radiation detectors in medical and biological applications1 · radiation detectors in medical and...

15
* E-mail: besch@vxdesy.desy.de. 1 Invited talk at the Vienna Wire Chamber Conference WCC98, Vienna, February 22-26, 1998. Nuclear Instruments and Methods in Physics Research A 419 (1998) 202 216 Radiation detectors in medical and biological applications1 H.J. Besch* Universita ( t-Gesamthochschule Siegen, Fachbereich Physik, Emmy-Noether-Campus, 57068 Siegen, Germany Abstract Basic definitions of imaging theory are given and their importance for detector development is demonstrated. Developments in two-dimensional and scanning radiographic systems are discussed. The substantial progress in non-invasive coronary angiography with synchrotron radiation is presented together with a comparison of the performance of a gaseous and a silicon detector. Synchrotron radiation is also applied in mammography. New modalities making use of coherence are discussed. Non-radiative methods competing with X-ray systems are also presented. Detectors for nuclear medicine and biological structure research are briefly mentioned. ( 1998 Elsevier Science B.V. All rights reserved. Keywords: Radiation detectors; Imaging theory; Radiography; Synchrotron radiation; Coronary angiography; Mam- mography 1. Introduction Since most of the medical and biological applica- tions of radiation detectors are concerned with imaging some basic properties of imaging systems and the terminology of the imaging community will first be discussed. The main purpose of this part is to convince the reader that a comparison of the performance of different detectors is not straight- forward and that for such a comparison appropri- ate methods have to be used. Since there is now general agreement in medical imaging that the radiation dose has to be kept at the lowest possible value compatible with the diag- nostic purpose, a medical image is always noise limited, at least in some region of interest. In the language of the imaging community Poisson statis- tics are included in the noise. 2. Characterisation of a noise limited imaging system The noise properties of an imaging system can be determined from the noise autocorrelation function C(x, y). This function is given by the correlation of the fluctuations *S of a constant (in the mean) signal as a function of the distance: C(x, y)" 1 a P a P *S(x@,y@) *S(x@#x, y@#y)dx@ dy@. (1) The variables x, y, x@, y@ are (Cartesian) coordinates in the image and a is its area. The Fourier transform 0168-9002/98/$19.00 ( 1998 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 8 - 9 0 0 2 ( 9 8 ) 0 0 7 9 4 - 3

Upload: others

Post on 24-Sep-2019

13 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Radiation detectors in medical and biological applications1 · Radiation detectors in medical and biological applications1 H.J. Besch* Universita(t-Gesamthochschule Siegen, Fachbereich

*E-mail: [email protected] Invited talk at the Vienna Wire Chamber Conference

WCC98, Vienna, February 22-26, 1998.

Nuclear Instruments and Methods in Physics Research A 419 (1998) 202—216

Radiation detectors in medical and biological applications1

H.J. Besch*

Universita( t-Gesamthochschule Siegen, Fachbereich Physik, Emmy-Noether-Campus, 57068 Siegen, Germany

Abstract

Basic definitions of imaging theory are given and their importance for detector development is demonstrated.Developments in two-dimensional and scanning radiographic systems are discussed. The substantial progress innon-invasive coronary angiography with synchrotron radiation is presented together with a comparison of theperformance of a gaseous and a silicon detector. Synchrotron radiation is also applied in mammography. New modalitiesmaking use of coherence are discussed. Non-radiative methods competing with X-ray systems are also presented.Detectors for nuclear medicine and biological structure research are briefly mentioned. ( 1998 Elsevier Science B.V.All rights reserved.

Keywords: Radiation detectors; Imaging theory; Radiography; Synchrotron radiation; Coronary angiography; Mam-mography

1. Introduction

Since most of the medical and biological applica-tions of radiation detectors are concerned withimaging some basic properties of imaging systemsand the terminology of the imaging community willfirst be discussed. The main purpose of this part isto convince the reader that a comparison of theperformance of different detectors is not straight-forward and that for such a comparison appropri-ate methods have to be used.

Since there is now general agreement in medicalimaging that the radiation dose has to be kept atthe lowest possible value compatible with the diag-nostic purpose, a medical image is always noise

limited, at least in some region of interest. In thelanguage of the imaging community Poisson statis-tics are included in the noise.

2. Characterisation of a noise limited imagingsystem

The noise properties of an imaging system can bedetermined from the noise autocorrelation functionC(x, y). This function is given by the correlation ofthe fluctuations *S of a constant (in the mean)signal as a function of the distance:

C(x, y)"1

aPaP*S(x@,y@)*S(x@#x, y@#y) dx@dy@. (1)

The variables x, y, x@, y@ are (Cartesian) coordinatesin the image and a is its area. The Fourier transform

0168-9002/98/$19.00 ( 1998 Elsevier Science B.V. All rights reserved.PII: S 0 1 6 8 - 9 0 0 2 ( 9 8 ) 0 0 7 9 4 - 3

Page 2: Radiation detectors in medical and biological applications1 · Radiation detectors in medical and biological applications1 H.J. Besch* Universita(t-Gesamthochschule Siegen, Fachbereich

¼( f, g) of the noise autocorrelation function isWieners noise power spectrum

¼( f, g)"P~=P

`=C(x, y) e~2p* > ( f >x`g >y) dxdy. (2)

Here f and g are the spatial frequencies with dimen-sions of inverse length related to x and y, respec-tively. Spatial frequencies are often quoted in line-pairs/mm. In the frequency domain it is possible tomake use of the convolution theorem to simplifythe analysis of composite systems. Consequently,instead of the point spread function PSF(x, y) (i.e.the detector resolution curve) its Fourier transform,the optical transfer function O¹F( f, g) given by

O¹F( f, g)"P~=P

`=PSF(x, y) e~2p*( f >x`g >y)dxdy

(3)

or its modulus, the modulation transfer functionM¹F( f, g) given by

M¹F( f, g)"DO¹F( f, g)D (4)

is used. Note that M¹F(0,0)"1 and also thatfor a detector of infinitely good resolutionM¹F( f, g)"1.

The total M¹F of a detector system is the prod-uct of the M¹Fs of its components. The use of the(spatial) frequency-domain permits in some casesan improvement in the spatial resolution by decon-volution, normally, at the expense of increasedimage noise.

The noise properties of an imaging detectionsystem, its quantum efficiency and its spatial res-olution are usually combined in the detective quan-tum efficiency DQE( f, g)

DQE( f, g)"ASNR

065( f, g)

SNR*/( f, g)B

2

"AS065

( f, g)p*/( f, g)

S*/( f, g)p

065( f, g)B

2, (5)

where SNR, S and p are the signal-to-noise ratio,the signal and the noise at the input and the outputof the detector system, respectively.

Obviously, for a detector for which the noise isgiven by Poisson statistics only and which has

infinitely good spatial resolution, the DQE is equalto ordinary quantum efficiency QE. In general, theDQE may be expressed by the functions alreadydefined, giving

DQE( f, g)"QEDM¹F( f, g)D2

¼( f, g). (6)

The ‘zero-frequency DQE’, DQE(0,0), is used todescribe the frequency-independent properties ofan imaging system. It is again equal to ordinary QEfor a Poissonian sampling system of finite pixel size.

The following example gives some insight intothe meaning of the zero-frequency DQE: for a one-dimensional integrating X-ray detector with a finite(intrinsic) single-event energy resolution, a resultingnormalised spectrum n(E) for the energy depositedper event and additional integrator noise ¼

!$$ex-

pressed in equivalent photons, the zero-frequencyDQE is given by

DQE(0)"QE

1#QE (:=

0n(E)E2dE!E2c )

E2c#

¼!$$

QE N

,

(7)

where N is the number of photons accumulatedand Ec the photon energy. The central term in thedenominator describes the loss of signal-to-noiseratio due to, e.g. escape losses from fluorescence ora Compton contribution. This loss can never berecovered by improving statistics. The additivenoise term, on the other hand, disappears withincreasing statistics. Therefore, DQE(0) depends onthe signal size and approaches QE for large signalsand a narrow spectrum n(E). The resulting curvesare shown in Fig. 1 for different noise levels. Also,shown are DQE curves for a counting system,where DQE approaches zero for a rate slightlyabove the inverse deadtime.

If the pixel noise in a detector is measured todetermine QE, naively using Poisson statistics, eas-ily an apparent QE'1 can be obtained. This ismost likely the case for a low-noise sampling sys-tem with some ‘cross talk’ in the detection system(wide presampling PSF), because a pixel would ‘see’part of the statistics of its neighbours. Therefore,even for this simple task the methods describedabove have to be used. For further information the

H.J. Besch/Nucl. Instr. and Meth. in Phys. Res. A 419 (1998) 202—216 203

I. INVITED TALKS

Page 3: Radiation detectors in medical and biological applications1 · Radiation detectors in medical and biological applications1 H.J. Besch* Universita(t-Gesamthochschule Siegen, Fachbereich

Fig. 1. Zero-frequency DQE as a function of the number ofphotons per pixel at the input of the detector for QE " 0.7.Curves (a—d): integrating systems with a noise of 10, 102, 103 and104 equivalent photons, curves (e—i): counting systems witha ratio of exposure time to deadtime of 103, 104, 105, 106 and107.

reader is referred to Refs. [1—4]. In these referencesthe effects of spatial sampling are also discussed.

The noise properties of an imaging system, obvi-ously, determine the radiation dose which a patienthas to accept for a given signal-to-noise ratio. Thefraction of the radiation dose a patient has totake due to imperfections of the imaging system is1!DQE. Therefore, for medical applications DQEshould be close to one.

For a diagnostic interpretation a signal-to-noiseratio of three to five is the minimum required. Fora sampling system with matched pixel size the skindose D

4,*/increases with the inverse fourth power

(1/w4) for a cubic object of size w, and also increaseswith the square of the signal to noise ratio SNR*Srequired [5]. D

4,*/is given by

D4,*/

"

2ekN )¸DQE (0)

SNR2*SkN 2w4C2k

EcAkNoB . (8)

Here ¸ is the length of the tissue traversed by thebeam, kN is the absorption coefficient of the tissueand o is its density. Ck"*k/kN is the object contrast,i.e. the relative change of the absorption coefficientin the object with respect to its neighbourhood.

Any dose quoted without proper considerationof the spatial and intensity resolution is thereforemeaningless.

As a consequence, there is only one reliable methodto compare the performance of different detectors,namely to compare DQE as a function of the spatialfrequencies at the most critical intensity.

3. Digital radiography

The by far most common application of medicalimaging is radiography, and to a lesser extent,fluoroscopy. The first is still largely dominated byX-ray film and screen film systems as detectors, butindustry has made a strong effort to replace themby digital systems with all the possibilities of pro-cessing and transferring digital data. These systems,if well made, have the additional advantage ofa much larger dynamic range compared to theseven bit of film. Therefore, by playing with the con-trast function, the medical doctor can see detailswhich otherwise would require several exposures,even for digitised film data. Present developmentsgo into two directions: two-dimensional generalpurpose systems and scanning systems.

3.1. Two-dimensional systems

The quantum efficiency of a 300lm Si-wafer forX-rays of medical energies is at best a few percent.Therefore, since QE is an upper bound for DQE, theidea of using standard Si-wafers as 2D-detectorscan immediately be abandoned.

With the advent of the amorphous silicon (a-Si)thin-film technique, detectors based on a converterand an a-Si active matrix became an attractivepossibility, because they can be built at a reason-able price and in sizes and weights compatible withcommon X-ray equipment. Ideally, they would fitinto a standard X-ray cassette. Such systems are atthe industrial production threshold and will soonbe produced in large quantities.

There are two types of such systems using in-direct or direct conversion of the X-ray photons.The indirect systems use a scintillator as converterand an active matrix as a detector for the visiblephotons emitted by the scintillator. The direct sys-tems use a heavy semi-conductor as a converterwith direct charge transfer to the active matrix. Thecharges stored in the capacitance of the pixels of the

204 H.J. Besch/Nucl. Instr. and Meth. in Phys. Res. A 419 (1998) 202—216

Page 4: Radiation detectors in medical and biological applications1 · Radiation detectors in medical and biological applications1 H.J. Besch* Universita(t-Gesamthochschule Siegen, Fachbereich

Fig. 2. Active matrix and CsI scintillator.

Table 1Properties of a CsI-a-Si detector system (Siemens/Thomson)

Size 20]23 cm2 or 40]46 cm2

(four may be combined)Thickness 15 cm including electronicsReadout speed 12.5 frames/sADC 14 bitDQE(0) 0.65 for 70 keV anode voltage and 0.7 mm

Cu filterM¹F 0.35 at 2.5 line-pairs/mm

Fig. 3. Schematic view of scanning X-ray systems.

active matrix are switched by thin-film transistors(TFTs) or thin-film diodes (TFDs) line by line to theoutput bus lines which follow the columns. Thecharges are transferred to an ADC via multiplexers,digitised and stored.

One such system (see Fig. 2), using a CsI-scintil-lator of 450lm thickness, grown in needles of6—10lm in diameter, has been developed by Sie-mens and Thomson [6]. Its properties are sum-marised in Table 1. The readout speed of12.5 frames/s makes the system also suitable forfluoroscopy.

A direct conversion system with a 2mm amorph-ous Se converter is, e.g. under development in Sas-katoon and in Toronto [7,8]. Images taken withthis system can be found in Ref. [8].

Considering the present status of industrial de-velopment, it does not seem promising to starta new project in this field.

3.2. Scanning systems

Scanning systems (see Fig. 3) have two advant-ages compared to 2D-systems: if properly designed,they have reduced scatter background and they useonly a small number of channels. But they needspecial scanning X-ray equipment and do not per-mit fluoroscopy.

The use of counting or integrating detectors isnot a question of principle, as can be seen fromFig. 1. In integrating systems the spatial resolution

is (de facto) always limited by the pixel size, while incounting systems an interpolation is possible. Butintegrating systems are always as good as countingsystems in intensity resolution, if the integratornoise ¼

!$$(see Eq. (7)) is smaller than the Poisson

noise. The improvement in cheap low-noise elec-tronics has continuously decreased the limit for theuse of integrating systems. In addition, integratinggaseous detector systems are robust, even with finecollecting structures.

The Novosibirsk commercial scanning systemhas been presented at this conference several times[9]. It uses a tapered pressurised gaseous detectororiginally working in the counting mode. With thetransition to an integrating mode it follows theabove mentioned line of development.

New developments use edge-on tapered Si-stripdetectors (see Fig. 4 and Fig. 11). With this tech-nique QE and DQE can be kept close to one due tothe long active length. In addition, the well-knownSi-technique can be used. These detectors workmainly in the counting mode.

A system of this kind has been developed inTrieste for application in digital mammography

H.J. Besch/Nucl. Instr. and Meth. in Phys. Res. A 419 (1998) 202—216 205

I. INVITED TALKS

Page 5: Radiation detectors in medical and biological applications1 · Radiation detectors in medical and biological applications1 H.J. Besch* Universita(t-Gesamthochschule Siegen, Fachbereich

Fig. 4. Principle of Si edge-on detectors.

Table 2Properties of the LBL-Berkeley Si edge-on strip detector

Thickness 300lmDepth 45mmStrip pitch 100lmCount rate '1 MHz/segmentSegment in depth 2, 6, 12, 24mm

tapered geometry Fig. 5. M¹F of the LBL-Berkeley Si-strip detector, solid line:edge scan, circles: phantom measurements, dash—dotted line:ideal curve for 100lm pixel size.

Fig. 6. Mouse skull.

with synchrotron radiation [10,41,11] and is de-scribed in Section 4.3.

At LBL-Berkeley [12,42] a similar system hasbeen developed, specially adopted to radiographywith X-ray tubes. Its properties are shown inTable 2. The segmentation in depth permits a stat-istical energy measurement and an increase in thetotal rate. Fig. 5 shows M¹F curves normal to thescan direction for this detector: a theoretical M¹Fderived from the pixel size and the results of anedge scan and a phantom measurement. Due to thevery short track of the photoelectrons in Si (asopposed to gas), the M¹F is mainly determined bythe pixel size. Using a very narrow collimated beamthe M¹F in the direction of the scan is similar toFig. 5 in spite of the wafer thickness of 300lm.Fig. 6 shows the image of a mouse skull taken withthis detector demonstrating its excellent resolution.

At Strasbourg work on a similar project is goingon using k-strips of 50mm depth and VLSI elec-tronics. The main emphasis is on continuousreadout, spectroscopic acquisition [13].

A different and promising approach is the use ofheavy semiconductors (e.g. GaAs or CdTe) as linedetectors. As an example, the fast counting GaAs

206 H.J. Besch/Nucl. Instr. and Meth. in Phys. Res. A 419 (1998) 202—216

Page 6: Radiation detectors in medical and biological applications1 · Radiation detectors in medical and biological applications1 H.J. Besch* Universita(t-Gesamthochschule Siegen, Fachbereich

Fig. 7. Side view of coronary angiography systems with synchrotron radiation.

system developed by a Cagliari/Catania group andpresented at this conference [14] is mentioned.

4. Synchrotron radiation in coronary angiographyand mammography

Synchrotron radiation in medical imaging isregularly used in intravenous coronary angiogra-phy and in several other projects in the develop-ment phase. All these projects make use of at leastone characteristic property of synchrotron radi-ation, e.g. high power per energy bandwidth, smallbeam divergence, good stability and even coher-ence. Monochromators are used to provide narrowband beams.

4.1. Coronary angiography

At DESY the NIKOS system for coronary an-giography [15,16] has been handed over to themedical doctors for evaluation [17]. Now, three tofour patients are examined per day, with a total of276 patients examined at the time of the conference.The system works with the now well-knownmethod of K-edge subtraction. Iodine contrastagent is injected into a peripheral vein.Fig. 7 shows a side view of the system with thepatient being moved through the beam. As detector

a two line high-pressure KrCO2

ionisation cham-ber is used with a pixel size of 400 lm and an activewidth of 150mm.

To avoid motion blurring, images have to betaken in less than 10ms for a 2D image and in lessthan 250 ms for a line scan system resulting in anexposure time of (1ms per line. The number ofphotons per pixel accumulated in the detector in1ms amounts to a few 102 behind a partially Iodinefilled ventricle and to about 106 behind the lungcorresponding to about 109 Hz. As can be seenfrom Fig. 1 or Eq. (7), this requires in case of integ-rating systems a noise as low as a few photons tokeep DQE close to QE at the lower end of therange.

Consequently, the system has to have a dynamicrange (maximum signal/integrator noise) of18—19bit. A counting system with a maximum rateclose to 109 Hz would need a very large numberof segments in the beam direction, making such asystem technically difficult and very expensive.Therefore, all systems developed for coronary an-giography now use integrating electronics withmore than 16 bit dynamic range and an integratornoise of a few photons.

The right coronary artery (RCA) and the leftanterior descending branch (LAD) are relativelyeasy to image, while the left circumflex branch(LCX) is always partially obscured by the left

H.J. Besch/Nucl. Instr. and Meth. in Phys. Res. A 419 (1998) 202—216 207

I. INVITED TALKS

Page 7: Radiation detectors in medical and biological applications1 · Radiation detectors in medical and biological applications1 H.J. Besch* Universita(t-Gesamthochschule Siegen, Fachbereich

Fig. 8. Energy image above (E2) and below (E1) the K-edge and subtraction image for the NIKOS coronary angiography system.

ventricle. Fortunately, the LCX has a ‘large’ dia-meter of 3—4mm in this region. Therefore it can bevisualised making use of structure dependent imageprocessing. The benefit of image processing in noiselimited images should not be overestimated, be-cause any improvement in intensity resolution is atthe expense of spatial resolution and vice versa.Clearly, the subtraction of large (average) struc-tures can help for the interpretation, as does theK-edge subtraction method in spite of a factor ofJ2 lost in the intensity resolution for constant skindose.

Fig. 8 shows two simultaneously taken images,one below (E1) and one above (E2) the K-edge, and

the resulting subtraction image, demonstratinghow the right coronary artery emerges from thebackground by the subtraction method. The stent(support mesh) in the proximal part of the RCA ismade visible by deliberately unbalancing theweighting factors for the subtraction.

The skin dose is about 40mSv per subtractionimage. The total skin dose for three scans in twopositions and a few low rate adjustment scansamounts to about 300mSv, somewhat below thedose for a standard invasive coronary angiography.

A detailed discussion about performance andpresent status of the NIKOS system developed atDESY and in Siegen can be found in Ref. [18].

208 H.J. Besch/Nucl. Instr. and Meth. in Phys. Res. A 419 (1998) 202—216

Page 8: Radiation detectors in medical and biological applications1 · Radiation detectors in medical and biological applications1 H.J. Besch* Universita(t-Gesamthochschule Siegen, Fachbereich

Fig. 9. M¹F (a) and Wieners noise power spectrum (b) for theionisation chamber.

The medical evaluation should be finished by theend of this year. Then it will be decided, if thistechnique will be implemented as standard medicalroutine.

To give an idea about the time scale and theeffort needed to arrive at this level, it is pointed outthat a group of about 10 people (including medicaldoctors) for 12 years was required. More than halfof them have worked full time on the project.

At the NSLS in Brookhaven/USA a similar sys-tem has been developed and about 30 patients havebeen examined with this system [19]. It uses, asdetector, a Si(Li) bar of 5 mm depletion depth,carrying two lines of pixels. Five mm of Si is equiva-lent to &1 absorption length at the Iodine K-edgeof 33.17 keV. A comparison of the NIKOS and theNSLS system is presented in Section 4.2.

At the ESRF a cooled Ge two line pixel detectorwith 2mm depth and 350lm pixel size has beendeveloped and is presently tested [20]. It uses 16 bitintegrating electronics with additional switchablegain. It has not yet been used for patient studies.

Synchrotron radiation has also been tried atNSLS Brookhaven for bronchiography usingXenon as a contrast agent to measure lung ventila-tion. The method used is similar to coronary an-giography but works at the Xenon K-edge. Thecontrast achieved is only moderate due to the lowXe density. Because xenon is a narcotic, it is diffi-cult to develop a safe procedure.

4.2. Comparison of a Si(Li) and a gaseous detectorfor coronary angiography

A careful comparison of the two detection sys-tems (NIKOS and the NSLS detector) has beenperformed [3] using the methods described inSection 2.

Fig. 9a and b show the M¹F and Wieners noisepower spectrum ¼ (denoted by NPS in the figure)for the ionisation chamber. Since the square of theM¹F drops faster with frequency than the noise,DQE decreases with increasing frequency. This be-haviour is shown in Fig. 10a and b for the twodetectors. As can be seen from the figure, DQE isbetter for the ionisation chamber than for the Si(Li)detector at signals of more than 20 000 photons,typical for a free standing artery. At smaller signals

the Si(Li) detector had a better DQE, but its dy-namic range was insufficient. As a consequence ofthis analysis both detectors were equipped withnew electronics, improving mainly the noise for theionisation chamber [18] and mainly the dynamicrange and minimum exposure time of the Si(Li)detector. A dynamic range of nearly 19 bit anda noise of three photons has been achieved for theionisation chamber. The DQE of the Si(Li) detectorwill always stay below 0.5 due to its insufficientX-ray absorption. It remains to remark that aftersome initial struggle the gaseous ionisation cham-ber turned out to be more reliable than the Si(Li)detector.

H.J. Besch/Nucl. Instr. and Meth. in Phys. Res. A 419 (1998) 202—216 209

I. INVITED TALKS

Page 9: Radiation detectors in medical and biological applications1 · Radiation detectors in medical and biological applications1 H.J. Besch* Universita(t-Gesamthochschule Siegen, Fachbereich

Fig. 10. DQE contours for the ionisation chamber (a) and theSi(Li) detector (b).

4.3. Mammography

From Eq. (8) it can be worked out that fora 30 lm in diameter micro-calcification, typical for

a very early stage of breast cancer, and a SNR of(only) three, the resulting skin dose would be50mSv and the mean glandular dose about 20mSv,even for an ideal detector with DQE"1. This is anunacceptable high dose for a screening procedure.Therefore, especially in mammography, methodswhich improve the object contrast are of greatimportance. Two possible candidates will be de-scribed in Section 5.

Synchrotron radiation is used for mammo-graphic studies mainly because a reduction of theradiation dose or a corresponding improvement inimage quality can be expected. This improvement iscaused by an optimised narrow band beam andscatter background reduction due to the line scanmethod used. Most advanced in this field is a col-laboration working at the medical beam lineSYRMEP at ELETTRA in Trieste [10,41].

The detector used is an edge-on Si strip detector(see Fig. 11) coupled to VLSI preamplifiers, dis-criminators and scalers developed in Strasbourg[11]. The mechanical layout is such that detectorsmay be extended sideways and stacked. The basicproperties of the detector are listed in Table 3.

Fig. 12 shows an image of an excised breast takenwith a narrow band beam at 18 keV with 100 lmstep size. The mean glandular dose was 0.2mSvonly, matched to the step size and to the relativelylarge vascular calcification visible in the frame.

5. New modalities

It is also possible to make use of the relativelygood coherence of mono-energetic synchrotronradiation beams to increase the object contrast.These methods are phase contrast imaging [21]and diffraction enhanced imaging (DEI) [22]. Bothmethods make use of the phase shift u of an electro-magnetic wave travelling in z direction through anobject with electron density o

%(see Fig. 13). The

phase shift is given in lowest order by

u(x, y)"!r%jPo%

(x, y, z) dz, (9)

where r%

is the classical electron radius and j thewavelength. The direction of the beam is denoted

210 H.J. Besch/Nucl. Instr. and Meth. in Phys. Res. A 419 (1998) 202—216

Page 10: Radiation detectors in medical and biological applications1 · Radiation detectors in medical and biological applications1 H.J. Besch* Universita(t-Gesamthochschule Siegen, Fachbereich

Fig. 11. Edge-on Si detector for mammography including electronics.

Table 3Edge-on Si strip detector at the medical beam line at ELETTRA

Depth 10mmDead region 250lmPitch 200lmWidth 51mmThickness 300lmDQE(0) at 20 keV &0.75Max. rate 105Hz/stripNoise (electrons) 100#7 )C

-0!$(pF)

Number of channels 32/chip

by z. If the phase shift depends on one of thetransverse variables, e.g. x, the wave is refracted byan angle *a given by

*a+j2n

Lu(x)

Lx. (10)

In phase contrast imaging the detector is simplyplaced at a substantial distance (up to a few meters)downstream of the object, instead of directly behindit, as it is in most cases optimal for an absorptionimage. Phase contrast imaging is most effective forsmall objects.

From Eq. (9) a very faint image at first could beexpected, because the phase shift is proportional too%and therefore proportional to the nuclear charge

Z, similar to the Compton cross section. Comptonimages are known to have a low contrast for thisreason. The photo cross section, the main contribu-tion to absorption images, on the other hand,is proportional to Z5 apparently providing amuch larger object contrast. But the expectationof a faint image is not justified for three reasons:(i) photo and Compton cross sections are propor-tional to r2

%instead of the much larger factor

r%j appearing in Eq. (9), (ii) the phase shift at

a given energy is only weakly coupled to absorp-tion (by a dispersion relation) and (iii) theinterference with the primary wave gives a strongenhancement.

Fig. 14 shows an absorption (top) and a phasecontrast (bottom) image of a bee taken at the SYR-MEP beam line at ELETTRA [23]. The detectorwas a high resolution X-ray film. The increase incontrast is impressive.

Techniques making use of Eq. (10), like DEI, willgive a strong edge enhancement due to the deriva-tive in the equation. This can be very useful for theinterpretation of images.

H.J. Besch/Nucl. Instr. and Meth. in Phys. Res. A 419 (1998) 202—216 211

I. INVITED TALKS

Page 11: Radiation detectors in medical and biological applications1 · Radiation detectors in medical and biological applications1 H.J. Besch* Universita(t-Gesamthochschule Siegen, Fachbereich

Fig. 12. Image of excised breast.

Fig. 13. Principle of phase contrast and diffraction enhancedimaging.

Fig. 14. Absorption image (top) and phase contrast image (bot-tom) of a bee.

In DEI a diffracting analyser crystal is placedbetween object and detector and two images aretaken, one on either side of the rocking curve [22]of the analyser crystal at the point of the steepestslope. For a good analyser the width of the rockingcurve is in the micro-radian range. Thus, the co-herently or incoherently scattered part of the beamdoes not appear on the detector, and the image isessentially determined by the refracted part follow-ing Eq. (10). For the formation of the image thescattered parts are as important as the absorbedpart of the beam, thus substantially increasing theobject contrast.

By taking the sum and the difference of the twoimages weighted with the slope of the rockingcurve, a refraction angle image and an extinction(or apparent absorption) image is obtained. Theseimages are shown in Fig. 15 for a mammographicphantom together with an ordinary absorption im-age [22]. The contrast in the extinction image is

more than 25 times better compared to the absorp-tion image. The very clear appearance of the fibrilsin the upper right corner of the angle image demon-strates that additional information is available

212 H.J. Besch/Nucl. Instr. and Meth. in Phys. Res. A 419 (1998) 202—216

Page 12: Radiation detectors in medical and biological applications1 · Radiation detectors in medical and biological applications1 H.J. Besch* Universita(t-Gesamthochschule Siegen, Fachbereich

Fig. 15. Absorption image (a), refraction angle image (b) and extinction image (c) of a mammographic phantom.

which cannot be extracted from the absorption orthe extinction image alone.

The increased contrast in angle and extinctionimages makes the method of DEI a very promisingcandidate for mammography.

The usefulness of phase contrast imaging formammography still has to be proven.

6. Non-radiative methods

In some fields, methods which do not use ionis-ing radiation compete with X-ray imaging. Themost common of these non-radiative methods areultrasound techniques and magnetic resonanceimaging (MRI). Both methods have a reputation ofbeing non-invasive, but do not allow the imaging ofsmall, rapidly moving structures like coronary arte-ries. Ultrasound, using Doppler techniques, per-mits also some functional imaging, e.g. flowmeasurements. This is the case for MRI only in veryspecial cases. Two examples will now be discussed.

6.1. Intra-coronary ultrasound

The technique of intra-coronary ultrasound[24,25], contrary to the synchrotron radiationmethod described in Section 4.1 and to commonultrasound techniques, is an invasive method using

a catheter introduced through a femural artery,similar to ordinary selective coronary angiography.It is technically interesting, because it contains ina 3.5F (,1.2mm diameter) catheter tip of 12mmlength a complete 64 element phase array sonarsystem including control electronics and space fora guide wire. The system works at 20MHz, permit-ting a sub-100lm resolution. It produces high-qua-lity cross sections through the coronary arteries,because the transducer moves with the artery, thusovercoming the problem of motion blurring.

In the Doppler mode the system also permitsblood flow measurements in arteries down to about2mm in diameter.

The technique is a good example for the progressallowed by micro-electronics.

6.2. Hyper-polarised 3He magnetic resonanceimaging

Tomographic MRI using protons does not per-mit imaging of gases and even of low-density tissue,because the product of polarisation and spin den-sity is too small. X-rays give a better image of thelung, but in both cases it is not possible to study thelung ventilation. As mentioned at the end ofSection 4.1, K-edge subtraction with xenon hasbeen tried with moderate success.

H.J. Besch/Nucl. Instr. and Meth. in Phys. Res. A 419 (1998) 202—216 213

I. INVITED TALKS

Page 13: Radiation detectors in medical and biological applications1 · Radiation detectors in medical and biological applications1 H.J. Besch* Universita(t-Gesamthochschule Siegen, Fachbereich

Fig. 16. 3He MRI images of a non-smokers (left) and a smokers (right) lung.

At Mainz a new and promising method has beendeveloped in a completely different field: MRI withhyper-polarised 3He [26,27]. For ordinary MRIthe proton spin polarisation at room (or body)temperature is simply given by the Boltzmann fac-tor and amounts to about 5]10~6 at 1.5 T. He-lium-3 (spin 1

2) can be polarised to a nuclear spin

polarisation of more than 0.5 by optical pumping.The procedure is briefly described: direct opticalpumping of the ground state is not possible, be-cause the lowest transition energy (&20 eV) wouldbe in the vacuum ultraviolet region. Therefore, themeta-stable 1s2s3S

1state is used. The state is popu-

lated in a gas discharge at a pressure of about1 mbar in a superimposed magnetic field. Then theF"3

2hyperfine level of this state is fully polarised

by optical pumping with an infrared laser (1083nm)resulting in a nuclear polarisation close to one. Byexcimer formation and decay the nuclear polarisa-tion is transferred to the ground state. The heliumis, again in a homogeneous magnetic field, com-pressed into a specially selected glass cell at threebar. After this manipulation the polarisation is stillabove 0.5, and the longitudinal relaxation time isabout 100h. The hyper-polarised 3He can then be

transported to a hospital in a ‘magnetic suitcase’.Slightly modified (to match the lower 3He frequen-cies) standard MRI equipment can then be used forthe tomographic imaging. Fig. 16 shows the differ-ence between a non-smokers (left) and a smokers(right) lung ventilation. The bright areas are thosewith high 3He spin density.

The 3He comes from tritium decay and is a by-product of Hydrogen bomb production and main-tenance. Since the total world production of 3Hegas is only about 1500 m3, more than 95% of the3He have to be recovered in regular hospital use.

As opposed to the Xenon K-edge images the 3Heimages have achieved diagnostic quality.

The example of 3He MRI shows also that, ifa new project in medical imaging is started, notonly the direct competition but also the develop-ments in, at first glance, far lying fields of researchshould be considered.

7. Detectors for nuclear medicine

Many developments have been going on for de-tectors in nuclear medicine over the last decade to

214 H.J. Besch/Nucl. Instr. and Meth. in Phys. Res. A 419 (1998) 202—216

Page 14: Radiation detectors in medical and biological applications1 · Radiation detectors in medical and biological applications1 H.J. Besch* Universita(t-Gesamthochschule Siegen, Fachbereich

improve positron emission tomography (PET) sys-tems in industry and in research institutes, aimingfor faster imaging and better resolution [28,29].While for PET decent DQEs can be achieved, this isnot the case for single photon emission tomogra-phy (SPET), for which the use of collimators limitsthe DQE to levels of some percent. The only wayout here would be a Compton imaging system.There are some occasional and some more perma-nent efforts in this direction, but, in general, pro-gress in this field seems to be slow and cumbersome[30].

Because of the many nuclides available for SPETand the corresponding large number of biologicalmarkers, good SPET detectors are very useful forfunctional imaging and can be a competition ora complement to morphological imaging.

A general trend is the increased importance ofsmall animal imaging for bio-medical research[31,32]. The new developments for PET and SPETare focused around new scintillators [33,34], likeLSO, GSO, YAP and CZT, and new position sensi-tive photodetectors, like multi-anode photomulti-pliers, or, just at the beginning, hybrid photo-detectors (HPDs) [35,36]. Depth of interactiontechniques are being further developed to improvethe spatial resolution making use of either spectralcoding or decay time coding [37].

8. Detectors for protein crystallography and smallangle X-ray scattering

There is no need to discuss the field of detectorsfor applications in biocrystallography and in smallangle X-ray scattering from biological samples hereagain in depth, because a number of relatively newsummary articles have been published to which thereader is referred, e.g. [38—40].

There are presently no detector systems availablewhich make full use of the performance of thirdgeneration synchrotron light sources. The availabledetectors do not permit time resolved studiesof dynamic processes or structural analysis of crys-tals with very large unit cells or large mosaicity.Therefore, in biological structure research there isa need and a challenge for new developments.

9. Summary and outlook

The successful development of imaging detectorsrequires some understanding of basic imaging the-ory. Otherwise a reliable comparison of new de-tectors with existing systems is not possible.

Digital detectors for standard medical X-rayequipment are at the industrial production thresh-old.

With Si-strip detectors the important step ofusing them ‘edge-on’ for scanning systems has beenmade.

The NIKOS system for intravenous coronaryangiography with synchrotron radiation is in themedical evaluation phase. About 300 patients havealready been examined.

Synchrotron radiation is also used in new prom-ising imaging modalities, namely in phase contrastand in diffraction enhanced imaging, opening upnew possibilities in medical diagnostics because ofincreased object contrast.

Detectors for nuclear medicine have made a bigstep away from Anger cameras, using new scintil-lators and position sensitive photodetectors.

In biocrystallography and in small angle X-rayscattering high-performance systems are still ur-gently required. There are many other aspects inbio-medical research, where new radiation de-tectors can make important contributions.

New projects, if properly chosen, are always use-ful and rewarding.

Acknowledgements

I would like to thank the members of the or-ganising committee of the WCC98 for their invita-tion and kind hospitality. I enjoyed to meet againmany old friends after I changed my field of workabout 10 years ago.

I also would like to thank F. Arfelli, E. Hell, S.Majewski, R.H. Menk, D. Nygren and E.W. Ottenfor useful information and many interesting dis-cussions.

My special thanks go to H. Plothow-Besch forcarefully reading the manuscript and for manycritical remarks and suggestions for improvements.

H.J. Besch/Nucl. Instr. and Meth. in Phys. Res. A 419 (1998) 202—216 215

I. INVITED TALKS

Page 15: Radiation detectors in medical and biological applications1 · Radiation detectors in medical and biological applications1 H.J. Besch* Universita(t-Gesamthochschule Siegen, Fachbereich

References

[1] K. Doi (Ed.), Recent Developments in Digital Imaging,New York 1985, ISBN 0-88318-463-X.

[2] G. Zanella, R. Zannoni, Nucl. Instr. and Meth. A 381(1996) 157.

[3] R.H. Menk et al., Nucl. Instr. and Meth. A 398 (1997) 351.[4] P. Ottonello et al., Nucl. Instr. and Meth. (1998). These

Proceedings.[5] K. Engelke, Ph.D. Thesis, Hamburg 1989, DESY Internal

Report, F41/89-12.[6] M. Spahn et al., Electromedica 65 (1997) 37.[7] W. Zhao et al., Medical Phys. 22 (1995) 1595.[8] J. Rowlands, S. Kasap, Phys. Today (1997) 44.[9] S.E. Baru et al., Nucl. Instr. and Meth. A 419 (1998)

295.[10] F. Arfelli et al., Nucl. Instr. and Meth. A 392 (1997) 188.[11] C. Colledani et al., Nucl. Instr. and Meth. A 395 (1997) 435.[12] E. Beuville et al., High resolution X-ray imaging using

a silicon strip detector, IEEE Trans. Nucl. Sci. (1997).[13] P. Fessler et al., A major step forward in continuous

spectroscopic imaging of ionising radiation, Strasbourg,preprint, 1998.

[14] S. Cadeddu et al., Nucl. Instr. and Meth. A 419 (1998) 270.[15] H.J. Besch, Nucl. Instr. and Meth. A 360 (1995) 277.[16] H.J. Besch et al., Phys. Med. 9 (1993) 171.[17] C.W. Hamm et al., Herz 21 (1996) 127.[18] M. Lohmann, Nucl. Instr. and Meth. A 419 (1998) 276.[19] W. Thomlinson et al., Rev. Scient. Instr. 63 (1992) 625.[20] A. Charvet, private communication and http://www.esrf.fr/

exp-facilities/bl14.

[21] A. Snigirev et al., Rev. Scient. Instr. 66 (1995) 5487.[22] D. Chapman et al., Phys. Med. Biol. 42 (1997) 2015.[23] F. Arfelli, private communication.[24] Y. Deychak et al., Am. Heart J. 129 (1995) 219.[25] D. Hausmann et al., Am. J. Cardiol. 74 (1994) 857.[26] H.U. Kauczor et al., Radiology 201 (1996) 564.[27] R. Surkau et al., Nucl. Instr. and Meth. A 384 (1997) 444.[28] W.W. Moses et al., IEEE Trans. Nucl. Sci. 44 (1997) 1487.[29] R.J. Ott, Nucl. Instr. and Meth. A 392 (1997) 396.[30] A. Bolodzyna et al., A cylindrical Compton SPET camera,

IEEE Trans. Nucl. Sci. (1997).[31] F. Vittori et al., IEEE Trans. Nucl. Sci. NS- 44 (1997)

1127.[32] S. Majewski, Contribution to the Satelite Workshop on

Applications, Vienna, February 28, 1998.[33] T. Malatesta et al., Nucl. Phys. B 61 (1998) 658.[34] M. Moszynski et al., Nucl. Instr. and Meth. A 372 (1996)

51.[35] A. DelGuerra et al., IEEE Trans. Nucl. Sci. NS- 43 (1996)

1958.[36] C. D’Ambrosio et al., Nucl. Phys. B 61 (1998) 638.[37] M. Dahlbom et al., IEEE Trans. Nucl. Sci. 44 (1997)

1114.[38] H.J. Besch et al., Twodimensional gas pixel detectors for

biological structure research, Proc. 36th INFN ELOISAT-RON Workshop on New Detectors, 1—7 November, 1997,Erice, Sicily, Rev. Scient. Instr., to be published.

[39] R. Fourme, Nucl. Instr. and Meth. A 392 (1997) 1.[40] R. Lewis, Phys. Med. Biol. 42 (1997) 1213.[41] F. Arfelli et al., IEEE Trans. Nucl. Sci. NS- 44 (1997) 2395.[42] E. Beuville et al., Nucl. Instr. and Meth. A 406 (1998) 337.

216 H.J. Besch/Nucl. Instr. and Meth. in Phys. Res. A 419 (1998) 202—216