the low-dose limits of lung nodule detectability in

The Low-Dose Limits of Lung Nodule Detectability

in Volumetric Computed Tomography

Jordan Silverman

A thesis submitted in accordance with the requirements

of the degree of Master of Health Science in Clinical Biomedical Engineering,

Institute of Biomaterials and Biomedical Engineering, University of Toronto

Supervised by:

Dr. Jeffery H. Siewerdsen

Department of Medical Biophysics, University of Toronto


© Copyright by Jordan Silverman, 2009

Lung Nodule Detectability in Low-Dose CT

ii

ABSTRACT

Jordan Silverman, MHSc in Clinical Biomedical Engineering


2009

Purpose. Low-dose computed tomography is an important imaging modality for screening and

surveillance of lung cancer. The goal of this study was to determine the extent to which dose

could be minimized while maintaining diagnostic accuracy through knowledgeable selection of

reconstruction techniques.

Methods. An anthropomorphic phantom was imaged on a 320-slice volumetric CT scanner.

Detectability of small solid lung nodules was evaluated as a function of dose, patient size,

reconstruction filter and slice thickness by means of 9-alternative forced-choice observer tests.

Results. Nodule detectability decreased sharply below a threshold dose level due to increased

image noise. For large body habitus, optimal (smooth) filter selection reduced dose by a factor of

~3. Nodule detectability decreased for slice thicknesses larger than the nodule diameter.

Conclusions. Radiation dose can be reduced well below current clinical protocols. Smooth

reconstruction filters and avoidance of large slice thickness permits lower-dose techniques

without tradeoff in diagnostic performance.


iii

To my family and friends

All those who have supported me


iv

ACKNOWLEDGEMENTS

First and foremost I would like to thank my research supervisor, Jeff Siewerdsen. Your support

and guidance throughout this project has afforded me the opportunity to present, publish, and

participate within the scientific community of medical imaging. You have taught me the true

meaning and value of research and opened many doors for my future as medical practitioner and

clinical researcher. I would also like to thank my clinical supervisor Dr. Narinder Paul for

showing me how this research had a true, positive impact on medical practice. The greatest

reward in my work is the clinical applications of these results. The contributions of Toshiba

Medical Systems technicians and engineers, including Noe Hinojosa, Vlad Drachinsky, Henrik

Andrulenas and Chris Porter are gratefully acknowledged. Thanks also to the helpful and

friendly CT staff at Toronto General Hospital and to volunteer observers from the Image Guided

Therapy (IGTx) Lab at the Ontario Cancer Institute. Your kindness is very much appreciated.

Thank you to my fellow students on the “Image Science Team” in Dr. Siewerdsen’s laboratory,

including Daniel Tward, Samuel Richard, and Grace Gang. Lastly, I would like to thank my

friend and colleague Nathaniel Hamming. Nate, your positive yet relaxed attitude has helped to

make my graduate experience in the Clinical Engineering program enjoyable and memorable.

Thank you for everything and I hope we will remain close friends wherever our lives take us.

This work was supported in part by the Radiology Research Fund at TGH, the Natural Sciences

and Engineering Research Council (PGS-M), the Division of Clinical Engineering in the Institute

of Biomaterials and Biomedical Engineering at the University of Toronto, and the National

Institute of Health (R01-CA112163).


v

TABLE OF CONTENTS

CHAPTER I: INTRODUCTION.................................................................................................... 1

1.1 Lung Cancer ................................................................................................................................ 2 1.1.1 Definition and Types................................................................................................................................2 1.1.2 Risk Factors .............................................................................................................................................2 1.1.3 Lung Cancer Staging ...............................................................................................................................3 1.1.4 Early Detection ........................................................................................................................................4

1.2 Dose Characterization in CT ..................................................................................................... 4 1.2.1 The Biological Effect of Radiation...........................................................................................................4 1.2.2 Dose in CT ...............................................................................................................................................5 1.2.3 Obesity and Image Quality ......................................................................................................................6

1.3 Imaging of the Chest and Computed Tomography ................................................................. 7 1.3.1 Chest Radiography to Computed Tomography........................................................................................7 1.3.2 Advances in Multi-Detector (Volumetric) CT ..........................................................................................8 1.3.3 The Need to Minimize Radiation Dose in CT ..........................................................................................8 1.3.4 CT for Early Detection and Surveillance of Lung Nodules ...................................................................11 1.3.5 CT Reconstruction .................................................................................................................................12

1.4 Previous Studies on Low-Dose Lung Nodule Detectability ................................................... 12

CHAPTER II: EXPERIMENTAL METHODS FOR EVALUATION OF LUNG NODULE

DETECTABILITY ....................................................................................................................... 14

2.1 CT Imaging and Reconstruction Techniques ......................................................................... 15 2.1.1 Volumetric CT Scanner – Toshiba Aquilion ONE

TM..............................................................................15

2.1.2 Acquisition and Reconstruction Techniques ..........................................................................................16

2.2 Phantom for Studies of Lung Nodule Detectability ............................................................... 17 2.2.1 Anthropomorphic Phantom with Simulated Lung Nodules....................................................................17 2.2.2 Anthropomorphic Phantom: Average and Obese Body Habitus............................................................19

2.3 Observer Tests and Data Analysis........................................................................................... 21 2.3.1 Observer Performance Test ...................................................................................................................21 2.3.2 Data Analysis.........................................................................................................................................25

CHAPTER III: CHARACTERIZATION OF RADIATION DOSE............................................ 27

3.1 Evolution of Dosimetry in Computed Tomography .............................................................. 28

3.2 Measurements of Dose .............................................................................................................. 30 3.2.1 Dose Reported by the Scanner (CTDIvol.e and DLP.e) ........................................................................30 3.2.2 Dose Measured Using Farmer Chambers .............................................................................................31 3.2.3 Dose Measured Using MOSFETs ..........................................................................................................33

3.3 Dosimetry Experiments ............................................................................................................ 33 3.3.1 Comparison of CTDIvol.e with Farmer Chamber and MOSFET Dose .................................................34 3.3.2 Effect of Beam Position and Scatter on Dose Profile ............................................................................37 3.3.3 Effect of Beam Width on Dose ...............................................................................................................40 3.3.4 Effect of Scan Length on Dose ...............................................................................................................43

3.4 Summary.................................................................................................................................... 47


vi

CHAPTER IV: CHARACTERIZATION OF CT RECONSTRUCTION FILTERS .................. 48

4.1 Reconstruction Filters in CT.................................................................................................... 49

4.2 Wire Phantom Scan .................................................................................................................. 49

4.3 MTF Calculation ....................................................................................................................... 51 4.2.1 The Radon Transform, R(i) ....................................................................................................................52 4.2.2 Determination of the Wire Signal Region ..................................................................................................52 4.2.3 Detrending the Non-Uniform Background ............................................................................................53 4.2.4 Determination of the Signal Profile Centroid ........................................................................................55 4.2.5 Application of Steps 1-4 for all Slices ....................................................................................................55 4.2.6 Align Centroids to Obtain an Over-Sampled LSF .................................................................................56 4.2.7 Fourier transform of the Over-sampled LSF .........................................................................................57

4.4 The MTF Associated with Various CT Reconstruction Filters ............................................ 59

CHAPTER V: LOW-DOSE LIMITS OF LUNG NODULE DETECTABILITY....................... 61

5.1 Experimental Parameters......................................................................................................... 62

5.2 Effect of Body Habitus on Detectability.................................................................................. 62

5.3 Effect of Reconstruction Technique ........................................................................................ 64 5.3.1 Image Quality for Various Reconstruction Filters and Slice Thickness ................................................64 5.3.2 Effect of Reconstruction Filter on Detectability ....................................................................................65 5.3.3 Effect of Reconstruction Slice Thickness on Detectability .....................................................................66 5.3.4 Statistical Comparison of Reconstruction Techniques ..........................................................................67

CHAPTER VI: DISCUSSION, CONCLUSIONS, AND FUTURE DIRECTIONS ................... 71

6.1 Factors Affecting Image Quality: Acquisition, Reconstruction Techniques, and Body

Habitus ................................................................................................................................................... 72

6.2 Optimal Reconstruction Technique Selection ........................................................................ 75

6.3 Limitations of the Current Study ............................................................................................ 77

6.4 Conclusions ................................................................................................................................ 78

APPENDICES .............................................................................................................................. 80

A.1. Observer Test Instructions........................................................................................................... 81

A.2. Calculation of BMI...................................................................................................................... 814

REFERENCES ............................................................................................................................. 89


vii

LIST OF FIGURES

FIG. 2.1. Photograph of the CT scanner.....................................................................................................................15

FIG. 2.2. Anthropomorphic phantom..........................................................................................................................17

FIG. 2.3. MOSFET dosimetry equipment....................................................................................................................18

FIG. 2.4. Anthropomorphic phantom: average vs. obese configuration.....................................................................20

FIG. 2.5. Illustration of 9AFC test for nodule detection .............................................................................................23

FIG. 2.6. Observer performance as a function of dose ...............................................................................................25

FIG. 3.1.Dosimetry experiment setup .........................................................................................................................32

FIG. 3.2. Experimental setup for comparison of measured and reported dose ..........................................................34

FIG. 3.3. Comparison of CTDIvol.e (as reported by the scanner) and Dcenter, or CTDIw ...........................................35

FIG. 3.4. Experimental setup for effect of beam position and scatter on dose profile ................................................37

FIG. 3.5. Measured dose as a function of longitudinal distance from the center of the beam (z-offset).....................39

FIG. 3.6. Experimental setup for effect of beam width on point dose measurement ...................................................41

FIG. 3.7. Effect of beam width on point dose..............................................................................................................42

FIG. 3.8. Experimental setup for measuring the effect of scan length on point dose measurement............................44

FIG. 3.9. Effect of scan length (number of beams) on measured dose........................................................................45

FIG. 3.10. Comparison of measured and reported dose for each beam configuration...............................................46

FIG. 4.1. Thin wire phantom.......................................................................................................................................50

FIG. 4.2. Wire phantom image....................................................................................................................................51

FIG. 4.3. Example 1D Radon transform of the wire image, superimposed over the corresponding ROI...................52

FIG. 4.4. Signal region for MTF calculation ..............................................................................................................53

FIG. 4.5. Quadratic detrend of the line spread function.............................................................................................54

FIG. 4.6. Area under the profile (within the signal region) and determination of centroid position..........................55

FIG. 4.7. Illustration of signal profiles for various image slices ................................................................................56

FIG. 4.8. Over-sampled line spread function..............................................................................................................57

FIG. 4.9. Modulation transfer function for FC1 filter.................................................................................................58

FIG. 4.10. Modulation Transfer Functions for all reconstruction filters in this study................................................59

FIG. 5.1. Observer performance measured as a function of dose for average and obese body habitus.....................63

FIG. 5.2. Example ROIs about a 3.2 mm nodule for each reconstruction filter and slice thickness investigated.......65

FIG. 5.3. Effect of reconstruction filter on detectability and Dthresh ............................................................................66

FIG. 5.4. Effect of slice thickness on Dthresh for average and obese habitus................................................................67

FIG. 5.5. Effect of reconstruction techniques on Dthresh ..............................................................................................68

FIG. 6.1. Images at similar reported and measured dose...........................................................................................73

FIG. 6.2. Axial images of 3.2mm nodule reconstructed at varying tslice ......................................................................76


viii

LIST OF TABLES

Table 1.1. Lung Cancer Staging.....................................................................................................................................3

Table 2.1. Summary of experimental parameters........................................................................................................16

Table 5.1. Summary of p-values from paired t-tests for comparison of reconstruction techniques.............................69


ix

LIST OF ABBREVIATIONS

9AFC Nine-Alternative Forced Choice

AAPM American Association of Physicists in Medicine

AFC Alternative Forced Choice

ALARA As Low As Reasonably Achievable

BMI Body Mass Index

CT Computed Tomography

CTDI Computed Tomography Dose Index

CXR Chest Radiography (X-ray)

FBP Filtered Back-Projection

FC# Filter Convolution # (Code denoting reconstruction filter selection)

HU Hounsfield Units

ICRP International Commission on Radiological Protection\

LNT Linear No-Threshold

LSF Line Spread Function

MDCT Multi-Detector Computed Tomography

MAFC Multiple-Alternative Forced Choice

MOSFET Metal-Oxide Semiconductor Field Effect Transistor

MSAD Multiple Scan Average Dose

MTF Modulation Transfer Function

NSCLC Non-small cell lung cancer

PMMA Polymethyl-Methacrylate

ROC Receiver Operating Characteristic

ROI Region of Interest

SCLC Small cell lung cancer

TLD Thermoluminescent dosimeters


x

LIST OF SYMBOLS

ε dose efficiency

image noise

standard deviation of Dthresh values

standard deviation in detectability vs. dose fit parameter a

fNyquist Nyquist frequency

Az observer performance

observer performance for average body habitus

observer performance for average obese habitus

CTDI∞ limiting equilibrium dose

CTDI100 CTDI measured at center of beam in a 100 mm scan

CTDIw weighted CTDI

CTDIvol.e extended Volume CTDI

Dcenter dose measured at center of cylindrical phantom

Ddetector dose reaching detector

Dlung dose delivered to the lung cavity

Do dose entering cylindrical phantom

Dperiphery dose measured at periphery of cylindrical phantom

Dthresh dose below which detectability falls to Az < 0.95

dose for average habitus below which detectability falls to Az < 0.95

dose to lungs below which detectability falls to Az < 0.95

dose for obese habitus below which detectability falls to Az < 0.95

D(z) dose profile

eµd

attenuation of object

K bandwidth integral

Pcorr proportion correct

R(i) Radon transform

tslice reconstruction slice thickness

threshDσ

aσ

ave

zA

obese

zA

ave

threshD

lung

threshD

obese

threshD

2σ


1

CHAPTER I: INTRODUCTION


2

1.1 Lung Cancer

1.1.1 Definition and Types

Lung cancer is the leading cause of cancer-related death. It kills approximately 1.3 million

people per year worldwide and is responsible for approximately 27% of all cancer deaths in

Canada – more than the combined total of the next three most common cancers (colon, breast

and prostate cancer).1 Lung cancer begins with the uncontrolled division of diseased cells in lung

tissue. It occurs predominantly in the lung epithelium and, like other tumors, is a result of genetic

abnormalities, such as the deactivation or inhibition of tumor suppressor genes or the activation

of proto-oncogenes.2 Lung epithelial tissue is found in the bronchi, bronchioles, and alveoli, the

apparatus of respiratory function. As a tumor continues to grow, it can infiltrate tissues adjacent

to the lungs and can metastasize throughout the body. Lung cancer is broadly characterized as

non-small cell or small-cell lung cancer (NSCLC or SCLC, respectively).3 The former is the

more frequently diagnosed and slower growing of the two, whereas the latter is less prevalent

and tends to grow and metastasize quickly to the lymphatic system and distant organs.3

1.1.2 Risk Factors

Lung cancer can present itself in several ways, including a persistent cough or change in cough,

dyspnea, dysphagia, hemoptysis, chest or shoulder pain, fatigue, weight loss, and swelling of the

face and neck.2 Lung cancer may be attributed to a combination of genetic and environmental

factors. The most common cause of lung cancer is long term exposure to tobacco smoke, which

is responsible for over 80% of lung cancers.4 Even non-smokers with environmental exposure to

tobacco smoke display increased lung cancer incidence. For example, spouses of smokers exhibit

a 30% greater risk of developing the disease than do spouses of non-smokers.4 Other risk factors


3

include genetic factors, increased exposure to Radon gas or asbestos and air pollution.5 It is

important to understand the implications of risk factors so that patients can receive the most

effective form of medical intervention in early stages of disease.

1.1.3 Lung Cancer Staging

Staging in lung cancer has several functions: it characterizes how far the cancer has progressed,

it helps to predict prognosis and therapeutic options, and it serves as a standard of comparison

for the retrospective evaluation of treatment effectiveness. Staging for SCLC is divided into only

two stages, as summarized in Table 1.1. SCLC tends to metastasize early, and doctors often

assume it has spread even if they do not see secondary cancers, although resection of solitary

SCLC has been reported.6 NSCLC exhibits five main stages, with A and B sub-stages, as

summarized in Table 1.1 along with the corresponding 5-year survival rates.6 Note the steep

reduction in 5-year survival beyond Stage I.

Type Stage Sub-

Stage

Description 5-year

Survival

Rate

SCLC Limited n/a Cancer only in one lung and nearby lymph nodes or pleural fluid 15-30%

Extensive n/a Cancer has spread outside of the lungs 0-2%

NSCLC 0 n/a A few cell layers are affected and do not penetrate surface lining 70-80%

I

A

B

The cancer is localized, not found in lymph tissue, and is

surrounded by normal tissue.

< 3.0 cm in diameter

> 3.0 cm in diameter

50%

II A

B

Progression of cancer to the lymph nodes near affected lung

Growth into chest wall or covering, diaphragm or heart

30%

III A

B

Advanced stages of IIB

Multiple tumors, pleural effusion, affected lymph nodes beyond

the adjacent lobe, or growth into another chest structure

5-15%

IV n/a Spreading of cancer to the other lung lobe or organs outside the

lung

0-2%

6

Table 1.1. Staging of Small Cell Lung Cancer (SCLC) and Non-Small Cell Lung Cancer (NSCLC) and associated

5-year survival rates.


4

1.1.4 Early Detection

Early detection is thus integral to survival, with five-year survival rates approximately double for

Stage I compared to that of advanced stages.6 However, about 78% of all newly diagnosed cases

are advanced.7 Additionally, 50% of these advanced-stage patients have metastases distant from

the lung tissue of the original tumor, which greatly diminishes survival rates.8 The key to early

detection then must lie in first understanding the risk factors, then implementing a safe and

effective method for screening and early diagnosis.

1.2 Dose Characterization in CT

1.2.1 The Biological Effect of Radiation

CT relies upon X-ray radiation for the formation of 3D images, and the amount of radiation

should be well understood and minimized due to the potential for inducing carcinogenic effects

on biological tissue. X-ray photons interact at the molecular level, knocking electrons out of orbit

and creating unstable radicals, the most common of which are hydroxyl radicals. These ions can

cause strand breaks in nearby deoxyribonucleic acid (DNA) or damage nucleotide bases. X-rays

may also ionize DNA directly.9 If such damage goes unrepaired, several other complications

may result including mutations, translocations, and gene fusions, all of which can lead to

uncontrolled cell growth and tumor formation.9 Radiation dose must therefore be well

characterized, justifiably delivered, and minimized such that the imaging task may still be

accomplished. This is particularly important in CT, which involves a higher radiation dose than

less sensitive modalities, such as radiography.


5

Diagnostic radiology (as well as other medical and non-medical applications of ionizing

radiation) employs the principle of “as low as reasonably achievable” (ALARA) dose as

described by the International Commission on Radiological Protection (ICRP).10

That is,

exposure to radiation must be justified (the benefits outweighing the potential risks), optimized

(reducing dose levels as low as possible while maintaining success of the task), and dose-limited

(bound by upper limits of dose that may be received by a specific procedure, taking into account

all man-made exposures).10

The ALARA principle guides the development and quality control of

low-dose diagnostic procedures as well as the procedures for ensuring the protection of workers

occupied in the use of ionizing radiation.

1.2.2 Dose in CT

Radiation dose assessment in CT has evolved considerably over the last 30 years as CT

technology has progressed. The unit of absolute dose (or absorbed dose) typically used in CT is

the milligray (mGy) where 1 gray (G) corresponds to the absorption of one joule (J) of energy in

one kilogram (kg) of matter (e.g., in water).11

For non-uniform irradiations, the “effective dose”

estimates the corresponding uniform whole-body absorbed dose that would result in similar

stochastic effects.12

Effective dose accounts for the relative sensitivity of various organs and

tissues and is measured in Sieverts (Sv).

Dose in a CT scan can be controlled by means of scan length as well as imaging technique

factors, including X-ray beam energy (kVp) and X-ray tube current (mA), which governs the X-

ray fluence. For single-slice axial and helical CT, in which the radiation beam is a narrow axial

fan (~1 cm thick), dose is characterized by the Multiple Scan Average Dose (MSAD), defined as

the average dose at a particular depth from the surface.11

It incorporates the dose to tissue from

radiation absorbed within the fan and from radiation scattered to adjacent slices. The latter is


6

important, since Compton scattering is the principal interaction at the photon energies

characteristic to CT, and scattered dose is significant and outside of the primary beam.11

MSAD

is commonly estimated by the CT Dose Index (CTDI), which can be measured experimentally

using a pencil ionization chamber (e.g., 10 cm length), and a cylindrical water phantom (e.g., a

32 cm diameter acrylic cylinder simulating the adult abdomen).11

Because CTDI is based on a

measurement in a phantom, the various CTDI metrics defined over the last few decades have

been effective for quality control and scanner-to-scanner dose comparison, but are not in

themselves an estimate of patient dose.13

The dose index associated with a CTDI phantom

probably underestimates MSAD, particularly for small patient size, and overestimates it for very

large ones because a 32 cm diameter phantom does not represent human anatomical geometry.

As CT scanners have evolved to volumetric multi-detector designs, as described in chapters

below, CTDI as a dosimetry metric is evolving as well to account for broad volumetric beams

that exceed the length of conventional ionization chambers (10 cm) and to account for the

significant contribution of out-of-field X-ray scatter to total dose. Chapter III examines this issue

specifically and identifies a shortfall in simple CTDI metrics as well as some alternative methods

used to improve patient dose estimation. Consistent with such finding are ongoing efforts to

update the standards for dosimetry in volumetric CT, as undertaken in the American Association

of Physicists in Medicine (AAPM) Task Group Number 111 investigating The Future of CT

Dosimetry.14

1.2.3 Obesity and Image Quality

In 2004, an estimated 23.1% of Canadians were found to be obese (body mass index, BMI,

greater than 30) and another 36.1% were deemed overweight (BMI between 25 and 30).15

This

corresponds to approximately 14.1 million adults. Obesity can impede accurate diagnosis, for


7

example, in medical imaging methods for which image quality is degraded for large patient size.

A study at Harvard Medical School by Raul et al. found a positive correlation between obesity

and the frequency of habitus-limited radiology reports in Massachusetts between 1991 and

2001.16

In CT, an obese body habitus is associated with increased X-ray scatter and a reduced

number of photons reaching detectors. These factors lead to image artifacts and high image

noise, each of which can compromise the diagnostic task. As dose decreases, image noise

increases (in square-root proportion, as discussed in Chapter VI), calling for alternative

approaches to maintain image quality in low-dose screening initiatives.

1.3 Imaging of the Chest and Computed Tomography

1.3.1 Chest Radiography to Computed Tomography

Early detection and diagnosis of lung nodules in chest radiography (CXR) has been a common

clinical challenge for radiologists over the last few decades. A long-term screening study at the

Mayo Clinic found that for diagnosed lung cancer patients, 90% of those cancers were visible on

earlier radiographs with a miss rate as high as 35% for nodules greater than 3 mm in diameter,

the minimum clinically suspicious nodule size.17

These high miss rates may be attributable to a

variety of factors such as tumor properties (size and contrast) and the complex dynamic anatomy

of the chest cavity. The lungs may move slightly during imaging due to normal respiratory

function, which incorporates lung expansion and contraction as well as muscular movements of

the diaphragm, in addition to involuntary motions caused by cardiac motion. The lungs also

contain several materials of various radiological attenuation, including bone, soft tissue,

vasculature, fluid (pleural and blood) and air.2 X-ray radiography and CT are the most common

modalities for chest imaging applications, but CT is gaining popularity due to its high contrast


8

resolution and because the cross-sectional nature of CT removes overlying anatomical clutter

from the image. CT exhibits improved performance for very small nodule detection tasks as

compared to CXR; in some soft tissue tumors, CT numbers can differ by only about 20

Hounsfield Units (HU, i.e. the measure of radiological attenuation of a given tissue) from

surrounding lung, which is often too small to be resolved by CXR.11

1.3.2 Advances in Multi-Detector (Volumetric) CT

Computed tomography (CT) is a promising imaging modality for the clinical detection and

surveillance of early-stage lung nodules. Recent developments in CT scanner technology include

the capability for volumetric scanning (i.e., 16 cm longitudinal coverage) in a single gantry

rotation (~0.35 s), offering sub-millimeter axial spatial resolution, faster scan times, and

potentially reduced patient dose.18

Such capability offers immediate application in cardiac

imaging as well as a host of other “whole-organ” imaging applications, ranging from brain or

liver perfusion scans to thoracic imaging.

1.3.3 The Need to Minimize Radiation Dose in CT

While CT offers increased sensitivity and speed compared to conventional chest radiography, the

radiation dose associated with CT is typically ~100 times greater.19

Furthermore, the number of

CT examinations performed per year has increased by over an order-of-magnitude worldwide in

the last two decades, although rates vary from country to country.19

This presents a significant

health concern, and several epidemiological studies have been conducted to evaluate the

carcinogenic potential of radiation. Data from atomic bomb survivors has served as the gold

standard in predicting radiation-induced carcinogenesis, as 30% of this cohort endured anywhere

from ~5-100 mSv, a dose range similar to that of single or multiple CT examinations.19


9

Several results have been indicated in these studies that motivate reduction and minimization of

radiation dose. The widely accepted linear no-threshold (LNT) hypothesis states that the risk of

radiation-induced harm decreases with decreasing radiation dose and exhibits no threshold (i.e.,

no minimum amount of radiation) for biological damage. This means that any amount of

radiation, no matter how small, increases cancer risk.20

It is worth acknowledging that the LNT

hypothesis is unproven, and insufficient data is available for exposure to very low doses (<5

mSv). In fact, some have posed that cancer risk deviates from linearity (suggesting a low-dose

threshold or even a hormesis effect) at very low doses.21

Still, the LNT hypothesis is the most

conservative with respect to health risk, is the most widely accepted, and is the basis for much of

the regulatory standards with which the health industry must comply.

The risk of radiation-induced cancers is, therefore, believed to increase with dose, and results

also suggest that risk increases with earlier age of exposure.19

Children are not only more

radiosensitive, but they have more time for the effects of genetic damage to manifest later in

their lifetime. However, radiation-induced lung cancer appears to be an exception, with relative

risk supposedly increasing with age up to middle-age.19

Another important finding is the

multiplicative (as opposed to additive) effect of radiation and smoking on the development of

lung cancer.22

Brenner et al. conducted a study on radiation risk from low-dose CT screening of

smokers based on the atomic bomb survivor data. They showed that if 50% of all current and

former smokers in the U.S. population received annual CT screening for lung cancer from ages

50-75 years, the incidence of radiation-induced lung cancers would be about 36,000, a 1.8%

increase over the otherwise expected number (with lower and upper bounds of this estimate

equal to a 0.5% and 5.5% increase from the expected number, respectively, at a 95% confidence


10

interval).23

Moreover, based on the risk estimates by CT and its current usage, Brenner et al.

predicted that radiation from CT is responsible for about 1.5-2% of all new cancers in the U.S.24

Results from these studies are relevant to patients at high risk for lung cancer. The results

suggest that CT screening should only be performed when a mortality benefit of above ~5% (the

upper limit of CT screening-induced lung cancer, as stated above) can be achieved by screening,

which is generally true for a 50 year-old smoker (~14% risk of developing lung cancer).19

Thus

screening is an effective tool for high-risk lung cancer patients, but radiation dose must still be

minimized. There is considerable effort to minimize radiation exposure in CT across the field of

medical imaging research and clinical practice. Approaches include modification of beam energy

and fluence (i.e., lower kVp and mAs), tighter collimation (less anatomy exposed to radiation),

fewer scans performed, and alternative diagnostic tests that do not involve X-rays.

With these potential risks of radiation exposure in mind, it is important also to consider the real

and potential benefits of CT. For example, since its widespread adoption since the late 1970s, it

is largely responsible for rendering obsolete the concept of “exploratory surgery” and other

invasive diagnostic approaches. The risk, recovery time, and mortality associated with such

procedures clearly outweigh even the most conservative estimates of radiation-induced cancer

risk. Furthermore, CT has become an essential component – arguably the most essential

component – in the medical imaging arsenal for a very broad range of diagnostic tasks, including

detection of cancer at an early curable stage. The benefits of CT may be more difficult to

quantify than the risks, but its importance to diagnostic medicine is clear. Considering both

potential risks and benefits of CT, the logical course of action is to minimize CT dose in a

manner that preserves the diagnostic capability of this modality. As discussed below, one


11

approach is to knowledgeably select image acquisition and reconstruction techniques that

minimize dose while providing image quality sufficient for the diagnostic task.

1.3.4 CT for Early Detection and Surveillance of Lung Nodules

Early-stage lung cancer diagnosis relies on accurate detection and characterization of subtle lung

nodules. Short-term follow-up is required for nodules of diameter greater than 5 mm, while

nodules less than 5 mm in size (but greater than 3 mm) without history of malignancy require

annual follow-up.25

Surveillance imaging to monitor nodule growth is typically performed at 3-

12 month intervals.26

While increased sensitivity has made CT a viable modality for the

detection of lung nodules, CT currently faces the challenge of inadequate specificity at low doses

and so may result in a large number of false positives. LDCT may be unable to discriminate

between malignant disease and benign lesions, especially for small structures. However,

detection is the first step to diagnosing lung cancer and CT presents arguably the most important

modality to early detection and monitoring – e.g., surveillance of suspicious nodules.

In addition to providing a means of early-stage nodule detection, CT has increased the accuracy

in monitoring nodule growth, which is particularly important at the low doses required to

facilitate regular follow-up.27,28

However, at low doses image quality is degraded significantly

due to noise and image artifact, particularly for large patients. Reconstruction software offers a

variety of options that can reduce such effects, such as spatial frequency filters, slice thickness

selection, artifact correction, and noise reduction algorithms. To delve the low-dose detectability

limits without jeopardizing diagnostic accuracy, the relationships of dose, body size, and

reconstruction parameters to diagnostic accuracy should be more fully investigated, particularly

as new scanner technologies emerge.29


12

1.3.5 CT Reconstruction

After a CT scan is acquired, the 3D volumetric image is reconstructed from raw X-ray projection

data processed by filtered back-projection (FBP), which includes the application of a high-

frequency ramp filter in combination with a smoothing filter (also called a convolution kernel,

apodization window, or simply “reconstruction filter”). The ramp filter is intrinsic to FBP and

overcomes the radial blur associated with back-projection reconstruction. Conversely, the

reconstruction filter (typically a low-pass filter) generally reduces the high-frequency noise that

is amplified by the ramp filter, with a corresponding reduction in spatial resolution.

Reconstruction filters are distinguished by their frequency-pass characteristics, ranging from

very smooth (low-pass) filters that strongly suppress noise at the cost of spatial resolution to

relatively sharp (higher-pass) filters that attempt to balance the tradeoff of noise and resolution.

Selection of slice thickness and sampling interval are also important reconstruction technique

parameters. Thinner slices (e.g., as small as 0.5 – 1.0 mm) improve spatial resolution, typically at

the cost of increased image noise. CT scanner reconstruction software may also feature optional

image processing algorithms (most of which are proprietary) to reduce noise or artifacts. An

understanding of how these numerous reconstruction parameters govern image quality is

important to selection of optimal reconstruction technique, particularly for lower-dose protocols

in which image quality may be severely diminished by quantum noise.

1.4 Investigation of Low-Dose Lung Nodule Detectability

Previous studies of low-dose CT lung nodule detection have demonstrated that for small solid

tumors, substantial dose reduction is possible given prior knowledge of nodule size and contrast


13

(analogous to the task of nodule surveillance).26

The study reported here investigates the low-

dose limits of lung nodule detectability in volume CT. The objectives of this study are to:

1) Determine the effect on lung nodule detectability associated with the following factors:

a. Imaging technique (kVp, mAs, and dose)

b. Patient body habitus (average and obese body size)

c. Reconstruction techniques (reconstruction filter and slice thickness)

2) Assess diagnostic performance for the detection of small lung nodules:

a. Determine low-dose detectability thresholds for a variety of imaging techniques.

b. Identify optimal techniques that manage tradeoffs in image noise and spatial

resolution for various body habitus.

Low dose detectability limits are identified by means of multiple-alternative forced choice

(MAFC) human observer tests. The results help to guide selection of technique factors

appropriate to low dose imaging protocols in a manner that accounts for body habitus and

maintains diagnostic accuracy.


14

CHAPTER II: EXPERIMENTAL METHODS FOR

EVALUATION OF LUNG NODULE DETECTABILITY


15

2.1 CT Imaging and Reconstruction Techniques

2.1.1 Volumetric CT Scanner – Toshiba Aquilion ONETM

Images were collected on the clinical multi-detector CT scanner (320-slice Aquilion ONETM

,

Toshiba Medical Systems, Tokyo) shown in Fig. 2.1. Images were acquired in volume mode (16

cm z-coverage per rotation). Consistent with the clinical protocol for chest scans on this CT

system, four volumes were combined axially to effect a scan length of 32 cm (beam overlap

required as described in Chapter III) from the lung apices to the diaphragm of an

anthropomorphic chest phantom (described below). Gantry rotation speed was 0.35 s per 360°

revolution with dose, slice thickness, tslice, and reconstruction filter varied as detailed below.

FIG. 2.1. Photograph of the CT scanner (Toshiba Aquilion ONETM

) with the anthropomorphic phantom shown in

the imaging position. As described below, the phantom incorporated simulated lung nodules and was imaged in both

average and obese body habitus configurations (simulated fat visible in this photograph). MOSFET dosimeters in

the right lung measured the radiation dose associated with each scan technique.


16

2.1.2 Acquisition and Reconstruction Techniques

A total of 1134 volume CT images were acquired—the product of two phantom sizes, three kVp,

nine mAs, three slice thicknesses, and seven reconstruction filters—as detailed in Table 2.1.

Experimental Parameters

Phantom Chest Thickness kVp mAs

Slice Thickness

@ Interval

Reconstruction

Filter

Average (22 cm) 80 3.5 1 mm @ 1 mm FC1

Clinically Obese (32 cm) 100 7 3 mm @ 1.5 mm FC2

120 10.5 5 mm @ 2.5 mm FC3

14 FC4

17.5 FC5

24.5 FC11

35 FC50

70

105

Table 2.1. Summary of experimental parameters: phantom size, image acquisition techniques (kVp and mAs), and

reconstruction techniques (slice thickness and reconstruction filter).

The parameters in Table 2.1 were selected to investigate the effects of dose, patient habitus, and

reconstruction technique on lung nodule detectability. The values of these parameters were

selected on the basis of current clinical protocols and were varied to examine trends in

detectability at extremes of dose, patient size, and image processing options. The scan techniques

(kVp and mAs) spanned the range of typical diagnostic scans and extended to ultra low-dose

levels which are of major interest in CT screening and surveillance. The slice thickness and filter

selection included those typically used for lung nodule detectability tasks (5 mm slice thickness

with FC50 sharp lung filter) and employed smaller slice thicknesses and smoother filters to

examine the effects of noise and contrast on detectability. A phantom simulating both average

and obese patient habitus was employed, allowing investigation of adult (non-pediatric) chest

thickness extremes for assessment of the dose reduction that can be achieved through

knowledgeable selection of reconstruction techniques.


17

2.2 Phantom for Studies of Lung Nodule Detectability

2.2.1 Anthropomorphic Phantom with Simulated Lung Nodules

A custom anthropomorphic phantom30

based on the RandoTM

chest phantom (The Phantom

Laboratory, Greenwich, NY) was used for all image data. The phantom contains a natural human

skeleton and other tissue-simulating materials. Figure 2.2 shows an illustration of the phantom

components and an axial image of the left chest cavity.

30

FIG. 2.2. Anthropomorphic phantom. (a) Coronal view of anthropomorphic phantom. (b) Axial view of left lung

with nodules. The alphanumerical codes denote nodules of varying size and contrast in various layers of the lung.

Figure adapted from Ref. 30 Chiarot et al. “An innovative phantom for quantitative and qualitative investigation of

advanced x-ray imaging technologies” with permission from publisher.

The left lung is composed of a heterogeneous microballoon-polyurethane mixture formulated to

give electron density approximating that of lung (-696 ± 10 HU). A variety of spherical nodules


18

ranging in diameter and contrast (~1.6 - 12.7 mm and ~ –496 to +20 HU) are incorporated within

the left lung. Nodules selected for the current study were 3.2 mm in diameter and -37 HU (660

HU contrast to background), approximating the smallest nodule diameter likely to be followed

up. Though larger nodules may also be of importance and difficult to detect,31

this selection

facilitated the investigation of reconstruction techniques at the limits most relevant to nodule

surveillance (small, suspicious lesions).

The right, air-filled lung was accessible via a hole in the shoulder and was used for measurement

of radiation dose for each scan. Metal-oxide semiconductor field-effect transistors (MOSFETs,

Thomson-Nielsen MobileMOSFET, Best Medical Canada, Ottawa, ON) dosimetry were used, as

illustrated in Fig. 2.3. The MOSFETs were placed on the phantom surface and on a Styrofoam

rod inserted into the right lung. In this way, the dose corresponding to each scan technique in

Table II could be directly measured.

(a) (b)(a) (b)

FIG. 2.3. MOSFET dosimetry equipment. (A) MOSFET reader and wireless transmitter. (B) Probe with three

MOSFETs, inserted in right phantom lung. (C) Strip with two MOSFETs, secured to phantom chest. (D) BlueTooth

Receiver. (E) User Interface.


19

Three MOSFETs were mounted on the probe labeled (B) and inserted in the empty cavity of the

right lung [visible in Fig. 2.3(a), each MOSFET encased in a thin layer (~5 mm) of SuperFlabTM

bolus material and separated by ~15 mm. In addition, two MOSFETs were secured to the

exterior of the phantom on the strip labeled (C), one at the anterior and one at the left lateral

chest, each encased in ~5 mm bolus. The box labeled (A) is the MOSFET reader, which

wirelessly transmits the MOSFET dose information to the BlueTooth receiver (D) in the

reporting area and enables dose reading from the user interface (E) [both labeled in Fig. 2.3(b)].

MOSFET measurements were converted to Dlung (mGy) by a calibration factor (32.2 mV/mGy)

determined in an independent calibration within the diagnostic energy range (80 – 120 kVp).

Through variation of kVp and tube current across the full array allowed by the scanner, the

radiation dose ranged from a minimum of ~0.1 mGy up through doses typical of LDCT (~5

mGy) and diagnostic CT (~10 mGy).

2.2.2 Anthropomorphic Phantom: Average and Obese Body Habitus

The anthropomorphic phantom presented an “average” body habitus with an anterior-posterior

(AP) chest thickness of ~22 cm at the sternum. To simulate an “obese” habitus, 10 cm

SuperFlabTM

polymer (~46.1 ± 7.9 HU) was added (5 cm anterior + 5 cm posterior) as shown in

Fig. 2.4(a).


20

FIG. 2.4. Anthropomorphic phantom: average vs. obese configuration. (a) Photograph of the phantom with 10 cm

SuperFlabTM secured to the torso to simulate an “obese” habitus. Axial images of the phantom in (b) the “average”

habitus configuration (without SuperFlabTM) and (c) the “obese” habitus configuration. Magnified views of a

simulated 3.2 mm nodule are shown in each case. Imaging techniques for example images in (b) and (c) were 100

kVp, 105 mAs (CTDIvol.e = 6.7 mGy), FC3 reconstruction filter and tslice = 3 mm. Figure adapted from Silverman

et al. “Investigation of lung nodule detectability in low-dose 320-slice computed tomography” with permission from

publisher.

The associated body mass index (BMI) was approximated as

2Height

WeightBMI =

(Eq. 2.1)

in [kg/m2] where weight and height were estimated by assuming the phantom to be of uniform

density (1 g/cm3) and extrapolating arms and legs of approximate cylindrical length and

diameter. The BMI calculation is detailed in Appendix II.


21

The estimated BMI for the “average” habitus was ~22.3 kg/m2, and that of the “obese” habitus

was estimated to be in the range ~32.8 kg/m2 (assuming fat on the torso only) to ~45.7 kg/m

2

(assuming fat to cover the limbs as well). Modelling the SuperFlabTM

as muscle, with density

1.06 g/cm3 instead of adipose tissue (0.9 g/cm

3) yields an obese habitus ranging from ~34.6

kg/m2 (assuming fat on the torso only) to ~49.8 kg/m

2 (assuming fat to cover the limbs as well).

These BMI values correspond to clinically average (18.50 – 24.99 kg/m2) and obese (≥ 30.00

kg/m2) body habitus, respectively, recognizing the upper bounds of the BMI estimate for the

obese habitus to be arguably “morbidly obese” (40.00 - 49.99 kg/m2).

32

2.3 Observer Tests and Data Analysis

2.3.1 Observer Performance Test

The detectability of phantom nodules in the phantom images was assessed in multiple-alternative

forced-choice (MAFC) observer tests. In an MAFC test, the observer is presented with M images

on a diagnostic display, one of which contains a signal, the M-1 others containing only “noise”

(i.e., normal anatomical background), and the observer is required to choose which image

contains the signal.33

MAFC tests are a good choice for simple phantom images, are well

tolerated by observers operating upon many images (hundreds or more), can reduce inter-

observer variability, and tend to reduce observer response bias (i.e., the tendency to report the

presence of sub-threshold stimuli in a yes-no trial), which could introduce inaccuracies in low-

dose threshold estimation.33

In an MAFC test, the observer is forced to select which image most

likely contains the stimulus (e.g., a nodule), and when the stimulus is too weak to be detected,

the observer is expected to guess (in which case the likelihood of a correct selection is 1/M).


22

Physicists (six total) were considered sufficiently expert readers for the fairly simple task of

nodule detection (requiring no real knowledge of disease or anatomy). Observer tests were

conducted on a 3 MP diagnostic-quality display monochrome LCD monitor (AM-QX21-A9300,

National Display, San Jose, CA) calibrated to the DICOM standard in a dark-controlled

radiology reading room [0.15 Cd/m2 ambient light measured by a photometer (LumaColor

Photometer, Tektronix, Beaverton, OR)]. Each test (2520 cases total) was completed in two 2-

hour sittings with three 5-minute breaks per sitting to avoid observer fatigue. Each of the two

sittings included 1260 cases and was preceded by a five-minute training session in which the

user was presented with 63 cases (~5 minutes) representative of those in the test.

Tests were conducted using custom software

(OPTEx)34

developed in MATLAB (The

Mathworks, Inc., Natick, MA) for randomization of reading order, control of image display, and

analysis of observer response. Each case was presented as a nine-alternative forced choice

(9AFC) detection task as illustrated in Fig. 2.5. Each of the nine sub-images within a case were

taken from an image acquired at the same dose (kVp and mAs), reconstruction technique (slice

thickness and filter), and phantom habitus (average or obese). Images were displayed using a

fixed “lung” window [W(1800), L(-500)] and cropped to 3.25 cm x 3.25 cm (65 x 65 pixels). As

illustrated in Fig. 2.5, one of the sub-images contained a 3.2 mm nodule, and the others were

“noise-only.” The position of the stimulus was randomized in the 3 x 3 selection matrix, and the

user was prompted to click on the sub-image containing the nodule. To better emulate the

clinical diagnostic task (and to dissuade observers from simply picking the sub-image with the

brightest central pixel), the position of the sub-image centers was randomized by up to 10% of

the sub-image size to shift the nodule slightly from the center then flipped at random in x and y.

A comfortable viewing distance of ~50 cm was suggested to observers, but not strictly enforced.


23

Observers were not allowed to adjust window-level settings or magnification. In order to

increase the inter-observer consistency in selection criteria, detailed instructions (to the

aforementioned effect) were provided before each sitting of the observer test. The observer test

instructions are found in Appendix I.

FIG. 2.5. Illustration of 9AFC test for nodule detection. In this example [120 kVp, 14 mAs (CTDIvol.e = 1.6 mGy),

FC2 filter, tslice = 3 mm, average body habitus] the stimulus is in the top-center sub-image. The streaks in several

ROIs are believed to be due to photon starvation and beam hardening artifacts associated with the spine, ribs,

sternum and mediastinum (shown in Fig. 2.4). Figure adapted from Silverman et al. “Investigation of lung nodule

detectability in low-dose 320-slice computed tomography” with permission from publisher.

Selection of the 2520 cases from the 1134 combinations [of beam energy (3), fluence (9), body

habitus (2), reconstruction filter (7) and slice thickness (3)] was weighted preferentially toward


24

lower-dose images, for which conspicuity decreased significantly. For all dose levels (i.e.,

combinations of beam energy and fluence) at least one nodule was shown to each observer per

parameter set (i.e., combination of habitus, reconstruction filter and slice thickness). For the

lower doses, 1-3 nodules were shown per parameter set with 1-3 repeats per nodule. The order in

which various cases were presented was randomized.

Results were analyzed by grouping the observer responses and determining the proportion

correct (Pcorr) for each set of phantom size, imaging and reconstruction techniques. To estimate

the corresponding area under the receiving operating characteristic (ROC) curve (denoted Az), a

lookup table of Az values as a function Pcorr for M = 9 was used. A three-parameter logarithmic

fit was employed to interpolate Az between tabulated Pcorr values. Specifically, a fit of the form:

)ln(* cPbaA corrz +−= (Eq. 2.2)

was used, where a, b, and c are fitting parameters. For M=9, a, b, and c were 1.00, -0.25 and

0.02, respectively. This equation assumes a binormal distribution typical of ROC analysis and a

unique relationship between Pcorr and Az such that the observer’s response characteristic

(decision threshold) does not change over the course of the test.

Values of Az ~0.95-1.0 represent conspicuity, while Az ~0.5 corresponds to pure guessing. Note

that the nodule shown in Fig. 2.5 is fairly conspicuous (for purposes of illustration), while the

images used throughout the observer tests ranged from conspicuous (Az ~1) to barely detectable

(Az ~0.7) to undetectable (Az ~0.5). The resulting Az were plotted as a function of dose for all

reconstruction slice thicknesses and filters as exemplified in Fig. 2.6. Error bars were determined

by specifying a 95% confidence interval on a discrete binomial distribution.


25

0.1 1 10

0.5

0.6

0.7

0.8

0.9

1.0

Az

CTDIvol.e (mGy)

Dthresh0.1 1 10

0.5

0.6

0.7

0.8

0.9

1.0

Az

CTDIvol.e (mGy)

DthreshDthresh

FIG. 2.6. Observer performance as a function of dose. Error bars represent 95% confidence intervals. The fit is a

one-parameter logistic function as discussed below. Figure adapted from Silverman et al. “Investigation of lung

nodule detectability in low-dose 320-slice computed tomography” with permission from publisher.

2.3.2 Data Analysis

A one-parameter logistic function was used to fit measurements of Az versus dose. A one-

parameter fit was chosen to minimize the error caused by parameter estimation and to ensure that

all curves were of the same shape to facilitate comparison of observer performance for varying

reconstruction techniques. A sigmoid fit of the form

)*exp(1

1

DaAz −+=

(Eq. 2.3)


26

was used, where D corresponds to the dose (specifically, the CTDIvol.e as described in the

following chapter) and ‘a’ is a fit parameter determined by minimizing χ2 on the data. The curve-

fitting software (OriginPro 8, OriginLab, Northampton, MA) returned an estimate and standard

error on ‘a’. A metric (denoted Dthresh) was defined as the dose at which detectability (Az)

decreased to a level of 0.95 (compared to 1.0 at arbitrarily high dose). The determination of

Dthresh is illustrated graphically by the dotted lines in Fig. 2.6 and may be determined analytically

from the inverse of the sigmoid fit as

aa

AAADD zz

zthresh

)05.0/95.0ln(])1/[ln()95.0( =

−===

. (Eq. 2.4)

The standard deviation in Dthresh was estimated from the error in the fit parameter and the

derivative

)(ada

dDthreshaDthresh

σσ = . (Eq. 2.5)

Thus, low Dthresh values represent high observer performance (i.e., high detectability) at low dose

levels, and high Dthresh indicates poor performance at low dose. To evaluate the effects of

reconstruction filter and slice thickness selection for average and obese body habitus, Dthresh

values were compared. For each combination of reconstruction settings the calculated Dthresh was

taken as the mean of a normal distribution with standard deviation σDthresh and compared to all

other combinations via paired, unequal-variance, two-tailed student t-tests. For each comparison,

this test returns the probability (p-value) that the measured results correspond to the null

hypothesis (no difference between two distributions), with p < 0.05 taken to indicate a

statistically significant difference.


27

CHAPTER III: CHARACTERIZATION OF RADIATION DOSE


28

3.1 Evolution of Dosimetry in Computed Tomography

As discussed in Chapter I, the improvement of CT technology over the last 30 years has had a

profound impact on radiation dose assessment. Several dosimetry devices and methods aimed at

measuring the Multiple Scan Average Dose (MSAD) have been outdated by the emergence of

wide-beam multi-detector (volumetric) designs. While these designs may improve speed and

suppress image artifacts, dose management is still a concern, and new standard methods are

being considered for accurate dose evaluation.

The Computed Tomography Dose Index (CTDI) has existed in several forms generally

distinguishable by the length along which dose is measured and by what instrumentation. The

theoretical CTDI∞ is the limiting equilibrium dose in a cylindrical PMMA (polymethyl

methacrylate; acrylic) phantom after which scatter tails outside the measurement area add

negligible increase in the accumulated dose.35

The first practical measurement to estimate CTDI∞

was the CTDIFDA which integrated the dose profile, D(z), over a distance of 14 beam widths

centered about the middle of the beam but the maximum beam width was generally only 10

mm.36

Soon afterward, the CTDI100, defined as the average dose to a phantom at the center of a

beam accumulated in a 100 mm scan, was introduced and widely accepted by physicists as an

estimate of patient dose.36

It was measured experimentally using a 100 mm pencil ionization

chamber and a 32 cm diameter cylindrical water phantom to simulate the adult abdomen.

However several recent studies have challenged both the pencil chamber methodology and the

success of CTDI100 at estimating the limiting equilibrium dose for volume CT.35,36,37

Though CTDI100 was reasonable for single-slice CT scanners with beam widths of 10 mm, it can

not sufficiently account for the scatter tails of the large beams associated with multi-detector CT


29

(MDCT) scanners. In a Monte Carlo simulation study published in 2007, Boone defined a

CTDI100 (as measured by pencil chambers) dose efficiency metric36

∞

=CTDI

CTDI100ε (Eq 3.1)

and determined it to be approximately 0.63 (63%) in body phantoms scanned at beam widths

varying from 10 to 40 mm (and decreasing only slightly with increasing beam width, since the

beams were sufficiently small that their scatter tails were mostly captured).36

In an experimental

validation of this simulation, Dixon et al. confirmed that CTDI100 underestimated dose for

longer, clinically relevant body scans and proposed a new method to measure dose profiles using

small Farmer-type ion chambers (instead of the conventional pencil chambers) in the center of a

sufficiently long body phantom.35

Dixon used extended body phantoms because the 15 cm body

phantoms (typically used) were not long enough to realistically represent the scatter-associated

dose in a patient.35

Mori et al. conducted similar measurements using point dosimeters to

measure the dose profile in a 900 mm phantom on a prototype 256-slice CT scanner (Toshiba

Aquilion, Toshiba Medical Systems).37

As new scanners emerge with increasingly wide beams

the challenge to estimate patient dose remains and new dose metrics are proposed.

In this study, the Aquilion ONETM

320-slice volumetric scanner was employed for dose

characterization and image acquisition, with dosimetry methods similar to those of Dixon.35

In

the sections below, we report measurements of dose to answer a number of basic questions,

including: 1.) How does the absolute dose measured by point dosimeters compare with the dose

reported by the scanner / manufacturer? 2.) What is the out-of-field dose (scatter penumbra)

associated with broad volumetric beams? 3.) How does the dose depend on the beam width and

scan length? The first is intrinsic to the measurements of low-dose thresholds of lung nodule


30

detectability that is the basic question in this thesis. The latter two are important additional

questions that should be considered from clinical standpoints (e.g., regarding the dose to

structures outside the field of view) and technical standpoints (e.g., in the development of new

dosimetry standards for volumetric CT).

3.2 Measurements of Dose

3.2.1 Dose Reported by the Scanner (CTDIvol.e and DLP.e)

The Aquilion ONETM

scanner reported dose in terms of CTDIvol.e (termed by the manufacturer

as the “extended” CT dose index) and an associated dose length product (DLP.e) to estimate the

MSAD in mGy and mGy·cm, respectively. According to the manufacturer, the CTDIvol.e is

calculated as

( )][10

.cm

dzzDeCTDIvol∫=

(Eq. 3.2)

where the numerator is the integral of the dose profile measured by a 10 cm pencil chamber

moved along the z-axis (at both the central and peripheral locations) in a beam of width 10 cm or

greater. For narrower beams, the denominator is equal to the nominal beam width. CTDIvol.e

varies as a function of beam energy, collimation, X-ray tube current, scan length, and phantom

size (head or body, as programmed within the scanner protocols). This extended dose metric is

intended to account for the long tails of scatter associated with volumetric imaging; however, to

the extent that the 10 cm pencil chamber may be insufficient to cover the long scatter tails in

volumetric CT (i.e., that the numerator in Eq. 2.2 does not correspond to the total integrated

dose), one may anticipate underestimation of the actual dose by this approach. Knowing previous


31

limitations of pencil chamber methodology, dose profiles were measured in this study using

point dosimeters, as described below.

3.2.2 Dose Measured Using Farmer Chambers

As a result of their high accuracy, robustness and ease of use, Farmer chambers have become a

common tool in point dose measurement of ionizing radiation. In these chambers an electric field

is applied to a known volume of air (typically ~ 0.6 cm3) and the electrode detects charged

particles from the interaction of X-ray photons with air and surrounding material. In the higher

energy ranges of such applications as radiotherapy, build-up caps are required to increase

interactions producing charged particles; however build-up caps and cable sleeves were not used

in this experiments as they have been shown to have little effect (0.6-0.9%)35

on the charge

measurement in the diagnostic energy range (80-140 kVp).

As a basis of comparison to the manufacturer-reported CTDIvol.e, the absolute dose was

measured using Farmer chambers placed within three CTDI body phantoms (32 cm diameter

acrylic; 48 cm total length; RTI Electronics) stacked along the z-axis of the scanner as shown in

Fig. 3.1. The long phantom configuration is sufficient to include the wide cone angle (16 cm

primary beam) as well as scatter tails. Two Farmer chambers (0.6 cc air ionization chambers,

Aluminum-electrode, graphite tip; Thomson-Nielsen, Best Medical Canada, Ottawa, ON.) were

inserted to 24 cm depth (half-depth of extended phantom) at the phantom center (16 cm from the

surface) and periphery (1 cm from the surface). The chambers were independently calibrated by

an accredited calibration laboratory (National Research Council, Ottawa, ON.).


32

FIG. 3.1.Dosimetry experiment setup. Three 32 cm diameter cylinders were stacked to 48 cm length to simulate a

human torso. Farmer chamber dosimeters were placed half-way inside the 48 cm phantom in the central and

peripheral positions.

Two electrometers (Advanced Therapy Dosimeter, FLUKE Biomedical, Everett, WA) recorded

the electrode charge, which was converted to absolute dose (mGy) by applying a calibration

factor (~45 mGy/nC) and temperature-pressure correction. For comparison to CTDIvol.e, the

weighted CTDI (CTDIw) was calculated from the doses measured at the center and periphery

(Dcenter and Dperiphery, respectively) as

peripherycenterw DDCTDI3

2

3

1+=

. (Eq. 3.3)

This weighted sum of central and peripheral dose is consistent with widespread definition of the

derived metric, CTDIw, and is also implied in the above definition of CTDIvol.e.


33

3.2.3 Dose Measured Using MOSFETs

As described in the previous chapter, the dose delivered inside the simulated lungs of the

anthropomorphic phantom was measured using MOSFETs (Thomson-Nielsen MobileMOSFET,

Best Medical Canada, Ottawa, ON.) with an active region of 0.2 x 0.2 mm2 as point dosimeters.

When a sufficiently large negative voltage is applied to the gate, minority carriers called holes

are attracted to the surface at the source and drain regions and current flows through the device.38

When irradiated, there is a build-up of trapped charge, and an increase in interface and bulk

oxide traps in the sensitive region.38

A secondary electron moves out of the gate electrode and

the hole becomes trapped causing a negative threshold voltage.38

The voltage shift before and

after exposure are measured and the difference, ∆VTH, is proportional to dose through a

calibration factor. For the diagnostic keV energy range, this calibration factor is 32.2 mV/cGy.39

The calibration factor does not change considerably for the beam energies utilized in this

experiment (80-120kVp).39

Some benefits for MOSFETs over TLD dosimeters are low energy

dependence, high sensitivity and immediate readout capabilities.

To evaluate the accuracy of the MOSFETs relative to the Farmer chambers, two MOSFETs were

incorporated in the long (48 cm) CTDI phantom and scanned at the same conditions described

above for the Farmer chambers. The setup was similar to that of the Farmer chambers in Fig. 3.1,

but with MOSFETs on Styrofoam probes in the place of the Farmer chambers.

3.3 Dosimetry Experiments

Dose measurements were performed at 80, 100, and 120 kVp, 200 mA, and 0.5 s gantry rotation

(100 mAs) and reported in terms of mGy/mAs. These experiments addressed several queries


34

surrounding Aquilion ONETM

dosimetry including: (1) How does the CTDIvol.e dose estimate

compare with measurements performed with Farmer chambers and MOSFETs? (2) What is the

longitudinal radiation profile of the beam and the out-of-beam scatter? (3) How does beam width

affect the radiation delivered to a point dosimeter within the beam? (4) What is the effect of

changing scan length on the point dose measurements?

3.3.1 Comparison of CTDIvol.e with Farmer Chamber and MOSFET Dose

As noted above, Farmer chambers were placed in the center and periphery of the 32 cm diameter

48 cm long body phantom as shown in Fig. 3.1. Figure 3.2 illustrates the experimental setup with

a beam 16 cm wide at the phantom center and symmetrical about the dosimeters’ longitudinal

position.

side view

*

Beam Width = 16 cm

side view

*

Beam Width = 16 cm

FIG. 3.2. Experimental setup for comparison of measured and reported dose. Three 16 cm wide acrylic body

phantoms are stacked longitudinally on the CT couch. The star illustrates the source of the beam while the diamonds

represent the central and peripheral point dosimeters (either Farmer chambers or MOSFETs).

Farmer chambers and MOSFETs were used to measure the point dose at the phantom center and

periphery (represented by the small diamond shapes in Fig. 3.2) at the beam center. The reported


35

CTDIvol.e and the measured doses (Farmer chamber and MOSFET) were compared for single

volume scans of the CTDI phantom. Figure 3.3 displays the mean and standard deviation of the

Dcenter and CTDIw measurements over five trials.

80 100 120

0.02

0.04

0.06

0.08

0.10

0.12

Do

se (

mG

y/m

As)

Beam Energy (kVp)

CTDIvol.e

Dcenter (Farmer)

Dcenter (MOSFET)

CTDIw (Farmer)

CTDIw (MOSFET)

80 100 120

0.02

0.04

0.06

0.08

0.10

0.12

Do

se (

mG

y/m

As)

Beam Energy (kVp)

CTDIvol.e

Dcenter (Farmer)

Dcenter (MOSFET)

CTDIw(Farmer)

CTDIw(MOSFET)

80 100 120

0.02

0.04

0.06

0.08

0.10

0.12

Do

se (

mG

y/m

As)

Beam Energy (kVp)

CTDIvol.e

Dcenter (Farmer)

Dcenter (MOSFET)

CTDIw (Farmer)

CTDIw (MOSFET)

80 100 120

0.02

0.04

0.06

0.08

0.10

0.12

Do

se (

mG

y/m

As)

Beam Energy (kVp)

CTDIvol.e

Dcenter (Farmer)

Dcenter (MOSFET)

CTDIw (Farmer)

CTDIw (MOSFET)

80 100 120

0.02

0.04

0.06

0.08

0.10

0.12

Do

se (

mG

y/m

As)

Beam Energy (kVp)

CTDIvol.e

Dcenter (Farmer)

Dcenter (MOSFET)

CTDIw(Farmer)

CTDIw(MOSFET)

80 100 120

0.02

0.04

0.06

0.08

0.10

0.12

Do

se (

mG

y/m

As)

Beam Energy (kVp)

CTDIvol.e

Dcenter (Farmer)

Dcenter (MOSFET)

CTDIw(Farmer)

CTDIw(MOSFET)

FIG. 3.3. Comparison of CTDIvol.e (as reported by the scanner) and Dcenter, or CTDIw (as measured by Farmer

chambers and MOSFETs in a long 32 cm diameter CTDI phantom for one X-ray tube revolution). For Dcenter and

CTDIw, the mean and standard deviation over five trials are shown. Figure adapted from Silverman et al.

“Investigation of lung nodule detectability in low-dose 320-slice computed tomography” with permission from

publisher.


36

The CTDIw determined from the Farmer chamber measurements were found to agree with

CTDIvol.e only to within ~30%. The source of systematic discrepancy (CTDIvol.e consistently

less than Farmer chamber CTDIw) is likely associated with an inability of the pencil chamber

method to completely integrate long tails of X-ray scatter. As CT is evolving to wider beams

with increased scatter penumbra, there exists a need for better standardization in CT dosimetry

so that such discrepancy is reduced, or at the very least, its causes widely understood and

accounted for.

While the discrepancy between CTDIvol.e and CTDIw as measured by Farmer chambers is

notable, the reported values of CTDIvol.e were taken as the abscissa in evaluation of Az versus

dose in the nodule detectability study, since this value of dose gives the most “portable”

interpretation of results (e.g., with respect to other scanners and other institutions, for which only

the scanner-reported CTDIvol.e is available). The resulting Dthresh values (i.e., the dose at which

Az was reduced to 0.95) therefore correspond to CTDIvol.e (mGy). To the extent that the Farmer

chamber CTDIw is a more accurate dose value, the resulting CTDIvol.e and Dthresh may be

related to it by Fig. 3.3.

Comparison of the dosimetry methods demonstrates that the Farmer chamber and MOSFETs

yield similar measurements in the center of the phantom (similar Dcenter). However the CTDIw

were higher for Farmer chambers, which suggests that they were more sensitive near the

periphery (given the calculation for CTDIw). This may be attributable to insufficient build-up

material on the MOSFETs in the CTDI phantom. The small holes for dose measurement do not

provide space for extraneous materials that may otherwise cover the active surface of the

MOSFETs. The MOSFETs likely underestimated dose compared to the Farmer chambers

particularly at higher beam energies as is supported by the findings reported in Fig. 2.3. For the


37

primary study with the anthropomorphic phantom, 5 mm of bolus material covered all

MOSFETs to ensure accurate dose measurement.

3.3.2 Effect of Beam Position and Scatter on Dose Profile

It is necessary to understand not only the magnitude but also the spatial distribution of dose

administered to patients in CT. Particularly, for the wide beam of volume CT it is important to

characterize the dose profile within and outside the primary beam. In an ideal beam the dose

would be constant within the primary beam and fall sharply outside of the beam. In fact, the

measurement of CTDIvol.e with a 10 cm pencil chamber implicitly makes this assumption.

*

Z-offset = 0

*

Z-offset

(a) (b)

*

Z-offset = 0

*

Z-offset

(a) (b)

FIG. 3.4. Experimental setup for effect of beam position and scatter on dose profile. The couch was moved such that

the distance between the dosimeters (diamond shapes) and beam center varied from (a) Z-offset = 0 cm to (b) Z-

offset > 0 cm (to a maximum of z-offset = 180 mm).

To characterize the dose profile the CTDIw was calculated by Farmer chamber measurements at

the center and periphery of an extended body phantom as described above. The couch was

positioned such that the dosimeters fell in the center of the beam, as illustrated in Fig. 3.4(a) and

the couch was moved in increments of 10 mm while single-volume scans were performed at each

position [e.g., Fig. 3.4(b)] up to 180 mm couch displacement. In this way, the ion chambers


38

measured dose as a function of longitudinal distance from the center of the beam (termed z-

offset).

This study demonstrated several interesting results which are illustrated in Fig. 3.5. In an ideal 16

cm beam, we would expect to see a constant dose within the beam and a sharp decrease in the

dose profile at the edge (z-offset = 80 mm). As shown in Fig. 3.5(a), the dose measured at the

center of the phantom is fairly constant, with broad penumbra beyond 80 mm that demonstrate a

very gradual fall-off. This corresponds to dose delivered to the patient but not contributing to the

image data. X-ray scatter within the body phantom contributes dose of ~10% of the central dose

at distances of ~10 cm from the edge of the primary beam. The peripheral dose shown in Fig.

3.5(b) exhibits a less uniform profile within the primary beam and a steeper dose profile near the

beam edge, because (unlike the central dosimeter) it is not uniformly irradiated throughout the

gantry rotation. The non-flatness of the peripheral dose profile within the beam may be

attributable to beam inhomogeneities as well as x-ray scatter, which cancel out for the central

dosimeters.

The peripheral dose was also observed to be approximately double that measured at the center

for each couch displacement, which affects the derived metric, CTDIw, to a large extent because

it is weighted preferentially toward the peripheral dose measurement. A thorough understanding

of the out-of-beam X-ray scatter fluence for wide volumetric beams is the subject of ongoing

work in CT imaging physics and scanner design, and such understanding will help not only to

improve scanner configurations, but also to guide clinicians in their selection of scan protocols,

such as the beam width or scan length.


39

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0 20 40 60 80 100 120 140 160 180

Z-Distance from Beam Center to

Chamber (mm)

Dose

Measu

red

(m

Gy

/mA

s)

●120 kVp

■100 kVp

▲80 kVp

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

0 20 40 60 80 100 120 140 160 180


Chamber (mm)

Do

se M

easu

red

(m

Gy

/mA

s) ●120 kVp

■100 kVp

▲80 kVp

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0 20 40 60 80 100 120 140 160 180


Chambers (mm)

CT

DI w

(m

Gy/m

As)

●120 kVp

■100 kVp

▲80 kVp

(a)

(b) (c)

Cen

tral

Do

se (

mG

y/m

As)

Per

iph

eral

Do

se (

mG

y/m

As)

Z-Offset (mm)

Z-Offset (mm) Z-Offset (mm)

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0 20 40 60 80 100 120 140 160 180


Chamber (mm)

Dose

Measu

red

(m

Gy

/mA

s)

●120 kVp

■100 kVp

▲80 kVp

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

0 20 40 60 80 100 120 140 160 180


Chamber (mm)

Do

se M

easu

red

(m

Gy

/mA

s) ●120 kVp

■100 kVp

▲80 kVp

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0 20 40 60 80 100 120 140 160 180


Chambers (mm)

CT

DI w

(m

Gy/m

As)

●120 kVp

■100 kVp

▲80 kVp

(a)

(b) (c)

Cen

tral

Do

se (

mG

y/m

As)

Per

iph

eral

Do

se (

mG

y/m

As)

Z-Offset (mm)

Z-Offset (mm) Z-Offset (mm)

FIG. 3.5. Measured dose as a function of longitudinal distance from the center of the beam (z-offset). Dose was

measured by Farmer chamber at constant (maximum) beam width of 16 cm. (a) Dose measured at the center of the

32 cm diameter (48 cm long) body phantom (b) Dose measured at the periphery of the phantom (c) CTDIw

calculated from (a) and (b).


40

Although the Farmer chamber is seen as an excellent choice for “point” dose measurement, it is

important to note that it is not of infinitesimal length. In fact, it is a small (~1 cm) pencil

chamber, and the dose value measured is the average dose administered over this distance.

Therefore, to obtain actual point dose data, the curves in Fig. 3.5 could be deconvolved with a

function corresponding to the Farmer chamber response over its finite length. Such

deconvolution was not performed in the current study, although as Farmer chambers become

more common in CT dosimetry for wide volumetric beams, such deconvolution may be standard

in determining accurate z-direction profiles.

3.3.3 Effect of Beam Width on Dose

Depending on the diagnostic task and requirements (e.g., scan time and longitudinal field of

view), the beam width may be changed to modify longitudinal coverage and avoid irradiating

tissue outside the region of interest. The z-offset experiment described above shows significant,

broad penumbral dose for volumetric beams. Of course, penumbral dose exists for narrow beams

as well, and such penumbra sum with each rotation of the scanner. Which scenario is

dosimetrically advantageous is an area of ongoing work.

To examine this point, Farmer chamber dosimeters were centered within the beam (constant z-

offset = 0) and the irradiation associated with six beam widths available on the scanner (from 4

cm to 16 cm coverage along the z-axis) was measured as illustrated in Fig. 3.6.


41

* *(a) (b)

Beam Width = 16 cm Beam Width < 16 cm

* *(a) (b)

Beam Width = 16 cm Beam Width < 16 cm

FIG. 3.6. Experimental setup for effect of beam width on point dose measurement. Farmer chambers are represented

by the diamonds at the center and periphery. The beam was varied from (a) 16 cm, the maximum beam width and

(b) below 16 cm (to a minimum of 4 cm).

Results are shown in Fig. 3.7(a), indicating a significant increase in dose at the center of the

beam as a function of beam width. In fact, the dose for the 16 cm beam is approximately double

that of the 4 cm beam. This dramatic increase is entirely attributable to the X-ray scatter dose. A

higher dose is generated at the periphery, as shown in Fig. 3.7(b), exhibiting a smaller

dependence of dose on beam width due to a smaller scatter dose contribution at the periphery

(compared to the center). The peripheral dose contributes significantly to the CTDIw, as in Eq.

3.3, so that the overall dependence is somewhat in between that of the central and peripheral

dose. Therefore, not only does location within or outside the beam affect dose levels, but the

width of the beam itself directly contributes to dose without necessarily improving image quality.

This must be considered when setting scan parameters for a specific diagnostic task, and the

requirements of speed, spatial resolution, and longitudinal coverage should be carefully

considered. For example, if the location of a lesion can be predicted within a few centimetres, it

may be unnecessary to utilize the full beam width or to scan the entire chest.


42

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

20 40 60 80 100 120 140 160 180

Width of Beam (mm)

Measu

red

Dose

(m

Gy/m

As)

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

20 40 60 80 100 120 140 160 180

Width of Beam (mm)

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

20 40 60 80 100 120 140 160 180

Width of Beam (mm)

CT

DI w

(m

Gy

/mA

s)

(a)

(b) (c)

● 120 kVp

■ 100 kVp

▲ 80 kVp

●120 kVp

■100 kVp

▲80 kVp

●120 kVp

■100 kVp

▲80 kVp

Per

iph

era

l D

ose

(m

Gy/m

As)

Cen

tral

Dose

(m

Gy/m

As)

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

20 40 60 80 100 120 140 160 180

Width of Beam (mm)

Measu

red

Dose

(m

Gy/m

As)

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

20 40 60 80 100 120 140 160 180

Width of Beam (mm)

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

20 40 60 80 100 120 140 160 180

Width of Beam (mm)

CT

DI w

(m

Gy

/mA

s)

(a)

(b) (c)

● 120 kVp

■ 100 kVp

▲ 80 kVp

●120 kVp

■100 kVp

▲80 kVp

●120 kVp

■100 kVp

▲80 kVp

Per

iph

era

l D

ose

(m

Gy/m

As)

Cen

tral

Dose

(m

Gy/m

As)

FIG. 3.7. Effect of beam width on point dose. Dose was measured by Farmer chambers at the center of a 100 mAs

beam. (a) Dose measured at the center of the 32 cm diameter 48 cm long body phantom (b) Dose measured at the

periphery of the phantom (c) CTDIw calculated from (a) and (b).


43

3.3.4 Effect of Scan Length on Dose

In addition to the understanding of dose profiles and beam characteristics gained from the z-

offset and beam width experiments, dose may be affected by the user-specified parameter of scan

length. At a constant beam width, the Aquilion ONETM

divides the volume into the minimum

number of beams required to sufficiently cover the specified scan length. This is not necessarily

straightforward, because some beam overlap is required to ensure sufficient data collection

toward the periphery of the scanned subject [e.g., under-sampled (grey) “cone” or “chamfer”

areas of the image between beams in Fig. 3.8(c)].

This experiment compared the dose measured at the longitudinal center of the phantom for four

scan lengths (four distinct 16 cm beam configurations). Fig. 3.8(a) is the nominal, single-beam

16 cm beam geometry, while Figs. 3.8(b) and (d) are configurations produced automatically by

the scanner when specifying scan lengths of 24, 32, and 48 cm. Note that the configuration in

Fig. 3.8(c) is not a typical scan geometry, as it would result in under-sampled “chamfer” regions

in the image. This geometry was forced simply for purposes of examining the effect on

dosimetry.


44

1

*

Scan Length = 16 cm Scan Length = 24 cm


* **

* ** * ***

3

3w 4

(a) (b)

(d)(c)

1

*



* **

* ** * ***

3

3w 4

(a) (b)

(d)(c)

FIG. 3.8. Experimental setup for measuring the effect of scan length on point dose measurement. Progressively

darker beam intersections represent overlapping regions of irradiation. The scan length and number of beams were

(a) one 16 cm wide beam (maximum) required for 320-slice image acquisition (b) 24 cm scan length achieved with

3 beams (c) 48 cm scan length achieved with 3 beams, and (d) 32 cm scan length achieved with four beams. All but

(c) are the actual scan configurations used clinically to achieve the given scan length.

The results are shown in Fig. 3.9 and can be well predicted by the understanding of the dose

profiles gained from the z-offset experiment along with a geometrical analysis of the beams and

dosimeter locations. As shown in Fig. 3.9, the average CTDIw was measured over three trials for


45

each scan configuration and beam energy shows a clear dependence on scan length (denoted 1, 3,

3w, and 4).

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

80 100 120

Beam Energy (kVp)

CT

DI w

(m

Gy

/mA

s)4

3

3w

1

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

80 100 120

Beam Energy (kVp)

CT

DI w

(m

Gy

/mA

s)4

3

3w

1

FIG. 3.9. Effect of scan length (number of beams) on measured dose. CTDIw was measured by Farmer chambers at

the center of an extended body phantom. Beams had width 16 cm, and beam energies of 80, 100 and 120 kVp. The

average CTDIw over three trials is plotted and error bars are standard deviations.

As expected, the single beam 16 cm scan length configuration in Fig. 3.8(a) yielded the lowest

dose. For the three-beam 24 cm scan length the dose increased as the dosimeters fell directly

within one beam but on the edge of two others. The dose was reduced slightly in the “3-wide”

(3w) beam configuration (scan length 48 cm), because the dosimeters were further away on the

scatter tails of the two outside beams, but those tails did add significant dose as the values were

still consistently higher than those of the single beam setup. The highest dose was measured in

the 4-beam configuration (32 cm scan length) as expected, because the dosimeters fell within


46

two primary beams. However, even after adding dose from the scatter tails of the two outside

beams, these doses were less than twice the dose measured for the single beam configuration.

This may be attributable to the non-homogeneous dose distribution within the primary beam,

since the maximum dose is expected at the beam center based on previous results.

Calculating CTDIw from two point dose locations is a conventional – but imperfect –

characterization of absolute volumetric dose. Moving the dosimeters in the longitudinal direction

changes the measured dose (due to variation of the dose profile). However the CTDIvol.e

reported by the scanner consistently underestimates the dose for all beam energies and

configurations as shown in Fig. 3.10.

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

1 3 3w 4 1 3 3w 4 1 3 3w 4

Number of Beams and Energy (kVp)

CTDIvol.e

CTDIw measured

CT

DI

(mG

y/m

As)

80 100 120

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

1 3 3w 4 1 3 3w 4 1 3 3w 4

Number of Beams and Energy (kVp)

CTDIvol.e

CTDIw measured

CT

DI

(mG

y/m

As)

80 100 120

FIG. 3.10. Comparison of measured and reported dose for each beam configuration. CTDIw was measured by

Farmer chambers and CTDIvol.e was reported by scanner.


47

This is attributable to the inability of a 10 cm pencil ion chamber to capture the full extent of the

X-ray scatter tails. A “point” dosimeter such as a Farmer chamber, on the other hand, makes no

assumption or beam width normalization as in Equation 3.2.

3.4 Summary

The evolution of scanner technology toward volume CT has been modified largely by reduced

scan time for longer scan lengths, but a compromise in out-of-field dose is identified in the

measurements above. The methodologies for dose measurement require a breadth of

considerations for proper adaptation to volumetric CT, including phantom design (viz., longer

phantoms) and dosimeter selection (e.g., pencil chambers versus point dosimeters). Dose profiles

in the Aquilion ONETM

appear to be complex: irradiation levels are not constant within the

primary beam, and scatter tails are very broad. The viability of CT in lung nodule screening and

surveillance lies in both high diagnostic accuracy and effective dose management. While the

field of CT dosimetry evolves to deal with the particular challenges of volumetric beams and

broad longitudinal penumbra associated with X-ray scatter, the work here has utilized a

methodology that may benefit the development of new volume CT dosimetry standards, namely

involving small point dosimeters (Farmer chambers) and long cylindrical phantoms. In light of

these results, CTDIvol.e is still a reasonable dose estimate for the Aquilion ONETM

, but its

absolute interpretation should consider the systematic underestimation implied in the

measurements above.


48

CHAPTER IV: CHARACTERIZATION OF CT

RECONSTRUCTION FILTERS


49

4.1 Reconstruction Filters in CT

The selection of reconstruction filter is an important consideration in terms of both image noise

and spatial resolution. It may be selected after image acquisition and, in theory, repeated and

iterated upon at will; however, in practice the reconstruction is filter is usually fixed for a given

scan protocol and is not iterated upon due to computation time in repeating the 3D

reconstruction. Therefore, it is important to identify reconstruction filters that perform optimally

under low-dose imaging conditions. In this chapter, the spatial resolution associated with various

reconstruction filters is characterized in terms of the modulation transfer function (MTF).

The MTF was measured to determine the spatial frequency response associated with the seven

reconstruction filters, or “convolution kernels,” listed in Table 2.1 in Chapter II. The methods

detailed below expand on prior methodology established for 2D radiography and single-slice CT,

including a description of the experimental apparatus and setup as well as analysis of image data.

A means of measuring the MTF is demonstrated in which axial signal profiles from multiple

slices of a slightly angled wire are interleaved to effect an “over-sampled” line-spread function

(LSF). The implications of MTF on image quality and detectability are discussed in Chapter V.

4.2 Wire Phantom Scan

A simple wire phantom was constructed to determine the MTFs of various filters. The phantom

incorporated a 250 µm steel wire suspended in a 5 cm diameter hollow acrylic cylinder at a small

angle relative to the central axis of the cylinder to reduce partial volume averaging effects. The

wire was tightened and secured by Styrofoam end caps as displayed in Fig. 4.1 below.


50

FIG. 4.1. Thin wire phantom. A 250 µm steel wire was suspended inside hollow acrylic cylinder at a small angle

relative to the central axis of the cylinder.

The wire phantom was positioned on a cushion and manipulated relative to positioning lasers to

align the central axis of the cylinder approximately along the scanner z-axis. The phantom was

scanned at 100 kVp and 300 mA (the average beam energy used in this study and the highest

tube current) at a gantry rotation time of 0.35 s. Volume images were reconstructed at 1 mm slice

thickness and 1 mm slice interval for seven reconstruction filters (FC1, FC2, FC3, FC4, FC5,

FC11, FC50) available within the scanner reconstruction software. A single slice from the image

reconstructed using the FC1 filter is shown in Fig. 4.2.


51

(a)

(b)

(a)

(b)

FIG. 4.2. Wire phantom image. (a) Axial slice of wire phantom on foam cushion scanned at 100 kVp, 105 mAs and

reconstructed with FC1 convolution kernel at 1 mm slice thickness (W 1800 L-500). The central axis of the cylinder

was coincident with the z-axis of the scanner, and the wire was suspended at a small angle relative to the z-axis axis

to reduce partial volume effects between axial slices. (b) 55 x 55 pixel region of interest for Radon transform

calculation.

4.3 MTF Calculation

A total of 46 adjacent slices were used to calculate the MTF for each reconstruction filter. Line

Spread Functions (LSFs) from each slice were interleaved by shifting such that the centroids of

each slice were coincident so as to obtain an over-sampled LSF analogous to that obtained in the

Fujita40

angled-slit method in 2D radiography. The MTF was calculated from the Fourier

transform of the over-sampled LSF. The details of the analysis are described below, with the

FC1 filter taken as an example.


52

4.2.1 The Radon Transform, R(i)

For each slice, the LSF of a 2.8 cm x 2.8 cm (55 x 55 pixel) region of interest (ROI) located

within the phantom and containing the wire was computed by Radon transform of the ROI – i.e.,

summation of pixel values (HU) along columns in the ROI. The ROIs contained 55 pixel

columns with 55 pixels per column. Figure 4.3 shows the Radon transform superimposed over

the ROI of slice 1.

0 4 8 12 16 20 24 28 32 36 40 44 48 52 56

-54000

-53500

-53000

-52500

-52000

-51500

-51000

-50500

-50000

Pix

el C

olu

mn

Sum

(H

U)

Pixel Column Index

0 4 8 12 16 20 24 28 32 36 40 44 48 52 56

-54000

-53500

-53000

-52500

-52000

-51500

-51000

-50500

-50000

Pix

el C

olu

mn

Sum

(H

U)

Pixel Column Index

FIG. 4.3. Example 1D Radon transform of the wire image, superimposed over the corresponding ROI. Image

brightness was adjusted for the purpose of illustration.

4.2.2 Determination of the Wire Signal Region

The maximum value of the wire profile (Radon transform) was determined, and the

corresponding pixel column index (i = 28 in Fig. 4.3) was labelled the central pixel column of

the signal region. Upper and lower (left and right) bounds of the signal region were somewhat


53

arbitrarily chosen such that the signal region spanned ~7 mm (14 pixels) with the maximum

value at the center as illustrated in Fig. 4.4.

0 4 8 12 16 20 24 28 32 36 40 44 48 52 56

-500

0

500

1000

1500

2000

2500

3000

3500

4000

L

ine

Sp

read

Funct

ion (

HU

)

Pixel Column Index

Signal Region Upper

Bound

Signal Region Center

Signal Region Lower

Bound

0 4 8 12 16 20 24 28 32 36 40 44 48 52 56

-500

0

500

1000

1500

2000

2500

3000

3500

4000

L

ine

Sp

read

Funct

ion (

HU

)

Pixel Column Index

Signal Region Upper

Bound

Signal Region Center

Signal Region Lower

Bound

FIG. 4.4. Signal region for MTF calculation [as determined from the maximum value of the Radon transform R(i)].

The center of the signal region is represented by the middle dotted vertical line at pixel column index i = 28 and the

lower and upper boundaries of the signal region are represented as the left and right dotted vertical lines,

respectively.

4.2.3 Detrending the Non-Uniform Background

A quadratic fit was computed for the region outside the signal region (i.e., the background

region) and subtracted from the signal profile to eliminate the offset and background non-


54

uniformities in the Radon transform for each slice. Each profile was corrected by subtraction of a

quadratic fit of the form:

Pixel Column Sum = a*(i)2 + b*(i) + c (Eq. 4.1)

to the data [i,R(i)] in the air-only region (e.g., i = 0 � 20 and 33 � 55 in the example slice) to

remove background trends. Figure 4.5(a) illustrates the quadratic fit on the Radon transform of

the air-only region of slice 1. This fit was subtracted from the Radon transform to produce the

detrended profile in Fig. 4.5(b).

0 4 8 12 16 20 24 28 32 36 40 44 48 52 56

-54000

-53500

-53000

-52500

-52000

-51500

-51000

-50500

-50000

Pix

el C

olu

mn S

um

(H

U)

Pixel Column Index

0 4 8 12 16 20 24 28 32 36 40 44 48 52 56

-500

0

500

1000

1500

2000

2500

3000

3500

4000

Lin

e S

pre

ad F

uncti

on

(H

U)

Pixel Column Index

(a) (b)

0 4 8 12 16 20 24 28 32 36 40 44 48 52 56

-54000

-53500

-53000

-52500

-52000

-51500

-51000

-50500

-50000

Pix

el C

olu

mn S

um

(H

U)

Pixel Column Index

0 4 8 12 16 20 24 28 32 36 40 44 48 52 56

-500

0

500

1000

1500

2000

2500

3000

3500

4000

Lin

e S

pre

ad F

uncti

on

(H

U)

Pixel Column Index

0 4 8 12 16 20 24 28 32 36 40 44 48 52 56

-54000

-53500

-53000

-52500

-52000

-51500

-51000

-50500

-50000

Pix

el C

olu

mn S

um

(H

U)

Pixel Column Index

0 4 8 12 16 20 24 28 32 36 40 44 48 52 56

-500

0

500

1000

1500

2000

2500

3000

3500

4000

Lin

e S

pre

ad F

uncti

on

(H

U)

Pixel Column Index

(a) (b)

FIG. 4.5. Quadratic detrend of the line spread function. Profiles (a) before and (b) after the quadratic detrend. The

curve in (a) is the quadratic fit to the background region and (b) shows the background-subtracted profile.


55

4.2.4 Determination of the Signal Profile Centroid

The area under the signal region of the profile was calculated and the centroid was defined as

the pixel column index that evenly divides the area in two. It is therefore a “centroid” in that is

central to the signal power of the profile. In the example slice, the centroid is at column index

27.90.

0 10 20 30 40 50

-500

0

500

1000

1500

2000

2500

3000

3500

Lin

e S

pre

ad F

unct

ion

(H

U)

Pixel Column Index

0 4 8 12 16 20 24 28 32 36 40 44 48 52 56

-500

0

500

1000

1500

2000

2500

3000

3500

4000

Lin

e S

pre

ad F

un

ctio

n (

HU

)

Pixel Column Index

27.90

0 10 20 30 40 50

-500

0

500

1000

1500

2000

2500

3000

3500

Lin

e S

pre

ad F

unct

ion

(H

U)

Pixel Column Index

0 4 8 12 16 20 24 28 32 36 40 44 48 52 56

-500

0

500

1000

1500

2000

2500

3000

3500

4000

Lin

e S

pre

ad F

un

ctio

n (

HU

)

Pixel Column Index

0 10 20 30 40 50

-500

0

500

1000

1500

2000

2500

3000

3500

Lin

e S

pre

ad F

unct

ion

(H

U)

Pixel Column Index

0 10 20 30 40 50

-500

0

500

1000

1500

2000

2500

3000

3500

Lin

e S

pre

ad F

unct

ion

(H

U)

Pixel Column Index

0 10 20 30 40 50

-500

0

500

1000

1500

2000

2500

3000

3500

Lin

e S

pre

ad F

unct

ion

(H

U)

Pixel Column Index

0 4 8 12 16 20 24 28 32 36 40 44 48 52 56

-500

0

500

1000

1500

2000

2500

3000

3500

4000

Lin

e S

pre

ad F

un

ctio

n (

HU

)

Pixel Column Index

27.90

FIG. 4.6. Area under the profile (within the signal region) and determination of centroid position. The centroid was

determined as the pixel column index that divides the area under the signal region in two.

4.2.5 Application of Steps 1-4 for all Slices

For each of the 46 adjacent slices in the volume, the Radon transform was computed, signal

regions were segmented, profiles were detrended, and centroids were determined. In addition to

the profile for slice 1 described above, Fig. 4.7 shows the LSF for slice 40, for which the centroid


56

is located at pixel column index 26.22 (compared to 27.90 for slice 1), since the wire was placed

at an angle and thus shifts in the axial plane from slice to slice.

0 4 8 12 16 20 24 28 32 36 40 44 48 52 56

-500

0

500

1000

1500

2000

2500

3000

3500

4000

Lin

e S

pre

ad F

unct

ion (

HU

)

Pixel Column Index

Slice 1

Slice 40

0 4 8 12 16 20 24 28 32 36 40 44 48 52 56

-500

0

500

1000

1500

2000

2500

3000

3500

4000

Lin

e S

pre

ad F

unct

ion (

HU

)

Pixel Column Index

Slice 1

Slice 40

FIG. 4.7. Illustration of signal profiles for various image slices. Shown here are LSFs of slice 1 (right peak) and

slice 40 (left peak).

4.2.6 Align Centroids to Obtain an Over-Sampled LSF

The centroid of each profile was aligned with that of slice 1 to produce the over-sampled LSF

shown in Fig. 4.8. The resulting profile is exactly analogous to the over-sampled LSF obtained in

the conventional “angled-slit” technique of Fujita et al.41

for 2D radiography, adapted here to the

3D case of CT.


57

0 4 8 12 16 20 24 28 32 36 40 44 48 52 56

-500

0

500

1000

1500

2000

2500

3000

3500

4000

Over

sam

ple

d L

ine

Spre

ad F

unct

ion

(H

U)

Pixel Column Index

FIG. 4.8. Over-sampled line spread function. The over-sampled LSF was obtained by aligning centroids of 46

adjacent slices.

4.2.7 Fourier transform of the Over-sampled LSF

The data in the over-sampled LSF are not, in general, equally sampled, requiring interpolation of

the profile at equal intervals prior to Fourier transform A linear interpolation was performed

between the measured values of the over-sampled LSF to produce an uniformly-sampled, over-

sampled LSF. The area under the resulting LSF was normalized to unity, and the MTF was

obtained as the magnitude of the Fourier transform. Figure 4.9 shows the MTF for the FC1 filter

plotted up to the Nyquist frequency, which was determined by the equation

fNyquist = 1 / (2 * pixel size) (Eq. 4.2)

where the pixel size was 0.5 x 0.5 mm2.


58

Two assumptions were made in the above analysis. First, the linear interpolation of the over-

sampled signal profile is assumed to have negligible effect, since the data points are along the

profile are close (i.e., the angle of the wire is small). Second, the width of the wire (0.25 mm)

was neglected in the analysis. Each of these effects (i.e., the linear interpolation and the transfer

function of the wire) could be modeled analytically and divided out as a small correction to the

measured MTF. The effects are believed negligible for purposes of the current work.

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Modula

tion T

ransf

er F

unct

ion

Spatial Frequency (mm-1)

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Modula

tion T

ransf

er F

unct

ion

Spatial Frequency (mm-1)

FIG. 4.9. Modulation transfer function for FC1 filter. MTF is plotted up to the Nyquist frequency. A polynomial fit

is superimposed on the measured data as a simple guide to the eye in joining the discrete measurement points.


59

4.4 The MTF Associated with Various CT Reconstruction Filters

This algorithm was implemented for each of the seven reconstruction filters used in this study to

determine their spatial frequency responses. The MTFs corresponding to the seven

reconstruction filters are plotted in Fig. 4.10.

-1

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1

Spatial Frequency (mm )

MT

F

FC50

FC5

FC4

FC3

FC2

FC1

FC11

-1

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1


MT

F

FC50

FC5

FC4

FC3

FC2

FC1

FC11

-1

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1


MT

F

FC50

FC5

FC4

FC3

FC2

FC1

FC11

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1


MT

F

FC50

FC5

FC4

FC3

FC2

FC1

FC11

FIG. 4.10. Modulation Transfer Functions for all reconstruction filters in this study. MTFs were plotted up to the

Nyquist frequency. “Sharp” filters (FC50 and FC5) exhibit a higher pass characteristic than their “smooth”

counterparts (FC11, FC1). The former are typical of scan protocols for which the task is visualization of fine, high-

contrast details, whereas the latter are more appropriate to visualization of large, low-contrast soft tissues. Figure

adapted from Silverman et al. “Investigation of lung nodule detectability in low-dose 320-slice computed

tomography” with permission from publisher.

Filters FC1 and FC11 exhibit the smoothest (low-pass) characteristic, descending to 10% MTF at

~0.55 mm-1

. Filters FC2 through FC5 exhibit increasingly higher frequency response, with FC4

and FC5 exhibiting a slight edge-enhancement effect (i.e., MTF slightly greater than 1.0 at low


60

spatial frequencies). The FC50 filter displays the highest MTF, with a strong edge-enhancement

effect and descending to 10% MTF at ~0.94 mm-1

.

The blur associated with smoother filters suppresses noise at the cost of spatial resolution, while

the sharper filters improve spatial resolution but also amplify high-frequency noise. Selection of

reconstruction filter has a direct impact on image quality (signal-to-noise ratio) and success of

certain clinical diagnostic tasks such as detectability and characterization. The effect of filter

selection on image quality and detectability of small nodules is discussed in Chapter V.


61

CHAPTER V: LOW-DOSE LIMITS OF LUNG NODULE

DETECTABILITY


62

5.1 Experimental Parameters

The previous chapters outlined the tools, methodology, and overall background required for the

primary study, including the scanner, the anthropomorphic phantom, the observer test, analytic

techniques, dosimetry protocols, and MTF corresponding to each reconstruction filter. Observer

performance was evaluated based on the detectability of 3.2 mm nodules in the left lung of the

anthropomorphic phantom and documented in terms of the scanner-reported CTDIvol.e. The

threshold dose, Dthresh, below which diagnostic performance drops significantly (below Az =

0.95), was assessed as a function of body habitus, reconstruction filter, and reconstructed slice

thickness (tslice). A larger body habitus may be expected to reduce beam penetration and increase

scatter, while smooth filters and increased slice thicknesses may reduce image noise. The work

reported in this chapter examines human observer performance as a function of these

experimental variables.

5.2 Effect of Body Habitus on Detectability

Figure 5.1 illustrates measurements of Az versus dose for both the “average” and “obese” body

habitus. For the “average” habitus, we observe that detectability is maintained to a fairly low

dose, below which detectability rapidly declines. For the “obese” habitus, the rapid decrease in

detectability occurs at a significantly higher dose for a given set of reconstruction parameters.

To reduce the scope of experimental parameters investigated yet maintain the broad range of

dose under consideration, it was found that results obtained at various kVp could be pooled

without significantly affecting the resulting curve fits or Dthresh values. Of course, image quality

in the range 80 – 120 kVp would be expected to vary somewhat with beam quality (e.g., due to


63

beam-hardening effects), but the curve fits of Az versus dose were not significantly affected by

choice of kVp. Specifically, removal of data points of specific beam energies from the fit did not

result in considerable shifts in the curve or in Dthresh. Therefore, data resulting from 80, 100, and

120 kVp were pooled in single curve fits of Az versus dose for a given body habitus,

reconstruction filter, and slice thickness. In addition, two scans were rejected as outliers from the

analysis – (80 kVp, 105 mAs) and (80 kVp, 35 mAs) – as they exhibited a significant ring-like

artifact in the image that interfered with visibility of the nodule.

0.1 1 10

0.5

0.6

0.7

0.8

0.9

1.0

Az

CTDIvol.e (mGy)

Dthresh

aveDthresh

obese

Average

Obese

[FC3]

[tslice = 3 mm]

0.1 1 10

0.5

0.6

0.7

0.8

0.9

1.0

Az

CTDIvol.e (mGy)

Dthresh

aveDthresh

aveDthresh

obeseDthresh

obese

Average

Obese

[FC3]

[tslice = 3 mm]

FIG. 5.1. Observer performance measured as a function of dose for average and obese body habitus. Logistic

functions were used to fit the measurement to a sigmoid (solid curves). Calculation of Dthresh is illustrated

graphically as the dose at which observer performance falls to 0.95. Examples shown are for fixed reconstruction

filter (FC3) and slice thickness (tslice = 3 mm). Figure adapted from Silverman et al. “Investigation of lung nodule

detectability in low-dose 320-slice computed tomography” with permission from publisher.


64

At the dose level ave

threshD where Az for the average habitus (ave

zA ) was 0.95, obese

zA was only 0.6,

indicating a nearly undetectable nodule. For the FC3 reconstruction filter obese

threshD was

approximately seven times greater thanave

threshD . Averaging over all reconstruction settings, obese

threshD

was found to be 8.6 times higher thanave

threshD , illustrating the significantly reduced observer

performance at low doses for the obese habitus.

5.3 Effect of Reconstruction Technique

5.3.1 Image Quality for Various Reconstruction Filters and Slice Thickness

The MTFs corresponding to the seven reconstruction filters used in this study were derived in

Chapter IV and are plotted in Fig. 4.10. The blur associated with smoother filters suppresses

noise, while the sharper filters amplify high-frequency noise. This is illustrated in Fig. 5.2, in

which the regions of interest (ROIs) for all reconstruction settings are displayed for the obese

body habitus.


65

FC3FC1 FC2 FC4 FC5 FC50FC11

1mm

3mm

5mm


1mm

3mm

5mm


1mm

3mm

5mm

FIG. 5.2. Example ROIs about a 3.2 mm nodule for each reconstruction filter and slice thickness investigated.

Examples were acquired at 100 kVp, 105 mAs (CTDIvol.e = 6.7 mGy) in the obese phantom configuration. For

purposes of illustration, the nodule is shown at the center of each image. Figure adapted from Silverman et al.

“Investigation of lung nodule detectability in low-dose 320-slice computed tomography” with permission from

publisher.

As expected, for a given slice thickness, smoother filters produce images with less noise, and for

a given filter, high-frequency fluctuations are reduced with slice thickness. Furthermore, at tslice =

5 mm contrast is noticeably reduced due to partial volume averaging of the small (3.2 mm

diameter) simulated nodule.

5.3.2 Effect of Reconstruction Filter on Detectability

The effect of reconstruction filter on observer performance is shown in Fig. 5.3(a). Figure 5.3(b)

presents the corresponding Dthresh for each filter.


66

FC3

FC1

FC11

FC2

FC5

FC4

FC50

0.1 1 10

0.5

0.6

0.7

0.8

0.9

1.0

Az

CTDIvol.e (mGy)

FC1 FC2 FC3 FC4 FC5 FC11 FC500

1

2

3

4

5

Det

ecta

bil

ity

Th

resh

old

(m

Gy

)

Reconstruction Filter

tslice = 3 mm

[Obese]

tslice = 3 mm

[Obese]

(a) (b)


FC3

FC1

FC11

FC2

FC5

FC4

FC50

0.1 1 10

0.5

0.6

0.7

0.8

0.9

1.0

Az

CTDIvol.e (mGy)


1

2

3

4

5

Det

ecta

bil

ity

Th

resh

old

(m

Gy

)


tslice = 3 mm

[Obese]

tslice = 3 mm

[Obese]

(a) (b)


FIG. 5.3. Effect of reconstruction filter on detectability and Dthresh (tslice = 3 mm; obese configuration). (a) Observer

performance plotted as a function of dose for seven reconstruction filters. (b) Comparison of Dthresh across seven

reconstruction filters. Figure adapted from Silverman et al. “Investigation of lung nodule detectability in low-dose

320-slice computed tomography” with permission from publisher.

Figure 5.3 suggests that for tslice = 3 mm and obese habitus, images reconstructed with the FC3

filter exhibited the best performance, followed closely by the smooth FC1 and FC11 filters. This

suggests that at low dose, noise becomes the limiting factor in the detection of small nodules.

The filters FC2, FC4, and FC5 differed only slightly in detection performance, whereas FC50

performed considerably worse than all other filters. These results demonstrate that

knowledgeable selection of reconstruction filters can reduce the dose required to detect nodules

in obese patients (e.g., by up to 2 mGy for the case in Fig. 5.3).

5.3.3 Effect of Reconstruction Slice Thickness on Detectability

The value Dthresh was found to depend also on slice thickness. Figure 5.4 shows an increase in

Dthresh with larger slice thickness for both the average and obese phantom configurations


67

reconstructed with FC11. While such a clear trend was not evident for all filters, the 5 mm slice

reconstructions generally resulted in poorer diagnostic performance (i.e., higher Dthresh), and the

1 mm and 3 mm slices exhibited comparable performance for the 3.2 mm nodule detection task.

Obese

Average

[FC11]

1 3 5

0.0

0.4

0.8

1.2

1.6

2.0

Det

ecta

bil

ity

Th

resh

old

(m

Gy

)

Slice Thickness (mm)

Obese

Average

[FC11]

1 3 5

0.0

0.4

0.8

1.2

1.6

2.0

Det

ecta

bil

ity

Th

resh

old

(m

Gy

)

Slice Thickness (mm)

FIG. 5.4. Effect of slice thickness on Dthresh for average and obese habitus. The cases shown correspond to the FC11

reconstruction filter. Figure adapted from Silverman et al. “Investigation of lung nodule detectability in low-dose

320-slice computed tomography” with permission from publisher.

5.3.4 Statistical Comparison of Reconstruction Techniques

The effect of all reconstruction filters and slice thicknesses on observer performance is

summarized in Fig. 5.5 for both the average and obese body habitus.


68


0.0

0.1

0.2

0.3

0.4

0.5D

etec

tab

ilit

y T

hre

sho

ld (

mG

y)


tslice = 1mm

tslice = 3mm

tslice = 5mm


0

1

2

3

4

5 tslice = 1mm

tslice = 3mm

tslice = 5mm


[Average] [Obese]

(a) (b)tslice= 1 mm

tslice= 3 mm

tslice= 5 mm

tslice= 1 mm

tslice= 3 mm

tslice= 5 mm

tslice= 1 mm

tslice= 3 mm

tslice= 5 mm

tslice= 1 mm

tslice= 3 mm

tslice= 5 mm

tslice= 1 mm

tslice = 3 mm

tslice= 5 mm

tslice= 1 mm

tslice = 3 mm

tslice= 5 mm

tslice= 1 mm

tslice = 3 mm

tslice= 5 mm

tslice= 1 mm

tslice = 3 mm

tslice= 5 mm


0.0

0.1

0.2

0.3

0.4

0.5D

etec

tab

ilit

y T

hre

sho

ld (

mG

y)


tslice = 1mm

tslice = 3mm

tslice = 5mm


0

1

2

3

4

5 tslice = 1mm

tslice = 3mm

tslice = 5mm


[Average] [Obese]

(a) (b)tslice= 1 mm

tslice= 3 mm

tslice= 5 mm

tslice= 1 mm

tslice= 3 mm

tslice= 5 mm

tslice= 1 mm

tslice= 3 mm

tslice= 5 mm

tslice= 1 mm

tslice= 3 mm

tslice= 5 mm

tslice= 1 mm

tslice = 3 mm

tslice= 5 mm

tslice= 1 mm

tslice = 3 mm

tslice= 5 mm

tslice= 1 mm

tslice = 3 mm

tslice= 5 mm

tslice= 1 mm

tslice = 3 mm

tslice= 5 mm

FIG. 5.5. Effect of reconstruction techniques on Dthresh for (a) average and (b) obese body habitus. Note the order-

of-magnitude scale factor between y-axes of (a) and (b), discussed below. Lower values of Dthresh correspond to

improved low-dose detectability. Figure adapted from Silverman et al. “Investigation of lung nodule detectability in

low-dose 320-slice computed tomography” with permission from publisher.

To compare performance among various cases, two-tailed, unequal variance paired t-tests were

conducted across all filter and slice thickness combinations. Table 3.1 summarizes a subset of the

p-values from these tests, with dark grey cells indicating a statistically significant difference in

performance at 95% confidence level.

The loss in detectability at larger body habitus is evidenced by an average increase in Dthresh by a

factor of 8.6 ± 2.8 for equivalent reconstruction settings. Even conditions yielding the worst

Dthresh for the average body demonstrated significantly improved performance compared to the

obese habitus (a factor of ~2.5 in Dthresh, p = 0.006). Thus, detectability is significantly reduced

for the obese habitus, and the dose required to achieve Az~0.95 differs by a factor of ~6-16

(depending on reconstruction technique). The higher level of statistical significance observed for


69

the average habitus results (i.e., lower p-values in Table III) are due to reduced observer

variability in nodule detection. The high noise and low detectability conditions of the obese

habitus increased the σDthresh and thus the p-values of comparison.

FC1 FC2 FC3 FC4 FC5 FC11 FC50 FC1 FC2 FC3 FC4 FC5 FC11 FC50

FC1 - - - - - - - - - - - - - -

FC2 0.000 - - - - - - 0.408 - - - - - -

FC3 0.009 0.049 - - - - - 0.480 0.363 - - - - -

FC4 0.000 0.140 0.010 - - - - 0.479 0.376 0.496 - - - -

FC5 0.001 0.220 0.155 0.044 - - - 0.389 0.500 0.308 0.344 - - -

FC11 0.053 0.000 0.000 0.000 0.000 - - 0.146 0.185 0.054 0.098 0.080 - -

FC50 0.000 0.000 0.000 0.000 0.000 0.000 - 0.017 0.010 0.012 0.015 0.007 0.002 -

FC1 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.263 0.337 0.165 0.212 0.264 0.262 0.004

FC2 0.004 0.000 0.000 0.000 0.000 0.075 0.000 0.298 0.210 0.276 0.300 0.166 0.043 0.036

FC3 0.292 0.000 0.004 0.000 0.000 0.197 0.000 0.190 0.244 0.080 0.135 0.125 0.357 0.003

FC4 0.008 0.260 0.201 0.059 0.489 0.001 0.000 0.267 0.203 0.255 0.270 0.178 0.072 0.069

FC5 0.002 0.491 0.074 0.148 0.251 0.000 0.000 0.240 0.169 0.218 0.239 0.134 0.040 0.058

FC11 0.012 0.135 0.330 0.027 0.330 0.001 0.000 0.301 0.381 0.224 0.258 0.339 0.258 0.006

FC50 0.003 0.093 0.344 0.018 0.265 0.000 0.000 0.043 0.031 0.038 0.042 0.026 0.012 0.486

FC1 0.434 0.001 0.016 0.000 0.001 0.174 0.000 0.171 0.125 0.157 0.170 0.106 0.042 0.131

FC2 0.002 0.169 0.022 0.499 0.068 0.000 0.001 0.141 0.107 0.131 0.141 0.094 0.044 0.207

FC3 0.008 0.121 0.032 0.314 0.065 0.004 0.038 0.254 0.190 0.240 0.256 0.164 0.063 0.070

FC4 0.000 0.347 0.013 0.209 0.104 0.000 0.000 0.099 0.074 0.091 0.098 0.064 0.029 0.284

FC5 0.001 0.040 0.007 0.157 0.017 0.000 0.097 0.055 0.032 0.040 0.050 0.020 0.005 0.223

FC11 0.003 0.133 0.023 0.392 0.060 0.001 0.007 0.425 0.319 0.426 0.440 0.272 0.065 0.018

FC50 0.003 0.026 0.008 0.072 0.015 0.002 0.451 0.016 0.009 0.011 0.014 0.005 0.001 0.434

1mm

AVERAGE BODY HABITUS OBESE BODY HABITUS

1mm

1mm

3mm

5mm

Table 3.1. Summary of p-values from paired t-tests for comparison of reconstruction techniques. This table

compares distributions with mean Dthresh and standard deviation σDthresh. Each p-value corresponds to the pair of

conditions specified by the rows and columns of the table, and gray cells correspond to cases for which a statistically

significant difference was observed (p-value < 0.05). Cases summarized here compare tslice = 1 mm to tslice = 1, 3,

and 5 mm for all filters and for average and obese habitus. [Similar analysis of Dthresh for tslice = 3 and 5 mm (not

shown) was also conducted.]

In addition to the effects of body size, Fig. 5.5 and Table 3.1 reveal the effects of reconstruction

slice thickness. For the average body habitus, tslice = 5 mm yielded poorer detectability than tslice

= 3 mm for the FC3, FC5, and FC11 filters (p < 0.042), and tslice = 3 mm demonstrated improved

detectability compared to tslice = 1 mm for FC1, FC2, FC3, and FC50 (p < 0.010). For the obese

habitus, the Dthresh for tslice = 1 mm and FC11 (second-to-last column of Table 3.1) is


70

significantly lower than that of FC2, FC5, and FC50 at tslice = 3 mm (p < 0.043), but also filters

FC1 and FC4 at tslice = 5 mm (p < 0.044). Although the results do not exhibit a statistically

significant difference between Dthresh values for tslice = 1 mm and 3 mm reconstructions for obese

body habitus, the trends clearly demonstrate inferior performance for tslice = 5 mm for both body

types.

While results indicate improved performance for tslice = 1 mm and 3 mm, they also demonstrate

the effects of reconstruction filter on detectability for these slice thicknesses. FC50 clearly

demonstrated poorer performance compared to other filters, as evident in Fig. 5.5 and throughout

Table III. For example, for the average body habitus at tslice = 1 mm, FC50 exhibited a higher

Dthresh than all other filters at both slice thicknesses (p < 0.001). For tslice = 3 mm, the smoother

filters (FC1, FC2 and FC3) performed better than the sharper ones (FC4, FC5, and FC50). For

the average body habitus, optimal performance was obtained using the smooth filters: FC1 and

FC2 at tslice = 3 mm followed by FC1 and FC11 at tslice = 1 mm. For the obese habitus, the FC50

filter at tslice = 1 mm was significantly worse than FC1, FC2, FC3, and FC11 (p < 0.036) for both

tslice = 1 and 3 mm reconstructions. At tslice = 3 mm, FC3 demonstrated optimal performance,

followed by FC1 and FC11. At tslice = 1 mm, the smooth filters did not differ from one another

significantly, although FC11 exhibited the lowest Dthresh.


71

CHAPTER VI: DISCUSSION, CONCLUSIONS, AND FUTURE

DIRECTIONS


72

6.1 Factors Affecting Image Quality: Acquisition, Reconstruction Techniques,

and Body Habitus

To evaluate diagnostic performance, lung nodule detectability was characterized in terms of Az

measured as a function of CTDIvol.e. Because CTDI characterizes dose specifically within an

acrylic 32 cm diameter phantom, it does not account for body size. One reason that the Dthresh

values increase by nearly an order of magnitude (8.6-fold) for the obese habitus is illustrated in

Fig. 6.1(a) and (b). The obese habitus image in Fig. 6.1(b) is significantly degraded by high

noise, blurred edges and reduced contrast. These scans involved an identical CTDIvol.e, but the

actual dose measured in the lung (Dlung) varied by a factor of two (3.6 mGy and 1.8 mGy for the

average and obese configurations, respectively). The reduced dose to the lung (i.e., higher

attenuation by surrounding fat) begins to explain the poor image quality. However, even at

similar lung dose, as in Fig. 6.1(c) and (d), we can see this is not the only factor; when

comparing detectability as a function of Dlung, lung

threshD [calculated analogous to Dthresh, but on a

plot of Az versus Dlung (instead of CTDIvol.e)] was still 2.5 ± 0.8 times higher for the obese

habitus.


73

Similar CTDIvol.e Similar Dlung

(a) (b) (c) (d)

Average

CTDIvol.e = 2.2 mGy

Dlung

= 3.6 mGy

Obese

CTDIvol.e = 2.2 mGy

Dlung

= 1.8 mGy

Average

Dlung = 1.4 mGy

CTDIvol.e = 0.8 mGy

Obese

Dlung = 1.5 mGy

CTDIvol.e = 2.0 mGy

Similar CTDIvol.e Similar Dlung

(a) (b) (c) (d)

Average

CTDIvol.e = 2.2 mGy

Dlung

= 3.6 mGy

Obese

CTDIvol.e = 2.2 mGy

Dlung

= 1.8 mGy

Average

Dlung = 1.4 mGy

CTDIvol.e = 0.8 mGy

Obese

Dlung = 1.5 mGy

CTDIvol.e = 2.0 mGy

FIG. 6.1. Images at similar reported and measured dose. Example images at similar CTDIvol.e (a,b) and measured

dose, Dlung (c,d) for average (a,c) and obese (b,d) configurations (reconstructed with the FC3 filter at tslice = 3 mm).

Images (a,b) were acquired at (100 kVp, 35 mAs), giving CTDIvol.e = 2.2 mGy and measured doses of (a) Dlung =

3.6 mGy and (b) Dlung = 1.8 mGy. By comparison, images (c,d) were acquired at (c) Dlung = 1.4 mGy, CTDIvol.e =

0.8 mGy (120 kVp, 7 mAs) and (d) Dlung = 1.5 mGy, CTDIvol.e = 2.0 mGy (120 kVp, 17.5 mAs). Figure adapted

from Silverman et al. “Investigation of lung nodule detectability in low-dose 320-slice computed tomography” with

permission from publisher.

Several factors contribute to lower detectability in large patients. The well-known relationship41

sliceo

d

t

K

D

e 2

µ

σ ∝ (Eq. 6.1)

describes the dependence of image noise on the dose at the center of a cylindrical phantom (Do)

and includes the attenuation of the object (eµd

) as well as slice thickness (tslice) and reconstruction

filter (contained in the bandwidth integral K). Even with all reconstruction parameters held fixed,

the dose delivered to the patient (e.g., Dlung measured in the organ) does not determine the noise;


74

rather, the noise depends on the dose to the detector, Ddetector, which depends on body habitus

[related to d in Eq. (8)] and is less than Dlung due to attenuation in the body. In a cylindrical

phantom, the relationship can be considered as σ2 α 1/Ddetector, where Ddetector = Dcenter e

-µd/2. For

larger patients, Ddetector is significantly reduced, and the noise increases (even for equivalent

Dlung),42

an effect evident in Fig. 6.1.

In addition, larger body habitus amplifies the effects of X-ray scatter, beam hardening, and

electronics noise. The additional 10 cm simulated tissue (SuperFlabTM

) increases the scatter-to-

primary ratio at detector, diminishing contrast and introducing cupping and streak artifacts.

Similarly, beam hardening increases with the bulk of material presented to the X-ray beam,

which may lead to cupping or streak artifacts, although such effects were not particularly evident

in the small ROIs considered in this study. Finally, the reduced X-ray fluence at the detector

increases the relative contribution of electronic noise at very low dose. Quantum and electronic

noise are proportional to 1/ D and 1/D, respectively,43

so whereas CT is typically quantum

noise limited, at low technique settings (kVp and mAs) and large body habitus the relative effect

of electronic noise rises.

In the results above, CTDIvol.e was used to analyze Dthresh, below which detectability rapidly

declines. The correspondence of CTDIvol.e and the CTDIw and Dcenter measured using Farmer

chambers in a long 32 cm diameter acrylic cylinder is given in Fig. 3.3 in Chapter III. Results

indicate that to achieve a high level of diagnostic accuracy (i.e., Az > 0.95 as in current clinical

techniques), the CTDIvol.e differed significantly – e.g., by a factor of ~8.6 between the average

and obese configurations modeled in this study. Results also demonstrate the potential for further

dose reduction in average-sized patients given knowledgeable choice of reconstruction

technique.


75

6.2 Optimal Reconstruction Technique Selection

Radiologists consider numerous complex characteristics, including shape, size, and contrast, in

searching for nodules. Small lung nodules of clinical consequence are abstractly considered to be

spheres of minimum diameter 3 mm and contrast ~600-700 HU to the background, as simulated

in this phantom study. Proper selection of reconstruction techniques may preserve the ability to

visualize these attributes, which become degraded by noise at low dose.

Selection of the reconstruction filter governs the spatial frequency response in the axial plane,

with smooth filters suppressing noise at the cost of spatial resolution. The results above suggest

that at very low doses, it is quantum noise (rather than spatial resolution) that limits detectability.

This is evidenced by the downward trend in Dthresh for progressively smoother filters. It explains

the poor performance of the FC50 sharp filter and the superior performance of FC1 and FC11 in

the obese and average configurations. For the obese habitus, FC3 also performed well, perhaps

because it provided a balance between noise reduction and edge-preservation. Although sharp

filters have become increasingly popular due to their ability to enhance visualization of fine

anatomical structure, in the dose-conscious present it is important to recognize the limitations of

sharp filters and focus on techniques that allow noise reduction for specific diagnostic tasks, such

as small-nodule detection and surveillance.

Similarly, the effect of slice thickness should be carefully considered in identifying low-dose

techniques. Reconstruction slice thickness affects not only the image noise, but also the contrast

of small nodules (due to partial volume averaging effects) and potentially a perceived loss of the

characteristic spherical shape. As evident in Fig. 6.2, large slice thicknesses smooth the noise but

can also decrease nodule contrast to a significant degree. The results above suggest inferior

performance for tslice = 5 mm compared to 1 and 3 mm, likely due to this loss of contrast.


76

(a) (b) (c)(a) (b) (c)

FIG. 6.2. Axial images of 3.2mm nodule reconstructed at varying tslice. Image in left lung acquired at 2.2 mGy,

obese habitus, FC3 reconstruction filter. Reconstruction slice thickness @ interval: (a) 1 mm @ 1 mm. (b) 3 mm @

1.5 mm. (c) 5 mm @ 2.5 mm.

The results are consistent with the notion that to optimize detectability in lung CT, the slice

thickness should be equal to or less than the minimum lesion size of interest. Results showed no

significant loss in detectability when tslice = 3 mm compared to when tslice = 1 mm in the obese

configuration. Combined with the higher performance of smoother reconstruction filters, this

suggests a tradeoff between the larger partial volume averaging effect and noise reduction for

nodules smaller than the slice thickness. For other non-pulmonary thoracic imaging tasks, such

as identification of lymphomas or metastases of the mediastinum, involving structures exceeding

3 mm, radiologists in clinical practice enjoy the efficiency of larger slice thickness as it reduces

the total number of slices without sacrificing diagnostic performance. However, results indicate

that to maintain nodule detectability in low-dose CT, the slice thickness should be no greater

than 3 mm, that is, at or below the minimum nodule size of interest.


77

The optimal reconstruction factors identified above yielded Dthresh values lower than current

clinical low-dose (5 mGy) and ultra low-dose (1 mGy) protocols. For average body habitus,

Dthresh was 0.23 ± 0.08 mGy. The highest (i.e., worst) value of Dthresh was 0.40 mGy for the sharp

FC50 filter at tslice = 1 mm, the noisiest combination of reconstruction techniques. These values

are significantly below the lowest dose protocols at our institution. For the obese habitus, Dthresh

was 1.90 ± 0.71 mGy. The worst case gave Dthresh = 3.28 mGy. While these values of Dthresh are

within the range of clinical low-dose protocols, only the best reconstruction technique selections

approached the ultra low-dose levels.

The clinical impact of such dose reduction is significant. Not only do ultra low-dose techniques

permit screening initiatives for high-risk populations, but upon detection of a suspicious lesion,

surveillance could be conducted at more frequent intervals (i.e., more than once annually as the

current clinical protocol suggests for a nodule smaller than 5 mm in diameter). A malignant

tumour can grow a great degree and progress through several disease stages over the course of a

year, so more frequent testing is valuable in the characterization of the disease, as well as the

treatment and prognosis.

6.3 Limitations of the Current Study

No study is without its limitations, and those associated with the current study should be

acknowledged. First and foremost, the study is based on a phantom lacking the rich anatomical

complexity of human lung. Second, the imaging task was specified and unvarying – detection of

a simulated small solid nodule – analogous to surveillance of a known nodule, but an

oversimplification of complex diagnostic tasks. Studies have demonstrated that nodules larger

than 3.2 mm in diameter may also be difficult to detect but such nodules were not considered


78

here.31

Thirdly, the 9AFC tests, though well suited to phantom studies involving a large number

of cases, do not reproduce the complexity of a true diagnostic search in a real chest image.

Furthermore, observers may have used different criteria to guide their 9AFC selections, such as

edge detection, shape or contrast, which may also be true in the clinic, but hinders our ability to

evaluate the effect of each criterion separately on detectability. Another difference from the

clinical scenario is the absence of motion artifacts, which are reduced, but not eliminated, in

volumetric CT with high gantry rotation speed. The cone-beam artifact may also affect the

resolution of nodules differentially based on their location in the beam; however, the nodules in

this study were in fairly close proximity (within a few cm in the z-direction) to the central plane,

so the effects of this artifact on detectability were probably small. Another limitation is the

inability to scroll adjacent slices in viewing suspicious structures in the ROIs. These limitations

considered together likely resulted in higher observer performance (i.e., lower Dthresh) than might

be expected in the clinic; however, the overall trends observed are expected to hold –

specifically, the reconstruction parameters (filter and slice thickness) leading to optimal

performance and the significant difference in detectability limits (a factor of ~8.6 in dose)

between average and obese body habitus.

6.4 Conclusions

Knowledgeable selection of acquisition and reconstruction techniques can reduce the dose

required for accurate detection of small lung nodules. The low-dose threshold, Dthresh, for nodule

detection in the obese body size increased in comparison to the average body size by a factor of

8.6 (in terms of CTDIvol.e) and 2.5 (in terms of absolute dose measured in the lung). Results


79

suggest that images are degraded in the larger habitus mainly due to increased noise (reduced

fluence reaching the detector) rather than from scatter, beam hardening, or spatial resolution

effects. To maintain a high level of detectability across a spectrum of patient sizes, it is necessary

to knowledgeably adjust the dose and reconstruction parameters. The low-dose detectability

limits identified in this study suggest that current clinical low-dose protocols [e.g., 120 kVp, 25

mAs, CTDIvol.e = 1.0 mGy] are appropriate for large patients and suggest the potential for

further dose reduction in average sized patients.

Proper selection of reconstruction filter and slice thickness was found to maintain detectability at

low doses. While sharp filters (e.g., FC50) demonstrated high MTF, performance in nodule

detection was significantly degraded at low doses. Smoother filters (e.g., FC1, FC3, and FC11)

performed best, reducing the threshold for detection by up to a factor of ~3 (e.g., for the obese

habitus and tslice = 3 mm, Dthresh = 1.1 mGy for FC3 compared to 3.2 mGy for FC50). Thicker

slices also suppressed image noise, but for slice thickness exceeding the nodule diameter,

detectability was degraded due to a loss in nodule contrast (partial volume averaging effect). For

a 3.2 mm diameter nodule, a slice thickness of 1 mm or 3 mm slices was superior to 5 mm by a

factor of ~2 in Dthresh (e.g., for the obese habitus and FC3 filter, Dthresh = 1.1 mGy for tslice = 3

mm, compared to 1.9 mGy for tslice = 5 mm). An empirical and theoretical understanding of the

dependence of lung nodule detectability on dose, reconstruction parameters, and patient size

facilitates the implementation of ultra low dose imaging protocols that are increasingly patient-

specific (particularly with respect to body habitus) and maintain a high level of diagnostic

accuracy in the detection, characterization, and surveillance of early stage lung nodules.


80

APPENDICES


81

A.1. Observer Test Instructions

The following instructions were given to each observer consistently with each sitting of each

MAFC test.

INTRODUCTION:

1) This is part one of a two-part study

2) Today you will perform one study which will involve looking at 1,260 sets of images.

This will be preceded by a training session to familiarize you with the test and the

imaging task.

3) Each set consists of nine images presented in a 3 x 3 grid.

4) One of the nine images presented will contain a nodule and the rest will present normal

simulated lung tissue. For all trials, the nodule is a spherical mass of constant size with

contrast brighter than the background and located near the center of the image.

5) You will be asked to determine which one of the nine images contains the nodule.

6) When you have chosen the image you suspect contains the nodule, use the mouse to

move the cursor and double-click in the region of the image.

7) After you double-click, it may take a moment for the next set of images to appear.

8) If you are unsure of which image most likely contains the nodule, take your best guess.

9) Don’t worry if you make a mistake. Please continue until the study is complete.

10) Do not click outside of the image. A good habit is to return the cursor to the center of the

grid between clicks. Otherwise, the cursor may be covering the nodule.

11) You will be timed but there is no need to rush. Try to maintain a fairly consistent pace in

image selection and avoid distractions during the study.


82

12) You are allowed to adjust your viewing position but do not adjust the brightness of the

screen or the image size.

13) Do you have any questions?

TRAINING:

1) The training will involve looking at 63 sets of images.

2) For the training session, the correct choice is provided after each case. If you select the

correct image region, text reading ‘CORRECT’ will appear in this region. If you select

the incorrect image region, text reading ‘INCORRECT’ will appear in the region with the

nodule. The solutions will remain on the screen for two to three seconds, and the next set

of images will then appear.

3) Results of the training session will not be recorded.

4) The training session will take approximately five minutes.

5) You may not take a break during the training session.


7) Let’s begin the training session.

(Observer performs training session)

PART I:

1) The study will be divided into three parts, each containing 420 sets of images and each

involving the same basic imaging task.

2) Solutions will not be shown after each case.

3) The results will be recorded. Do your best to remain focused and to maintain a fairly

consistent pace in selection.


83

4) The study should take approximately one hour and forty five minutes.

5) You may take two five-minute breaks one third and two thirds of the way through the

study. The investigator will let you know when it is time for a break.


7) Let’s begin the first part of the study.

(Observer performs Part I of study)

(Five-minute break)

PART II:

1) Let’s begin the second part of the study.

(Observer performs Part II of study)

(Five-minute break)

PART III:

1) Let’s begin the third part of the study.

(Observer performs Part III of study)


84

A.2. Calculation of BMI

The goal of this calculation was to estimate the BMI of the anthropomorphic phantom illustrated

in Chapter II (missing arms and legs) relative to a full-sized human. The BMI was estimated for

two body types, namely average and obese.

Phantom 1: Average Body Habitus

The following basic approximations and calculations were performed to estimate the BMI:

- The torso and limbs were idealized as cylinders.

- The head was idealized as a sphere of diameter (Rhead)

- The existing limb lengths (lleg, larm) and circumferences (cleg, carm) were measured on the

phantom.

- The extended limb lengths and diameters were estimated from a quick sampling of three

average-sized volunteers: Lleg, Larm, Ltorso, Dleg, Darm, Dtorso

- The phantom height height (H) was measured = Lleg + Ltorso + Dhead = 170.4 cm = 1.704 m

- The phantom volume (Vave) was computed as:

Vave = 2*( π *Dleg2 * Lleg / 4) + 2* (π *Darm

2 * Larm / 4) + π * Dtorso

2 * Ltorso / 4 +

(4/3) * π * Rhead3

to be: 64646.97 cm3 = 0.0646 m

3

- The body density (ρbody) was taken equal to 1000 kg / m3, the density of water (an

approximation of human body density).

- The phantom mass (M) was computed as Vave * ρbody = 64.6 kg

- The BMI is therefore:

BMI = M / H2 = 64.6 / 1.704

2 = 22.3 kg/m

2


85

Phantom 2: Obese Body Habitus

- A layer of 5 cm of SuperFlab was placed on either side of the phantom’s chest (10cm

total). This increased the torso thickness Dtorso from Dtorso.ave = 20 cm to Dtorso.ave = 30 cm.

- For the lower limit on BMI estimation, it was assumed that all fat is localized on the torso

(no fat on the arms and legs), and the ratio of Dtorso.fat: Dtorso.ave = 1.5

- For the upper limit on BMI estimation, the ratio of Dtorso.fat: Dtorso.ave = 1.5 was applied to

the arms and legs as well (idealized as cylinders) – i.e., to the whole body, excluding the

head.

- Obese Habitus BMI: Lower Limit

o The diameter of the obese torso (Dtorso.fat) was estimated by:

Dtorso.fat = Dtorso.ave * 1.5 = 30 cm

o The volume of the obese torso (Vtorso.fat) was determined by:

Vtorso.fat = π * Dtorso.fat2

* Ltorso / 4 = 0.0611 m3

o The volume of the flab on the torso (Vflab.torso) was determined by:

Vflab.torso = Vtorso.fat – Vtorso.ave = 0.06110m3

- π *Dtorso2 * Ltorso / 4 = 0.06110 -

0.02716 = 0.03394 m3

o The density of adipose tissue (ρfat) was approximated as 0.9g/cm3 = 900 kg/m

3

o The mass of the fat on the torso (Mflab.torso) was estimated by:

Mflab.torso = Vflab.torso * ρfat = 30.546 kg

o The lower limit of mass (Mlower.fat) was calculated by:

Mlower.fat = M + Mflab.torso = 64.6 + 30.5 = 95.1 kg

o The lower limit of BMI (BMIlower) was therefore:


86

BMIlower = Mlower.fat / H2

= 95.1 / 1.7042 = 32.8 kg/m

2

If the SuperFlabTM

were modeled instead as muscle:

o The density of muscle adipose tissue (ρmuscle) was approximated as 1.06 g/cm3 =

1060 kg/m3

o The mass of the muscle tissue on the torso (Mflab.torso) was estimated by:

Mflab.torso = Vflab.torso * ρmuscle = 35.976 kg

o The lower limit of mass (Mlower.fat) was calculated by:

Mlower.fat = M + Mflab.torso = 64.6 + 35.98 = 100.58 kg

o The lower limit of BMI (BMIlower) was therefore:

BMIlower = Mlower.fat / H2

= 100.58 / 1.7042 = 34.6 kg/m

2

- Obese Habitus BMI: Upper Limit

o The diameter of the obese limbs were calculated as:

Dleg.fat = Dleg.ave * 1.5 and Darm.fat = Darm.ave * 1.5.

o The volume of the large phantom body (Vfat) was determined by:

Vfat = 2*(π*Dleg.fat2 * Lleg / 4) + 2* (π *Darm.fat

2 * Larm / 4) + Vtorso.fat + Vhead =

0.1401 m3

o The volume of fat (Vflab) was determined by:

Vflab = Vfat - Vave = 0.1401 – 0.0646 = 0.0755 m3

o The density of adipose tissue (ρfat) was approximated as 0.9g/cm3 = 900 kg/m

3

o The mass of adipose tissue (Mflab) was estimated by:


87

Mflab = Vflab * ρfat = 0.0755*900 = 67.95 kg

o The upper limit of mass was calculated as:

(Mupper.fat) = M + Mflab = 64.6+67.95 = 132.55 kg

o The upper limit of BMI (BMIupper) was therefore:

BMIupper = Mupper.fat / H2

= 132.55 / 1.7042 = 45.7 kg/m

2

If the SuperFlabTM

were modeled instead as muscle:

o The density of muscle tissue (ρmuscle) was approximated as 1.06 g/cm3 = 1060

kg/m3

o The mass of muscle tissue (Mflab) was estimated by:

Mflab = Vflab * ρmuscle = 0.0755*900 = 80 kg

o The upper limit of mass was:

(Mupper.fat) = M + Mflab = 64.6+80 = 144.6 kg

o The upper limit of BMI (BMIupper) was therefore:

o BMIupper = Mupper.fat / H2

= 144.6 / 1.7042 = 49.8 kg/m

2

Summary:

The BMI for the Average body habitus phantom configuration was ~22.3 kg/m2. This is in

agreement with typical clinical classification of average body type (BMI = 18.50-24.99 kg/m2).

32

The BMI for the obese body habitus phantom configuration was ~32.8 – 45.7 kg/m2 when

SuperFlabTM

was modeled as adipose tissue and 34.6 - 49.8 kg/m2

when SuperFlabTM

was

modeled as muscle. A patient with this minimum BMI would be considered obese (obese class I:


88

BMI = 30.00 - 34.99 kg/m2) while a patient with the maximum BMI would be considered

morbidly obese [obese class III: BMI ≥ 40.00 (“morbidly obese” BMI = 40.00 – 49.99, above

which is termed “super obese.”)].32

Overall, the phantom and added SuperFlab configurations

were consistent with average and obese body habitus, respectively.


89

REFERENCES

1 Canadian Cancer Statistics 2007. Canadian Cancer Institute - National Cancer Institute of Canada, 2007.

2 N. Shkumat, High-Performance Dual-Energy Imaging with a Flat-Panel Detector. Department of

Medical Biophysics, University of Toronto. 2008.

3 What is Lung Cancer? Canadian Cancer Society. 2007. Accessed on August 1

st, 2008.

http://www.cancer.ca/ccs/internet/standard/0,3182,3172_10175_86915_langId-en,00.html

4 H. H. Hansen, Textbook of Lung Cancer, Martin Dunitz Ltd., London, 2000.

5 Causes of Lung Cancer. Canadian Cancer Institute. Accessed on April 17, 2009.

http://www.cancer.ca/Canada-

wide/About%20cancer/Types%20of%20cancer/Causes%20of%20lung%20cancer.aspx?sc_lang=en

6 The Stages of Lung Cancer. Cancer Research UK, 2002. Accessed on August 1

st, 2008.

http://www.cancerhelp.org.uk/help/default.asp?page=6706

7 R.J. Van Klaveren et al., “Lung cancer screening by low-dose spiral computed tomorgraphy,” Eur.Respir.J. 18[5]:

857-866 (2001).

8 S.J. Swensen et al., “Lung Nodules: Dual-Kolovolt Peak Analysis with CT – Multicenter Study,” Radiology

214[1]: 81-85 (2000).

9 Sources and Effects of Ionizing Radiation: United Nations Scientific Committee on the Effects of Atomic

Radiation: UNSCEAR 2000 report to the General Assembly. New York: United Nations, 2000.

10 What is the ALARA Principle? Health Physics Society. Accessed on April 8, 2009.

http://www.hps.org/publicinformation/ate/q435.html 11

J.T. Bushberg et al., The Essential Physics of Medical Imaging, Lippincott Williams & Wilkins, Philadephia, PA,

2002.

12 D.J. Brenner and W. Huda, “Effective Dose: A useful concept in diagnostic radiology?” Radiation Protection

Dosimetry 128 (4): 503-508 (2008).

13 R.L. Dixon et al., “Restructuring CT dosimetry-A realistic strategy for the future requiem for the pencil chamber,”

Med. Phys. 33 (10): 3973-3976 (2007).

14 The Future of CT Dosimetry (AAPM Task Group # 111). The American Association of Physicists in Medicine,

2008. Accessed on August 1st, 2008. https://www.aapm.org/org/structure/default.asp?committee_code=TG111

15 Adult Obesity in Canada – Measured Height and Weight. Statistics Canada, 2005. Accessed on August 1

st, 2008.

http://www.statcan.ca/english/research/82-620-MIE/2005001/articles/adults/aobesity.htm#1

16 R.N. Uppot et al., “Effect of obesity on image quality; fifteen-year longitudinal study for evaluation of dictated

radiology reports,” Radiology 240: 435-439 (2006).

17 J.S. Lin et al., “Reduction of false positives in lung nodule detection using a two-level neural classification.”

Institute of Electrical and Electronics Engineers, 1996. 0278-0062 (1996).


90

18 M. Remy-Jardin et al., “Pulmonary nodules: Detection with thick-section spiral CT versus conventional CT,”

Radiology 187: 513–520 (1993).

19 E J Hall and D J Brenner, “Cancer risks from diagnostic radiology,” The British Journal of Radiology 81: 362-378

(2008).

20 NCRP. Evaluation of the linear nonthreshold dose-response model for ionizing radiation. Bethesda, MD: National

Council on Radiation Protection and Measurements; 2001: Report No 136.

21 M. Tubiana, “Dose-effect relationship and estimation of the carcinogenic effects of low doses of ionizing

radiation: The joint report of the Académie des Sciences (Paris) and of the Académie Nationale de Medecine.” Int J

Radiat Oncol Biol Phys 63: 317-319 (2005).

22 E.L. Kaufman et al., “Effect of breast cancer radiotherapy and cigarette smoking on risk of second primary lung

cancer.” J Clin Oncol 26: 392-398 (2008).

23 D.J. Brenner, “Radiation risks potentially associated with low-dose CT screening of adult smokers for lung

cancer,” Radiology 231: 440-445 (2004).

24 D.J. Brenner, “Computed Tomography-An increasing source of radiation exposure,” N Engl J Med 357: 2277-

2284 (2007).

25 N. Piyaviset et al., “Small incidental pulmonary nodules: How useful is short term interval CT?” J. Thorac.

Imaging 20: 5–9 (2005).

26 N.S. Paul et al., “Investigating the low-dose limits of multidetector CT in lung nodule surveillance,” Med. Phys.

34: 3587-3595 (2007).

27 D.F. Yankelevitz et al., “Small pulmonary nodules: Evaluation with repeat CT—Preliminary experience,”

Radiology 212: 561–566 (1999).

28 M.P. Revel et al., “Pulmonary nodules: Preliminary experience with three-dimensional evaluation,” Radiology

231: 459–466 (2004).

29 J. Valentin (International Commission on Radiation Protection), “Managing patient dose in multi-detector

computed tomography (MDCT),” Ann ICRP 37: 1-79 (2007).

30 C.B. Chiarot et al., “An innovative phantom for quantitative and qualitative investigation of advanced x-ray

imaging technologies,” Phys. Med. Biol. 50: 297 (2005).

31 A.E. Burgess et al., “Mass discrimination in mammography: Experiments using hybrid images,” Acad. Radiol. 10:

1247-1256 (2003).

32 BMI Classification. World Health Organization, 2009. . Accessed on April 23, 2009.

http://www.who.int/bmi/index.jsp?introPage=intro_3.html

33

Gescheider, G. A. Psychophysics: The Fundamentals – Third Edition. Lawrence Erlbaum Associates, Mahwah,

NJ, 1997.

34 D. J. Tward et al., “Soft-tissue detectability in cone-beam CT: Evaluation by 2AFC tests in relation to physical

performance metrics,” Med. Phys. 34(11): 4459 – 4471 (2007)

35 R.L. Dixon and A.C. Ballard, “Experimental validation of a versatile system of CT dosimetry using a

conventional ion chamber: Beyond CTDI100,” Med. Phys. 34 (8): 3399-3413 (2007).


91

36 J.M. Boone. The Trouble with CTDI100. Med. Phys. 34 (4): 1364-1371 (2007).

37 S. Mori et al., “Enlarged longitudinal dose profiles in cone-beam CT and the need for modified dosimetry,” Med.

Phys. 32(4): 1061-1069 (2005).

38 Introduction to the MOSFET Dosimeter; Technical Note 4. Best Medical Canada, Thomson-Nielsen. Accessed on

August 1st, 2008. www.thomson-elec.com/global/pdf/technotes/te_4.pdf

39 M.K. Islam et al., “Patient dose from kilovoltage cone beam computed tomography imaging in radiation therapy,”

Med. Phys. 33 (6): 1573-1582 (2006).

40 H. Fujita et al, "A simple method for determining the modulation transfer function in digital radiography," IEEE

Trans. Medical Imaging 11: 34-39 (1992).

41 H.H. Barrett et al., “Statistical limintations in transaxial tomography,” Comput. Biol. Med. 6, 307-323 (1976).

42 W. Huda et al. “Technique factors and image quality as functions of patient weight in abdominal CT,” Radiology

217: 430-435 (2000).

43 D.J. Tward and J.H. Siewerdsen, “Cascaded systems analysis of the 3D noise transfer characteristics of flat-panel

cone-beam CT,” Med. Phys. 35: 5510-5529 (2008).

the low-dose limits of lung nodule detectability in

Documents