, indicating no additional PpIX signal in the
injected tumour (Figure 33). To eliminate background influence, all data points were
normalized against pre-injection and control data points. Prior to normalization, in all 3
mice the absolute pixel intensity as measured by ImageJ showed a decrease over time,
suggesting that both GFP and PpIX signals (if any) were photobleached over time.
a)
- 98 -
Figure 35 – AxioObserver images for a) an 5-ALA-injected window chamber mouse and b) a control window chamber mouse. Both tumours showed a similar GFP signal. Signal was detected in the PpIX channel in control tumour as well as in 5-ALA-injected tumour pre-injection, indicating a flood of GFP signal in the PpIX channel. Under this constant background signal, PpIX signal was not detected in the injected tumour over time. The FL intensity in injected tumour remained relatively constant visually.
Figure 36– PpIX FL signal comparison in all 3 mice as measured by ImageJ. This analysis aimed to visualize any difference in trend between control and injected mice. All data points from the same mouse were normalized against pre-injection intensity. Between mice, the two 5-ALA-injected mice points were normalized against control to give a common starting point. The general trends for all 3 lines were consistent, indicating that the signals were similar between control and injected mice. If injected mice showed an increase in FL intensity over time, the normalized graph would display a positive slope much greater than one from the control mouse.
b)
- 99 -
7.5.3. Discussion and Conclusion
It was unclear why liver and kidney accumulated FL signal over time in both
control and injected mice. In injected mice, this phenomenon could have been attributed
to PpIX being removed from the system over time. In control mice, this phenomenon
may have occurred from impurities in food. Although purified food was already fed 24
hours prior to imaging, there still may be impurity that could cause autofluorescence in
the PpIX channel.
In the first trial, the skin AF signal was in the same channel as PpIX channel,
interfering with the detection of PpIX signal. The lack of PpIX signal observed in this
trial could be attributed to several reasons, some of which are not engineering-related
issues. First, the PpIX signal might have been very weak compared to skin AF, and was
masked by the AF signal. Maestro data indicated that there was a PpIX signal present in
the injected mice, but it was difficult to isolate the signal against a background signal of a
similar wavelength. To alleviate this issue, in trial 2, a different filter setting was used for
AF signal. This proved to be much better setting for PpIX-detection. In the second trial,
PpIX signal was readily observable using the prototype device because skin AF was
moved into the green channel, which was distinctively different from PpIX’s red signal.
Injected mice in this trial had a unique red FL signal that was absent in all control mice,
indicating the presence of PpIX. However, in both trials, PpIX signal showed no tumour-
specificity. In particular, the tumour signal was extremely weak to the point that it was
- 100 -
hardly visible under the prototype device. The majority of PpIX signal was observed on
the skin instead. The dominance of signal in the skin, however, may not be an issue in the
actual application of the device on human trials in the OR, because the device is expected
to operate in the surgical field where skin would have already been removed from the
field of view. It is more critical to enhance the tumour signal for clinical application. If
the camera is linear, tumour signal may be more apparent; but the intensity difference
between tumour and skin signals is likely also amplified, making the tumour signal
difficult to visualize when compared against background skin signal. Intra-operatively,
with the skin removed, a linear imaging system may benefit from the increased
concentration-to-signal response compared to a logarithmic system.
Point spectroscopy was able to detect a signal in the tumour, indicating that PpIX
was present and that 5-ALA-to-PpIX conversion mechanism was functional in the
tumour cells, but at a concentration of 5-ALA that may be too low for the camera or
human inspection. Comparing my protocol to published literature, the concentration used
in the literature was much higher ( >200 mg/kg) [72]. The concentration used in my
experiment was selected for clinical relevance, since the clinical applicable concentration
in humans was 20 mg/kg. Given that mouse metabolism is much higher than human
metabolism, an increased 5-ALA concentration may be an option worth exploring in the
future to see if a visible tumour signal can be obtained. As presented by existing results,
the prototype device was already able to detect a PpIX signal in vivo when the
concentration of PpIX was high enough. A human trial can be conducted to see if PpIX
- 101 -
concentration would be high enough in vivo in humans at 20mg/kg initial 5-ALA
concentration.
The window chamber model used in trial 3 was set up to allow multiple-dosage
testing on the same mice. The optimal concentration can be found by repeating in vivo
injections at different dosages, and allow PpIX to be flushed from the system between
trials. Unfortunately, the AUP associated with trial 3 was not yet approved at the time of
the experiment to try concentrations other than 20 mg/kg. There may be several
contributing factors to the lack of visible PpIX signal (at 20 mg/kg injection) when
viewed under PRODIGI™. First, MCF7-EGFR cells were used in this model. The
purpose of using a GFP-transfected cell line was to enable me to delineate the tumour
region easily. While MCF7-EGFR successfully achieved this purpose, the GFP signal in
these cells was leaking into the PpIX channel, causing a strong background noise in the
PpIX channel. While this noise was much weaker than the signal in GFP channel, it was
much stronger than the PpIX signal if there was any. This background noise effectively
masked PpIX signal and made it extremely difficult to adjust microscope setting in order
to capture PpIX signal. To complicate the matter, the PpIX signal was naturally weaker
than GFP signal and had the tendency to photobleach rapidly. Thus there was very little
time to optimize the microscope for PpIX signal. This experimental protocol needed to be
improved by using a tumour model that did not have an extremely strong intrinsic FL
signal.
- 102 -
My in vivo results so far validated that 5-ALA-induced-PpIX fluorescence
mechanism was compatible with breast cancer cells, and that a signal was detectable at 3
hours post administration in injected mice. The presence of tumour signals was
confirmed by point spectroscopy, but the intensity of the signals had not reached the
detection threshold of the camera in the prototype device. In the future, trial 3 could be
repeated at different 5-ALA dosages to find the optimal 5-ALA dosage in vivo in a
xenograft model. Ideally, a different cell line should be used for the new experiment,
preferably one with a weaker GFP (or any other indicative fluorescence) signal such that
less to no noise would be leaked into PpIX channel.
8. Summary and Future Work
8.1. Summary of Results
My thesis was aimed to determine if PRODIGI™, the prototype imaging device
that my lab works with was a suitable tool for intra-operative real-time margin detection
for breast cancer. This device has been previously involved in several different clinical
trials. My goal is to adapt it from its current application to the breast cancer application.
The number of incidences of positive tumour margin in current breast cancer
surgeries is alarmingly high, presenting a real need in the medical field for a device that
would be able to delineate tumour from benign tissues intra-operatively. If PRODIGI™
- 103 -
was able to achieve that, the rate of secondary surgeries to remove residual tumours
would be significantly decreased.
As this device has never been used in breast application, I started with some pilot
test to see if it was feasible to use this device on human breast samples. In this pilot test, I
attempted to use PRODIGI™ to differentiate between tumour and normal tissues using
autofluorescence alone. Clinical results showed that PRODIGI™ was able to differentiate
between normal and tumour tissues based on AF alone in ex vivo samples. Normal
adipose and connective tissue produce consistent AF signatures (red and green AF,
respectively). However, tumours produce a range of AF signals depending on the
concentration of blood, connective tissues and level of necrosis. If the tumour signal was
distinctively different from the characteristic adipose and connective tissue signals, the
tumour-normal tissue boundary could be delineated. In other cases, AF did not have
sufficient diagnostic specificity across different tumour subtypes to delineate tumours in
an unknown sample.
To address this limitation, in the next stage of my thesis project, I adopted a
clinical-grade contrast agent, 5-ALA. 5-ALA has been as is used in other human cancer
studies, but was never used in breast cancer when I began this arm of the study. To
ensure that 5-ALA was compatible with breast cancer cells, and would produce PpIX
fluorescence in vivo, I first tested this exogenous prodrug in vitro in several breast cancer
cell lines (MCF7, SK-BR-3, MDA-MB-231-H2N). All cell lines responded to 5-ALA
and produced visible PpIX fluorescence after 3-6 hours of incubation. With proven
- 104 -
compatibility between 5-ALA and breast cancer cell lines, I moved into in vivo validation
experiments. Nude mice were inoculated with previously tested breast cancer cells, and
5-ALA was injected when tumours grew to 1 cm in diameter. Between several imaging
platform used (Maestro, PRODIGI™, AxioObserver), I demonstrated that 5-ALA
induced PpIX signal in vivo in a xenograft tumour models. With the concentration I
tested (20 mg/kg), the PpIX signal was at the strongest in skin, followed by organs such
as liver and kidney. Tumour cells contained PpIX signal, as validated by point
spectroscopy. However, this tumour signal was not strong enough to be visible without
imaging processing using PRODIGI™. My in vivo results validated that 5-ALA-induced-
PpIX fluorescence was viable in breast cancer cells, and could be used to isolate tumour
cells from normal cells. However, further experiments needed to be conducted to
optimize the parameters for 5-ALA administration.
Unfortunately, the linearity validation experiment revealed that the digital camera
used for the prototype device was not linear, but logarithmic. This result was expected of
a commercial consumer digital camera. This observation did not significantly impact the
results from this project, due to the fact that most of the analysis was not strictly
quantitative. However, if future experiments involving robust quantitative analysis were
to be performed, the camera must be set up in a way that would produce a linear
photometric response.
Overall, this project served as a pilot project to adapt PRODIGI™ for breast
cancer clinical uses. I was able to confirm that PRODIGI™ could differentiate between
- 105 -
different breast tissues, and could isolate PpIX fluorescence signal if visible. To continue
this project, there are two areas where additional work could immediately push the
project forward. First, in the clinical trial arm, more patients would need to be recruited to
generate core biopsy and point spectroscopy data sets. These data would help quantify the
AF signal obtained by PRODIGI™, and may uncover a relationship between tumour
composition and its apparent AF signature. In the 5-ALA/PpIX arm, different dosages of
5-ALA need to be tested in vivo to uncover the optimal 5-ALA dosage that would
produce a visible PpIX signal when viewed using PRODIGI™.
8.2. Contribution to the Field
As a pilot study for translating an existing device from its current application into
breast cancer application, this project laid the basic groundwork. In my thesis work, I
completed all the administrative steps required to move the device into the clinical trial
and into the OR. I also tested the device against a variety of specimens, cell lines and
animal models, and confirmed that the device was able detect the expected signals in a
breast sample. Furthermore, I identified areas that needed immediate work, most of
which were biological in nature and were not limitations in the device itself.
8.3. Future Work
In the immediate future, experiments could be conducted to identify the optimal
5-ALA dosage in a murine xenograft model. The model has already been prepared in in
- 106 -
vivo experiment #3 as outlined in this thesis. Additional administration concentrations of
5-ALA needs to be approved in an AUP amendment, and experiment #3 can be repeated
with newly approved dosages. Once this optimal dosage has been identified, the
prototype device PRODIGI™ can be used to visualize PpIX fluorescence signals,
perhaps with different filter settings to see if an optimal tumour-to-normal contrast can be
identified.
The CTA for clinical-grade 5-ALA is expected to be approved by Health Canada
in fall of 2012. Once this approval is received and the drug is delivered to UHN from the
German manufacturer, human clinical trials with 5-ALA can commence. This arm of the
project will attempt to determine the feasibility of using the prototype device as an intra-
operative margin detection tool. Patients can ingest 5-ALA 3-4 hours prior to their
surgery, and the device can be used to screen for PpIX signal intra-operatively. While this
arm has been approved to recruit up to 30 patients, an initial 5-patient cohort should be
recruited to test if PpIX signal is present and intense enough in breast cancer to be
detected by PRODIGI™. The number 5 is chosen because this is the smallest number
where a consistent positive or negative result is statistically significant by student’s T-test
(P < 0.05). Additional tests may need to be performed to optimize route of entry for 5-
ALA (e.g. oral, intra-venous injection) or the administration time.
The long-term goal of this project is to provide the surgeons with a margin-
detection tool that can seamlessly integrate into the surgical environment while providing
real-time, visual feedback about tumour margins. Once the prototype device can
- 107 -
confidently delineate tumour cells from normal cells in breast cancer intra-operatively, it
can be upgraded to allow hands-free operation (e.g. head-mounted projection goggles,
surgical arms). An improved detection sensitivity and spatial resolution can also be
explored by replacing the camera system in the device with one that is more advanced, or
incorporating the capabilities to image microscopically. The future of breast cancer
surgeries could be immensely improved when such technologies were optimized for
intra-operative uses.
- 108 -
9. References
[1] Canadian Cancer Society: Breast Cancer Statistics at a Glance. [Online] http://www.cancer.ca/canada-wide/about%20cancer/cancer%20statistics/stats%20at%20a%20glance/breast%20cancer.aspx
[2] D.T. Ramsay, J.C. Kent, R.A. Hartmann, P.E. Hartman. Anatomy of the lactating human breast redefined with ultrasound imaging. J Anat. 2005 June; 206(6): 525–534.
[3] Wikipedia, the free encyclopedia: File:Breast anatomy normal scheme.png [Online] http://en.wikipedia.org/wiki/File:Breast_anatomy_normal_scheme.png
[4] Women’s Health Zone: Lobes, Lobules and Breast Cancer [Online] http://www.womenshealthzone.net/cancer/breast-cancer/lobes-lobules/
[5] A. Gabriel, J. N. Long, “Breast Anatomy” [Online] http://emedicine.medscape.com/article/1273133-overview#aw2aab6b4
[6] University of Wisconsin School of Medicine and Public Health: Arrangement of Connective Tissue Predicts Breast Cancer Prognosis [online] http://www.med.wisc.edu/news-events/news/arrangement-of-connective-tissue-predicts-breast-cancer-prognosis/30720
[7] PubMed Health: Breast cancer [Online] http://www.ncbi.nlm.nih.gov/pubmedhealth/PMH0001911/
[8] Breastcancer.org: Non-Invasive or Invasive Breast Cancer [Online] http://www.breastcancer.org/symptoms/diagnosis/invasive.jsp
[9] Canadian Cancer Society: Staging and Grading for Breast Cancer [Online] http://www.cancer.ca/canada-wide/about%20cancer/types%20of%20cancer/staging%20and%20grading%20for%20breast%20cancer.aspx?sc_lang=en
[10] M. Banys, A.D Hartkopf, N. Krawczyk, S. Becker, T. Fehm. Clinical implications of the detection of circulating tumor cells in breast cancer patients. Biomark Med. 2012 Feb;6(1):109-18.
- 109 -
[11] M. Gnant, P. Dubsky, P. Hadji. Bisphosphonates: prevention of bone metastases in breast cancer. Recent Results Cancer Res. 2012;192:65-91.
[12] A.G. Garcia, H. Nedev, K. Bijian, J. Su, M.A. Alaoui-Jamali, H.U. Saragovi. Reduced in vivo lung metastasis of a breast cancer cell line after treatment with Herceptin mAb conjugated to chemotherapeutic drugs. Oncogene. 2012 Jul 16.
[13] L. Padovani, X. Muracciole, J. Régis. γ knife radiosurgery of brain metastasis from breast cancer. Prog Neurol Surg. 2012;25:156-62.
[14] M. Mistrangelo, P. Cassoni, M. Mistrangelo, I. Castellano, E. Codognotto, A. Sapino, G. Lamanna, F. Cravero, L. Bianco, G. Fora, S. Sandrucci. Obstructive colon metastases from lobular breast cancer: report of a case and review of the literature. Tumori. 2011 Nov-Dec;97(6):800-4.
[15] American Cancer Society: Learn About Cancer [Online] http://www.cancer.org/Cancer/BreastCancer/OverviewGuide/breast-cancer-overview-survival-rates
[16] M. Nothacker, V. Duda, M. Hahn, M. Warm, F. Degenhardt, H. Madjar, S. Weinbrenner, U.S. Albert. Early detection of breast cancer: benefits and risks of supplemental breast ultrasound in asymptomatic women with mammographically dense breast tissue. A systematic review. BMC Cancer 2009, 20;9:335.
[17] E. Sickles. Mammographic detectability of breast mircrocalcification. AJR 1982, 139:913-918
[18] M.T. Mandelson, N. Oestreicher, P.L. Porter. Breast density as a predictor of mammographic detection: comparison of interval- and screen-detected cancers. J Natl Cancer Inst 2000, 92:1081-1087
[19] V. McCormack, I. dos Santos Silva. Breast density and parenchymal patterns as markers of breast cancer risk: a meta analysis. Cancer Epidemiology Biomarkers and Prevention 2006, 15:1159-1169.
[20] N. Boyd, L. Martin, S. Chavez, A. Gunasekara, A. Salleh, O. Melnichouk, M. Yaffe, C. Friedenreich, S. Minkin, M. Bronskill. Breast-tissue composition and other risk factors for breast cancer in young women: a cross-sectional study. The Lancet Oncology 2009, 10;6:569-580
- 110 -
[21] J. Reiland-Smith. Diagnosis and surgical treatment of breast cancer. S D Med. 2010;Spec No:31-7.
[22] N. Houssami, S. Ciatto, R. M. Turner, H. S. Cody. Preoperative staging of the axilla in women with invasive breast cancer. Breast Cancer Management, May 2012, 1(1):65-72
[23] M.C. Smitt, K. Nowels, R.W. Carlson, S.S Jeffrey. Predictors of reexcision findings and recurrence after breast conservation. Int J Radiat Oncol Biol Phys. 2003 Nov 15;57(4):979-85.
[24] S.J. Schnitt. Risk factors for local recurrence in patients with invasive breast cancer and negative surgical margins of excision. Where are we and where are we going? Am J Clin Pathol. 2003 Oct;120(4):485-8.
[25] J-P.Pignol, B. M. Keller, A. Ravi. Doses to internal organs for various breast radiation techniques - implications on the risk of secondary cancers and cardiomyopathy. Radiation Oncology. 2011, 6:5
[26] Breastcancer.org: When Is Radiation Appropriate? [Online] http://www.breastcancer.org/treatment/radiation/when_appropriate.jsp
[27] American Cancer Society: Learn About Cancer [Online] http://www.cancer.org/Cancer/BreastCancer/DetailedGuide/breast-cancer-treating-chemotherapy
[28] Breastcancer.org: Hormonal Therapy [Online] http://www.breastcancer.org/treatment/hormonal/
[29] E.K. Valdes, S.K. Boolbol, J.M. Cohen, S.M. Feldman. Intra-operative touch preparation cytology; does it have a role in re-excision lumpectomy? Ann Surg Oncol. 2007 Mar;14(3):1045-50.
[30] A. H. Fischer, K. A. Jacobson, J. Rose, R. Zeller. Cryosectioning Tissues. Cold Spring Harb Protoc 2008.
[31] C.S. Kaufman, L. Jacobson, B.A. Bachman, L.B. Kaufman, C. Mahon, L.J. Gambrell, R. Seymour, J. Briscoe, K. Aulisio, A. Cunningham, F. Opstad, N. Schnell, J. Robertson, L. Oliver. Intraoperative digital specimen mammography: rapid, accurate results expedite surgery. Ann Surg Oncol. 2007 Apr;14(4):1478-85.
- 111 -
[32] E.R. Fisher, A.M. Gregonio, B. Fisher. C. Redmond, F. Vellios, S.C. Sommers. The pathology of invasive breast cancer. A syllabus derived from findings of the National Surgical Adjuvant Breast Project (protocol no. 4). Cancer 1975;36: 1-85
[33] C.M. Holloway, A. Easson, J. Escallon, W.L. Leong, M.L. Quan, M. Reedjik, F.C. Wright, D.R. McCready. Technology as a force for improved diagnosis and treatment of breast disease. Can J Surg. 2010 Aug;53(4):268-77.
[34] J.T. McCormick; A.J. Keleher; V.B. Tikhomirov; R.J. Budway; P.F. Caushaj Analysis of the use of specimen mammography in breast conservation therapy The American Journal of Surgery October 2004, 188 (4), pg. 433-436
[35] D.N. Bimston, G.G. Bebb, L.D. Wagman. Is Specimen Mammography Beneficial? Arch Surg. 2000 Sep;135(9):1083-6.
[36] R.M. Moadel. Breast Cancer Imaging Devices Seminars in Nuclear Medicine. May 2011, 41 (3), pg. 229-241
[37] S.J. Kim, J.M. Chang, N. Cho, S.Y. Chung, W. Han, W.K. Moon. Outcome of breast lesions detected at screening ultrasonography. Eur J Radiol. 2012 May 14.
[38] F.D. Rahusen, A.H. Taets van Amerongen, P.J. van Diest, P.J. Borgstein, R.P. Bleichrodt, S. Meijer. Ultrasound-guided lumpectomy of nonpalpable breast cancers: A feasibility study looking at the accuracy of obtained margins. J Surg Oncol. 1999 Oct;72(2):72-6.
[39] R.A. Graham, M.J. Homer, C.J. Sigler, H. Safaii, C.H. Schmid, D.J. Marchant, T.J. Smith. The efficacy of specimen radiography in evaluating the surgical margins of impalpable breast carcinoma. AJR 1994;162:33–36
[40] W. Jung, J. Kim, M. Jeon, E.J. Chaney, C.N. Stewart, S.A Boppart. Handheld optical coherence tomography scanner for primary care diagnostics. IEEE Trans Biomed Eng. 2011 Mar;58(3):741-4.
[41] A. M. Mohs, M. C. Mancini, S. Singhal, J. M. Provenzale, B. Leyland-Jones, M. D. Wang, S. Nie Hand-held Spectroscopic Device for In Vivo and Intraoperative Tumor Detection: Contrast Enhancement, Detection Sensitivity, and Tissue Penetration Analytical Chemistry. November 2010, 82 (21): 9058-9065
- 112 -
[42] R.G. Pleijhuis, G.C. Langhout, W. Helfrich, G. Themelis, A. Sarantopoulos, L.M. Crane, N. J. Harlaar, J.S. de Jong, V. Ntziachristos, G.M. van Dam. Near-infrared fluorescence (NIRF) imaging in breast-conserving surgery: assessing intraoperative techniques in tissue-simulating breast phantoms. Eur J Surg Oncol. 2011 Jan;37(1):32-9.
[43] Health Canada: Application Information - Medical Devices [Online] http://hc-sc.gc.ca/dhp-mps/md-im/applic-demande/index-eng.php
[44] Health Canada: Medical Devices Regulations [Online] http://laws-lois.justice.gc.ca/eng/regulations/SOR-98-282/page-2.html#h-7
[45] Notice of decision for Metvix. Health Canada, Ottawa; 2009.
[46] M. Peterka, I. Klepácek. Light irradiation increases embryotoxicity of photodynamic therapy sensitizers (5-aminolevulinic acid and protoporphyrin IX) in chick embryos. Reprod Toxicol. 2001 Mar-Apr;15(2):111-6.
[47] M. Moisan1, J. Barbeau, M.C. Crevier, J Pelletier, N. Philip, B Saoudi. Plasma sterilization.Methods and mechanisms. Pure Appl.Chem., Vol.74, No.3, pp.349–358, 2002.
[48] Microbiology, Jacquelyn Black, Prentice Hall,1993 pg 334
[49] Guidance for Industry: Device Licence Applications for Ultrasound Diagnostic Systems and Transducers [Online] http://www.hc-sc.gc.ca/dhp-mps/md-im/applic-demande/guide-ld/ultrasound_ultrasons-eng.php
[50] E.W. Gunter, W.E. Turner, D.L. Huff. Investigation of protoporphyrin IX standard materials used in acid-extraction methods, and a proposed correction for the millimolar absorptivity of protoporphyrin IX. Clin Chem. 1989 Aug;35(8):1601-8.
[51] J.C. Mullikin, L.J. van Vliet, H. Netten, F.R. Boddeke, G. van der Feltz, I.T. Young. Methods for CCD Camera Characterization. Proc. SPIE 2173, Image Acquisition and Scientific Imaging Systems, 73. 1994.
[52] Digital Photography Review: Sensor Linearity [Online] http://www.dpreview.com/learn/?/Glossary/Camera_System/sensor_linearity_01.htm
- 113 -
[53] A. Juzeniene; M. Kaliszewski, A. Bugaj, J. Moan. Clearance of protoporphyrin IX induced by 5-aminolevulinic acid from WiDr human colon carcinoma cells. Photodynamic Therapy. 2009 7380:73802Q-73802Q-9
[54] M. Monici. Cell and tissue autofluorescence research and diagnostic applications. Biotechnol Annu Rev. 2005;11:227-56.
[55] G. Tang. A. Pradhan, W. Lam, D. Choy, E. Opher. Fluorescence spectra from cancerous and normal human breast and lung tissues. IEEE Journal of Quantum Electronics, 23(10):1806-1811
[56] Y. Wu, J. Y. Qu. Two-photon autofluorescence spectroscopy and second-harmonic generation of epithelial tissue. Opt Lett. 2005 Nov 15;30(22):3045-7.
[57] A. Douplik, W.L. Leong, A.M. Easson, S. Done, G. Netchev, B.C. Wilson. Feasibility study of autofluorescence mammary ductoscopy. J Biomed Opt. 2009 Jul-Aug;14(4)
[58] S. Kamali, O. Bender, G.H. Kamali, M.T. Aydin, O. Karatepe, E. Yuney. Diagnostic and therapeutic value of ductoscopy in nipple discharge and intraductal proliferations compared with standard methods. Breast Cancer. 2012 Jun 6.
[59] Wikipedia: Methylene Blue [Online] http://http://en.wikipedia.org/wiki/Methylene_blue
[60] A.A. Onitilo, J.M. Engel, R.T. Greenlee, B.N. Mukesh. Breast cancer subtypes based on ER/PR and Her2 expression: comparison of clinicopathologic features and survival. Clin Med Res. 2009 Jun;7(1-2):4-13.
[61] K. Chang, F. Jaffer. Advances in fluorescence imaging of the cardiovascular system. J Nucl Cardiol. 2008 May-Jun;15(3):417-28.
[62] K. Sano, M. Mitsunaga, T. Nakajima, P.L. Choyke, H. Kobayashi. In vivo breast cancer characterization imaging using two monoclonal antibodies activatably labeled with near infrared fluorophores. Breast Cancer Res. 2012 Apr 17;14(2):R61.
[63] M. Wachowska, A. Muchowicz, M. Firczuk, M. Gabrysiak, M. Winiarska, M. Wańczyk, K. Bojarczuk, J. Golab. Aminolevulinic Acid (ALA) as a Prodrug in Photodynamic Therapy of Cancer. Molecules. 2011, 16, 4140-4164
- 114 -
[64] S.R. Millon, J.H. Ostrander, S. Yazdanfar, J.Q. Brown, J.E. Bender, A. Rajeha, N. Ramanujam. Preferential accumulation of 5-aminolevulinic acid-induced protoporphyrin IX in breast cancer: a comprehensive study on six breast cell lines with varying phenotypes. J Biomed Opt. 2010 Jan-Feb;15(1):018002.
[65] D.P. Ladner, R.A. Steiner, J. Allemann, U. Haller, H. Walt. Photodynamic diagnosis of breast tumours after oral application of aminolevulinic acid. Br J Cancer. 2001 Jan 5;84(1):33-7.
[66] R. Sailer, W.S. Strauss, M. Wagner, H. Emmert, H. Schneckenburger. Relation between intracellular location and photodynamic efficacy of 5-aminolevulinic acid-induced protoporphyrin IX in vitro. Comparison between human glioblastoma cells and other cancer cell lines. Photochem Photobiol Sci. 2007 Feb;6(2):145-51. Epub 2006 Nov 21.
[67] S.L. Gibbs, B. Chen, J.A. O'Hara, P.J. Hoopes, T. Hasan, B.W. Pogue. Protoporphyrin IX level correlates with number of mitochondria, but increase in production correlates with tumour cell size. Photochem Photobiol. 2006 Sep-Oct;82(5):1334-41.
[68] A.H. Ali, H. Takizawa, K. Kondo, H. Matsuoka, H. Toba, Y. Nakagawa, K. Kenzaki, S. Sakiyama, S. Kakiuchi, Y. Sekido, S. Sone, A. Tangoku. 5-Aminolevulinic acid-induced fluorescence diagnosis of pleural malignant tumor. Lung Cancer. 2011 Oct;74(1):48-54. Epub 2011 Feb 26.
[69] A. Johansson, G. Palte, O. Schnell, J. C. Tonn, J. Herms, H. Stepp. 5-Aminolevulinic acid-induced protoporphyrin IX levels in tissue of human malignant brain tumors. Photochem Photobiol. 2010 Nov-Dec;86(6):1373-8.
[70] M. Uekusa, K. Omura, Y. Nakajima, S. Hasegawa, H. Harada, K.-I. Morita, H. Tsuda. Uptake and kinetics of 5-aminolevulinic acid in oral squamous cell carcinoma. Int J Oral Maxillofac Surg. 2010 Aug;39(8):802-5. Epub 2010 Apr 20.
[71] European Medicines Agency: Human medicines - EU/3/02/121 [Online] http://www.ema.europa.eu/ema/index.jsp?curl=pages/medicines/human/orphans/2009/11/human_orphan_000324.jsp&mid=WC0b01ac058001d12b
[72] A.M. Dorward, K.S. Fancher, T.M. Duffy, W.G. Beamer, H Walt. Early neoplastic and metastatic mammary tumours of transgenic mice detected by 5-aminolevulinic acid-stimulated protoporphyrin IX accumulation. Br J Cancer. 2005 November 14; 93(10): 1137–1143.