Impact of high energy resolution detectors on the performance of a PET system dedicated to breast cancer imaging

2004 Workshop on the Nuclear Radiology of Breast Cancer Rome (Italy) October 22-23, 2004

,Physica Medica,, 9 Vol. XXI, Supplement 1, 2006

Impact of High Energy Resolution Detectors on the Performance of a PET System Dedicated to Breast Cancer Imaging Craig S. Levin, Angela M. K. Foudray, Frezghi Habte Department of Radiology and Molecular Imaging Program Stanford University School of Medicine, Stanford (USA)

Abstract We are developing a high resolution, high sensitivity PET camera dedicated to breast cancer imaging. We are studying two novel detector technologies for this imaging system: a scintillation detector comprising layers of small lutetium oxyorthosilicate (LSO) crystals coupled to new position sensitive avalanche photodiodes (PSAPDs), and a pure semiconductor detector comprising cadmium zinc telluride (CZT) crystal slabs with thin anode and cathode strips deposited in orthogonal directions on either side of each slab. Both detectors achieve 1 m m spatial resolution with 3-5 m m directly measured photon interaction depth resolution, which promotes uniform reconstructed spatial resolution throughout a compact, breast-size field of view. Both detector types also achieve outstanding energy resolution (< 3% and < 12~ respectively for LSO-PSAPD and CZT at 511 keV). This paper studies the effects that this excellent energy resolution has on the expected system performance. Results indicate the importance that high energy resolution and narrow energy window settings have in reducing background random as well as scatter coincidences without compromising statistical quality of the dedicated breast PET data. Simulations predict that using either detector type the excellent performance and novel arrangement of these detectors proposed for the system facilitate -20% instrument sensitivity at the system center and a peak noise-equivalent count rate of > 4 kcps for 200 microCi in a simulated breast phantom. K~YwoRDs: Breast cancer imaging, PET, detectors, cadmium zine telluride, scintillation detectors, position sensitive avalanche photodiodes. 1. INTRODUCTION

Positron Emission Tomography (PET) is a noninvasive, in-vivo, functional and molecular imaging technology that has shown promise for more specific identification of cancer due to its unique ability to sense and visualize increased biochemical and molecular changes in malignant compared to healthy tissue. However, to date conventional PET has not been incorporated into standard practice for breast cancer management due to: [1] awkward, low photon efficiency geometry for breast imaging and relatively long scan times/low throughput of standard PET systems, [2] non-optimal spatial and contrast resolutions for early breast cancer identification, [3] relatively high cost; and [4] the lack of tracers with adequate specificity. Low coincident photon detection efficiency ( 1%) of the conventional clinical PET system limits the statistical quality of the data and achievable reconstructed spatial resolution, which reduces detection sensitivity for small lesions. Low efficiency also means relatively long scan times, which, together with the inappropriate large diameter ring geometry, make conventional PET impractical for breast imaging. Low detection efficiency together with inadequate spatial resolution (6-12 mm) mean poor visualization of lesions smaller than 0.5 cm 3, which further reduces PET's sensitivity for early detection of breast cancer, associated metastases, or monitoring after therapeutic interventions or treatments.

Improved delineation of smaller lesions is important since it translates to earlier detection and improved prognosis of malignant tumors and their metastases [1]. The open geometry and relatively poor energy and coincident time resolutions of conventional PET systems increases acceptance of background scatter and random photon coincidence events from both within and outside the camera field-of-view (FOV), which reduces achievable contrast resolution. Finally, the relatively high cost for standard PET instrumentation limits PET's cost-effectiveness for a variety of breast cancer indications. We are developing a PET system dedicated to breast cancer imaging that will address the issues of low photon detection efficiency, long scan times, awkward geometry, non-optimal spatial and energy resolutions, and high cost of PET for breast cancer imaging. Just as a dedicated system (mammography) is required to optimize X-ray breast imaging, we argue that a dedicated camera is crucial to optimize PET breast imaging, and without such a system, we are far from reaching PET's potential to play a role in breast cancer management. A new high performance detector will be configured into a compact system dedicated to close-proximity, breast and/or axillary node imaging with very high coincident photon detection efficiency and ultra-high spatial, energy, and coincident time resolutions. If used in conjunction with a highly specific breast cancer tracer, this high photon efficiency and high spatial resolution imaging system would facilitate the early

Address for correspondence: Craig S. Levin, Ph.D., Stanford University School of Medicine, 300 Pasteur Drive, MC 5128 Alway Building, Room MOO1 Stanford, CA 94305-5128, Phone: 650-736-7211, Fax: 650-724-1499. E-mail: [email protected] 3. New Camera Instrumentation

<,Physica Medica,, . Vol. XXI, Supplement 1, 2006 identifica:ion of developing, spreading or recurring breast cancer. Using a tracer with non-specific uptake (e.g Fluorodeoxyglucose (FDG)], there will likely be activity generated from the nearby heart and liver. Excellent energy and time resolutions and flexible positioning of the camera to reduce out o f FOV acti':ity will help to reduce the resulting high backgro Jnd scatter and r a n d o m coincidence rates that can degrade lesion contrast resolution. As is true for ~nammography, imaging the chest wall at close pr(~::imity is challenging. However, a dedicated breast im~Lging g e o m e t r y may be opened wider and oriented !avorably to contain the chest wall. Finally, a small, r,_qatively inexpensive camera dedicated to breast im Lging could significantly reduce study costs of the technique to the point that it becomes cost-effective fir: a variety of breast cancer indications. The 511 keV p h o t o n detectors used and their geometti::al configuration about the patient determine the resulting performance characteristics o f the dedictted breast imaging system. The system performance goals are 1 m m 3 spatial resolution achievec ] n i f o r m l y within the FOV, > 10% coincidence del:ection efficiency (a.k.a. photon sensitivity), and > 12% energy resolution for 511 keV photons. In this pi~per we discuss the importance o f using 511 keV d erectors with very high energy resolution configure.] into a high p h o t o n sensitivity system for dedicated PET imaging o f the breast. We demonstrate tl';t for a PET system built with detectors

that have excellent energy resolution, one may use a very n a r r o w energy w i n d o w setting to significantly reduce both scatter and r a n d o m coincidence events to obtain high contrast resolution, while also achieving high counting statistics and signal-to-noise ratio (SNR) within that n a r r o w window. 2. MATERIALS AND METHODS

2. 1. 511 keV photon detectors under study In this paper we studied two types o f high performance 511 keV p h o t o n detector modules that are candidates for the breast imaging system. The first detector m o d u l e design uses 1 m m lutetium oxyorthosilicate (LSO) scintillation crystal pixels coupled to n e w position sensitive avalanche photodiodes (PSAPD) [2, 3]. The crystals are coupled to the PSAPD in a novel configuration that promotes nearly complete (> 90%) scintillation light collection efficiency independent of p h o t o n interaction depth [2]. Incoming photons see > 2 cm thick LSO for high (> 82%) intrinsic (single 511 keV photon) detection efficiency. The PSAPD and crystal segmentation facilitate direct m e a s u r e m e n t of p h o t o n interaction depth within the scintillation crystal array [2, 3] (Figure 1, left). The other detector under study is c a d m i u m zinc telluride (CZT) semiconductor crystal with a cross-strip cathode (top surface) and anode (bottom surface) electrode patterns. These C Z T arrays are arranged in an innovative "edge-

511 pho

T ooi 8O0

22Na energy

60O

spectrum in

500[

10.82% 40C

2 cm

thick

1 ": i • J

iiiiii'i

fwhm at 511 keV

30( 20( 10(

2

3

4

5

6

7

8

9

10

Pulse height (Volts)

m,r ~LSO & &/b & 4~/I~/b & Need: extremely thin PSAPD

FIG. 1. Left: Depiction of proposed position sensitive 511 keV photon scintillation detector module. The detector comprises layers of 1 mm L!;O crystal arrays (white) coupled to PSAPDS (gray) in a novel sideways arrangement that promotes uniform and high light collection efficiency.Each layer has two separate crystal arrays coupled to separate PSAPD chips. Incoming photons see a minimum :hickness of -2 cm of LSO with directly measured interaction depth resolution of 3 mm. Extremely thin PSAPDS are requi~e;t to preserve a high crystal packing fraction. Right: 22Na energy spectrum extracted from one 1 x 1 • 3 mm LSO crystal fro ra one LSO-PSAPD array layer. 29

Craig s. Levin et alii: Impact of High Energy Resolution Detectors on the Performance of a PET System on' orientation (Figure 2, left) that allows incoming photons to see > 4 cm thick C Z T for high (> 86%) intrinsic 511 keV p h o t o n detection efficiency [4]. These detectors also have three-dimensional (3-D) p h o t o n interaction positioning capabilities within the detector volume [4]. Figures 1 and 2 (right) also show 511 keV p h o t o n energy spectra acquired in the two novel detectors demonstrating the excellent energy resolution performance of < 3 and < 12% fullwidth-at-half-maximum (fwhm) energy resolution achieved, respectively, for the C Z T and LSO-PSAPD detectors. The measured coincidence time resolution spectra measured for the two detector types had typical values of 2 and 8 ns fwhm, respectively for the LSO-PSAPD, and C Z T detectors. The time resolution of the C Z T is relatively poor compared to LSO since semiconductor detectors rely on charge transport rather than scintillation light transport for their signal formation. 4 x 4 x 0.5 cm 3 CZT crystals

We used Monte Carlo simulations [5] to study the geometrical configuration and performance of a dedicated breast imaging PET system built from the proposed position sensitive p h o t o n detector arrays. We studied dual-plate [6] as well as "box' [7] configurations of the detector modules (Figure 3). In this paper we focus on results of p h o t o n sensitivity and SNR of PET data for the box configuration using Monte Carlo simulated PET acquisitions that incorporated the accurate geometry, including all expected dead gaps, of the PET detector modules arranged in the box arrangement. The inner box gantry FOV dimensions for both types of detector systems (CZT or LSO-PSAPD) were 16 x 16 x 12 cm ~. The box thickness is 2 cm using the LSOPSAPD or 4 cm using C Z T (Figures 1 and 2). For all simulations we assumed the experimentally meas-

511 keV photon

-

40 Ano( strir

2.2. Monte Carlo Simulations of Detector Gantry

T

I

4

"7

/////

Cross Strip Detector

f ,

Collimatedbeam at 511 keV (Na 22) Correctedand selectedfor Cathode/Anode 0.1 to 1.03

3~o

_

.kd_ .....

;

Red: corrected-

FWHM

<

~

= 2.5%

I~

e

em

-1

ji

Cathode strips (on back, orthogonal to anode strips)

AAA/

0

~00

2oo

30o

400

5o0

PHA Channel

FIG. 2. Left: Depiction of proposed position sensitive 511 keV photon semiconductor detector module. The detector comprises layers of 40 • 40 x 5 m m 3 CZT (grey). Orthogonal strips of cathodes and anode electrodes are deposited on each side of each CZT slab. Signals induced on the anodes, cathodes, as well as their ratio are used to determine the 3-D interaction coordinates. The detector arrays are arranged edge-on so that incoming 511 keV photons see a minimum thickness of - 4 cm of CZT with directly measured interaction depth resolution of 3-5 mm. The CZT layers can be packed together with > 99% packing fraction. Right: =Na energy spectrum extracted from one CZT slab layer before (black) and after (grey) correction for variations in charge collection for different interaction locations between the anode and cathode using the ratio of cathode/anode signals.

Dual-Plate: 10 x 15 cm2plates

Box: 16 x 16 x 12 c m 3 F O V | I l l

. . . .

/

i .... , i,

i,,

I ii l I II I

FIG. 3. Two potential configurations of the proposed LSO-PSAPDor CZT detector modules. Left: Dual-plate configuration; Right: Box configuration. This paper focuses on performance results of the box configuration. 30

<> 9Vol. XXI, Supplement 1, 2006 ured valu ~s o f 3 and 12% f w h m energy resolution at 511 ke ~/, and 8 and 2 ns f w h m coincidence time resolutiol:., respectively for the LSO-PSAPD and C Z T detector configurations. For the p h o t o n sensitivity studies, we translated (in simulations) a point source o t annihilation photons from the center to the edge of the FOV of the box along the center "axis' of the box and measured the total n u m b e r o f coincider t 511 keV p h o t o n pairs detected within the detectors for various energy and coincidence time w i n d o w :!ettings. For the measure of SNR of the dedicated breast PET system data, the 16 x 16 x 12 cm 3 sen,
culations, completely filling the active volume o f the system with activity represents the worse case scenario, however out o f FOV activity such as from the hot heart or liver that could have important impact on detected event rates in a dedicated breast g e o m e t r y was not simulated in this study. 3. RESULTS Figure 4 plots coincident p h o t o n detection efficiency (photon sensitivity) as function of position along the axis from the center to the edge for b o t h the LSO-PSAPD and C Z T systems for two different energy w i n d o w s (2 x and 3 x the energy resolutions) and coincident time w i n d o w s equal to 2 x and 1 x the coincidence time resolutions, respectively for the LSO-PSAPD and C Z T detector types. Near the system center the predicted point source p h o t o n sensitivity is ~20% for b o t h detector types. Figures 5-8 plot the true, scatter, r a n d o m and N E e coincidence rate as a function o f energy w i n d o w for different time windows and as a function of coinComparison of Scnsitivity for Box Geometry

i

~2-4

,'

i i

....... .. '"

25

I

,'<

16cm

!

,,

!

.->, >=_

....

'i~ i..~.......i~_ I~-I .I-i......

"" . . . . . . . . . . . . . .

12 c m I

Jr . . . . . . . . .

i

[ ...................

~

CZT,488-534 keV,8ns LSO,481-541 keV,4ns LSO,450-572keV,4ns

201

I

1

O9 t--

. . . . . . . . . . . . . .

. . . . . .

---:

/ ....

[- CZT'~176176176176 ]1

> iicm

o 15 n

z-axis

0

1

2

3

4

5

6

A x i a l position from the center [cm]

Fig. 4. Left: A point I~F source was translated from the center to the FOV edge along the axis of a box configuration of detectors. Right: Photon sensitivity (coincident photon detection efficiency) vs axial position using the proposed CZT and LSO-PSAPD position serLsitive detectors systems for the narrow energy window settings shown. Trues vs Energy Window for 200 uCi in Breast

Trues vs Time Window for 200 uCi in Breast

300"

300 CZT

LSO

250

250

200!

200

LSO

LSO

8- 150q m O

o

100ii 50

,,~

i

~150

CZT, 8 n s CZT, 12ns CZT, 16ns LSO, 2 n s LSO, 4 n s LSO, 6 n s

*~ 100

CZT, 503-518 keV CZT, 496-526 keV CZT, 488-534 keY LSO, 480-541 keV ~ - L S O , 447-574 keV LSO, 419-603 keV

3 50

10

20

30

40

E n e r g y W i n d o w [% c e n t e r e d

50

at 511 keV]

60

00

2

4

6

8

10

Time Window

12

14

16

18

20

[ns]

Fig. 5. Trm coincidence rate for CZT and LSO-PSAPD detector systems vs energy window setting for various coincidence time windows [l:ft) and vs coincidence time window for various energy windows (right). For the simulations, 200 gCi of activity was uniformly ,:[istributed in simulated breast tissue that completely filled the 16 • 16 • 12 cm3 box volume. 31

Craig s. Levin et alii : Impact of High Energy Resolution Detectors on the Performance of a PET System cidence time window for different energy windows for the two detector systems. The peak NEC values are 4.2 and 4.3 kcps using 24 and 6% energy window and 4 and 8 ns time window, respectively, for LSO-PSAPD and CZT detectors and 200 btCi total activity uniformly distributed within the 16 • 16 • 12 cm 3 simulated breast tissue volume.

fwhm) compared to the CZT (8 ns fwhm) configuration, the random coincidence rate is generally lower. However, for the proposed CZT coincidence time window of 8 ns and energy window of 6% (496-526 keV), the random coincidence rate is comparable for both the CZT ( - 9 kcps) and LSOPSAPD (-11 kcps) designs. The reason for this is that that the random coincidence rate varies as the square of the singles rate, but only to the first power of the coincidence time window. The singles rate varies with the energy window width. Thus, the random coincidence rate varies as the square of the energy window and using a narrow energy window will reduce the random coincidence rate significantly. This is a result of the fact that many detected single photons undergo scatter in the tissue before absorption in the detector material and a narrow energy window helps to reject such events. Thus, even though the coincidence time resolution is relatively poor for CZT, the excellent energy resolution and narrow energy window setting keep the random coincidence rate under control. Figure 8 shows that both the LSO-PSAPD and CZT box configurations have similar peak NEC values, assuming a total of 200 btCi uptake in the breast tissue. The LSO-PSAPD achieves this peak at lower coincidence time window setting of 4 ns compared to 8 ns for the CZT design, but at a higher energy window of 24% compared to 6% for the CZT. The CZT's excellent energy resolution (3% fwhm) and narrow window (6%) allows superior rejection of scatter coincidences and makes up for the relatively poor coincidence time resolution (8 ns fwhm) to control random coincidence rate. Thus, the CZT system achieves similar peak NEC performance to the LSO-PSAPD system. Similar effects of reduced randoms rate with higher energy resolution and narrow energy window setting were seen with a mouse and rat phantom in a dedicated CZT small

4. DISCUSSION Figure 4 shows that unprecedented photon sensitivity (20%) near the center is possible with a dedicated breast PET geometry that collects photons at close proximity, using relatively thick, tightly packed detector crystals for either the LSO-PSAPD or CZT detectors that are proposed. As with any 3-D PET system, the sensitivity drops off rapidly as the point source approaches the edge of the sensitive FOV. Figure 5 shows that relative to the CZT design, the LSO-PSAPD box configuration has a 10-15% higher true coincident rate for an energy window equal to twice the energy resolution for each detector type and a time window equal to 4 ns for the LSO-PSAPD design and 8 ns for the CZT system. This higher trues rate is likely due to the fact that LSO has a higher probability for photoelectric absorption and the coincidence time window setting for the simulated LSO-PSAPD system was 2 x the time resolution rather than 1 x the time resolution for the simulated CZT design. However, as seen in Figure 6, the scatter coincidence rate is nearly a factor of eight lower for the CZT compared to the LSO-PSAPD system, since the 511 keV photopeak energy resolution and window setting is significantly narrower (3% fwhm and 6%, respectively for CZT and 12% fwhm and 24%, respectively for LSO-PSAPD). Figure 7 shows that due to the superior coincidence time resolution for the LSO-PSAPD (2 ns Scatter vs Energy Window at 200 uCi,

Scatter vs Time Window at 200 uCi,

200 18(] 160

14(] "~

200

CZT,503-518keY CZT,496-526keV CZT,488-534keV LSO,480-541keV 140 --~- LSO,447-574keV LSO,419-603keV 120

I_.~_ CZT, 8ns -~- CZT,12ns CZT,16ns

,, ~,,

160

LSO, 2ns LSO, 4ns LSO, 6ns

12(]

LSO

_100

100 80

o

.o

i!

LSO

w 60 0

40

J

40

LSO

20

~

lo

20

Energy Window

3'0

1o

so

o c e n t e r e d a t 511 keV] [%

60

0

0

2

4

, 6

, 8

.

10

.

.

12

14 T i m e W i n d o w [ns]

.

.

16

18

20

Fig. 6. Scatter coincidencerate for CZT and LSO-PSAPDdetector systemsvs energy window setting for various coincidencetime windows (left) and vs coincidencetime window for various energy windows (right). For the simulations, 200 gCi of activity was uniformly distributed in simulated breast tissue that completelyfilled the 16 • 6 • 12 cm3 box volume. 32

< 9Vol. XXI, Supplement 1, 2006 R a n d o m s vs Energy W i n d o w at 200 uCi,

R a n d o m s vs T i m e W i n d o w at 200 uCi,

,1.c,

6(]

CZT, 8ns -..e.-,, CZT,12ns ,10 CZT, 16ns LSO, 2ns I.t; - ~ - LSO, 4ns LSO, 6ns

~

~-"~5C

~

-

CZTf.~........

311

I!i

--*'- CZT,488-534 keV LSO, 480-541 keV - ~ LSO,447-574 keV I~ LSO,419-603 keV

~ ~ .j-/ . / * ( /--~

//

4(]

/

o

CZT, 503-518 keV

~-~ CZT,496-526 keV

3(]

~

CZT ~

"

LSO

~,

20

.....

I(I

1(]

10

20

30

40

50

0

60

2

4

6

Energy Window [% centered at 511 keV]

8

10

12

14

16

18

20

Time Window [ns]

Fig. 7. Random coincidence rate for CZT and LSO-PSAPD detector systems vs energy window setting for various coincidence time windows (left) and vs coincidence time window for various energy windows (right). For the simulations, 200 btCi of activity was unifor: nly distributed in simulated breast tissue that completely filled the 16 • 16 • 12 cm3 box volume. N E C v s E n e r g y W i n d o w at 2 0 0 uCi,

N E C vs T i m e W i n d o w at 2 0 0 uCi,

9L ! I.

4.5 r

CZT

CZT

LSO 3..= 3

~

_~.!;

~2.5

O

CZT

cz-r, 8ns CZT,12ns CZT, 16ns LSO, 2ns "~'o' LSO, 4ns LSO, 6ns

c-

-~

I.

).5

r

/

CZT, 503-518 keY CZT, 496-526 keY CZT, 488-534 keY LSO, 480-541 keV LSO, 447-574 keV LSO, 419-603 keV

1.5

0.5

o

;o

2'0

;o

4o

~o

Energy Window [% centered at 511 keV]

60

%

i

a

6

~

1'o

1'= 1'4 1'6 l'a

20

Time Window [ns]

Fig. 8. Noi!e-equivalent count rate (NEC) for CZT and LSO-PSAPD detector systems vs energy window setting for various coincidence time windows (left) and vs coincidence time window for various energy windows (right). For the simulations, 200 gCi of activity,~vas uniformly distributed in simulated breast tissue that completely filled the 16 x 16 x 12 cm3 box volume.

animal PET system [4]. More realistic simulated breast p h a n t o m configurations should also include out o f FOV activities from organs such as the heart or liver, which this study did not address. However, with a m o r e realistic p h a n t o m we expect the same general trends we have observed here to hold for both detector systems under study.

resolution is poor. The detector configuration proposed yields nearly 20% instrument sensitivity that facilitates the reconstructed spatial resolution goal of 1 m m 3 with high SNR. These qualities o f outstanding spatial resolution and contrast resolution translate into superior lesion visualization and quantitative accuracy.

5. CONC f,USION

ACKNOWLEDGEMENTS

Results f i o m simulation o f a dedicated PET breast imaging g e o m e t r y indicate that using detectors with excellent (< 12% fwhm) 511 keV energy resolution, one mav use a very narrow energy w i n d o w setting to rejed ~'andom in addition to scatter coincidences for high contrast resolution, while maintaining high true coir cidence counts and SNR. This i m p o r t a n t effect o( energy resolution makes the use o f C Z T feasible i a PET even t h o u g h its coincidence time

The authors would like to t h a n k Dr. James Mattesson at the UC San Diego Center for Astrophysics and Space Studies (CASS) for his help with developing and testing C Z T detectors and Richard Farrell from RMD, Inc. for his efforts to develop an extremely thin PSAPD device. This work was supported in part by NIH-NCI grants P,21 CA098691 and RO1CA119056 and UC Breast Cancer Research P r o g r a m grant 12IB-0092. 33

Craig S. Levin et alii: Impact of High Energy Resolution Detectors on the Performance of a PET System

REFERENCES [1]

[2]

[3]

[4]

34

Silverstein M J, Lagios M D, Groshen S, Waisman J R, Lewinsky B S, Martino S, Gamagami P, Colburn W J. The influence of margin width on local control of ductal carcinoma in situ of the breast. N EnglJ Med 1999: 340; 1455-1461. Levin C S. Design of a high-resolution and high-sensitivity scintillation crystal array for PET with nearly complete light collection. IEEE Transactions On Nuclear Science Part 1 Oct 2002:49 (5); 2236-2243. Levin C S, Foudray A M K, Olcott P D et alii. Investigation of position sensitive avalanche photodiodes for a new high-resolution PET detector design. IEEE Transactions On Nuclear Science Part 2Jun 2004 51 (3); 805-810. Levin C S, Matteson J, Habte F, Skelton T, Pelling M, Duttweiler F. Promising Characteristics and Performance of Cadmium Zinc Telluride Detectors for Positron

[5]

[6]

[7]

Emission Tomography. Presented at the 2004 IEEE Nuclear Science Symposium and Medical Imaging Conference, Rome (Italy), October 2004. Abstract # M2-117, Book of Abstracts, P. 136, IEEE Nuclear and Plasma Sciences Society. Strul D, Santin G, Lazaro D, Breton V, Morel C. gate (GEANT4 Application for Tomographic Emission): a PET/SPECT general-purpose simulation platform. Nucl Phys B (Proc Suppl) 2003; 75-79. Weinberg I N, Beylin D, Zavarzin V e t alii. Positron emission mammography: High resolution biochemical breast imaging. Technology in Cancer Research and Treatment 2005 Feb: 4 (1). Moses W W, Qi J Y. Instrumentation Optimization For Positron Emission Mammography Nuclear Instruments & Methods. In: Physics Research Section A, Accelerators Spectrometers Detectors And Associated Equipment Jul 11 2004:527 (1-2); 76-82.