A detector for use in a dedicated PET scanner for cancer applications,
particularly breast cancer applications, using a LSO scintillator,
a lightguide coupling arrangement, and an efficient way to construct
the scintillator array, which provides a flexible imaging system
for breast cancer applications with high sensitivity and high spatial
resolution in a compact, cost effective, design.
1. A positron emission tomography imaging apparatus comprising:
at least two, opposed detectors, said detectors having an array
of scintillation crystals, a plurality of photomultiplier tubes
positioned adjacent said plurality of arrays, and a lightguide having
an end positioned adjacent to said array of scintillation crystals
and having an opposing end adjacent to said photomultiplier tubes.
2. The apparatus of claim 1, wherein each array comprises at least
3. The apparatus of claim 1, wherein said crystals are lutetium
oxyorthosilicate (LSO) or light-output equivalent crystals.
4. A PET imaging apparatus comprising at least two detector plates,
each plate comprised of at least one detector, said detector having
a scintillator coupled to one end of a lightguide, the opposing
end of said lightguide coupled to a photomultiplier tube.
5. The apparatus of claim 4 wherein said scintillator comprising
an array of lutetium oxyorthosilicate (LSO) scintillator crystals.
6. The apparatus of claim 4 wherein said lightguide is an optical
7. A method for examining a body part comprising: providing an
internal image of the body part including, a positron emitting radioisotope
and a positron recording apparatus between which the body part is
to be disposed; and placing at least two detector plates, each plate
comprised of at least one detector, said detector having a scintillator
coupled to one end of a lightguide, the opposing end of said lightguide
coupled to a photomultiplier tube, said detector is capable of detecting
gamma-rays emitted by the radioisotope infiltrated into the body
part in an adjacent relationship with said recording apparatus for
providing the internal image.
8. The method of claim 7, wherein said scintillator comprises an
array of lutetium oxyorthosilicate (LSO) scintillator crystals or
light-output equivalent crystals.
9. The method of claim 7, wherein said lightguide is an optical
CROSS-REFERENCE WITH RELATED APPLICATIONS
 This application claims the benefit of U.S. Provisional
Application No. 60/170,746, filed Dec. 14, 1999, which is herein
incorporated by reference.
BACKGROUND OF THE INVENTION
 1. Field of the Invention
 This invention relates generally to an apparatus and method
for medical examination, in particular, a lutetium oxyorthosilicate
(LSO) or light-output equivalent positron emitting tomography (PET)
 2. Description of Related Art
 The American Cancer Society has predicted that there will
be more than 181,000 new breast cancer cases and more than 40,000
deaths from breast cancer in the United States in 2000. [American
Cancer Society, "Cancer Facts and Figures-1999," American
Cancer Society, Atlanta, Ga. (1999).] Breast cancer is also the
second leading cause of cancer death in women. Currently, mammography
and physical breast examination, provide the two most effective
methods for screening potential breast cancer patients. Although
mammography allows the detection of very small, non-palpable lesions,
it has a limited diagnostic accuracy for detecting cancer and image
interpretation is subject to considerable inter-observer and intra-observer
variability. The incidence of positive biopsies performed after
mammographic findings ranges from 9% to 65%, with most investigators
reporting a 15 to 30% positive biopsy rate. The sensitivity of detection
by mammography drops considerably in women with dense, fibrocystic
 Microcalcifications, one of the classic signs of occult
malignancies, have a low predictive value of only 11.5% for the
presence of cancer. The predictive value of masses that are thought
to definitely represent malignancies is about 74%, but masses thought
to be possibly malignant turn out to be carcinoma in only 5.4% of
the cases. [M. Moskovitz, "The predictive value of certain
mammographic signs in screening for breast cancer," Cancer,
51, 1007-1011 (1983)]. Also, several studies have reported substantial
variability among radiologists in interpretation of mammographic
examinations. [K. Kerlikowske, et al., "Variability and accuracy
in mammographic interpretation using the American College of Radiology
Breast Imaging Reporting and Data System," J. Natl. Can. inst.,
90, 1801-1809 (1998)]. Therefore, mammography is a useful screening
tool for detecting cancer, but it is limited by a large number of
false positive tests, which result in unnecessary biopsies. Mammography
is also limited by a considerable number of false negative tests,
which result in the missed diagnosis of cancer.
 It is also possible to use radio-pharmaceutical and radio-nuclide
imaging to detect cancers, such as [.sup.18F]fluoro-2-deoxy-D-glucose
(FDG). FDG is a radioactive analog of glucose, which is phosphorylated
and trapped within cells. After a patient receives a dose of FDG,
she may be examined with a detector that senses the gamma rays produced
by .sup.18F. Positron emission tomography (PET), using FDG as a
tracer of tumor glucose metabolic activity, is an accurate non-invasive
imaging technology which probes tissue and organ function rather
than structure. [See U.S. Pat. No. 5,453,623 and U.S. Pat. No. 5,961,457].
The increased rate of glycolysis in neoplastic cells, independent
of the oxygen concentration present, has been previously reported.
[O. Warburg, "On the origins of cancer cells," Science,
Vol. 123, 309-314 (1956) and U.S. Pat. No. 5,969,358]. This information
is fundamental to the utility of FDG for imaging human neoplasms.
 Whole body PET scanners are used clinically to diagnose
and to stage a wide variety of cancers. [C. K. Hoh, et al., "PET
in oncology: will it replace the other modalities?" Sem. Nucl.
Med., 27, 94-106 (1997)]. PET scanners detect breast cancer with
sensitivities between 70 and 90% and with specificities of 84-97%.
[N. Y. Tse, et al., "The application of Positron Emission Tomographic
imaging with fluorodeoxyglucose to the evaluation of breast disease,"
Ann Surg., 216, 27-34 (1992); O. E. Nieweg, et al., "Positron
Emission Tomography of Glucose Metabolism in Breast Cancer: Potential
for Tumor Detection, Staging, and Evaluation of Chemotherapy,"
Ann. N. Y. A. Sci., 698, 423-448 (1993); and R. L. Wahl, et al.,
"Primary and Metastatic Breast Carcinoma: Initial Clinical
Evaluation with PET with the Radiolabeled Glucose Analogue 2-[F-18]-Fluoro-2-deoxy-D-glucose,"
Radiology, 179, 765-770 (1991)]. A high diagnostic accuracy of PET
imaging for staging of axillary lymph node involvement has also
been reported. [L. Adler, et al., "Axillary lymph node metastases:
screening with F-18 2-deoxy-2fluoro-D-glucose (FDG) PET," Radiology,
203, 323-327 (1997)]. The. lower than desired diagnostic accuracy
reported for PET imaging is due to relatively poor accuracy for
detecting tumors of less than 1 cm in size. [N. Avril, et al., "Metabolic
characterization of breast tumors with positron emission tomography
using F-18 fluorodeoxyglucose," J Clin. Onc., 14, 1848-1857
 Most PET imaging technology is currently based on scintillation
detectors. Radiation detection begins by injecting isotopes with
short half-lives into a patient's body. The isotopes are absorbed
by target areas within the body, causing the isotope to emit positrons
that are detected when they generate gamma rays. When in the human
body, the positrons collide with electrons and the two annihilate
each other, releasing gamma rays. The emitted rays move in opposite
directions, leave the body and strke the array of radiation detectors.
In the majority of commercial PET systems, a "block" design
composed of a high-density, partially-segmented (for weighted light
sharing) scintillation crystal (bismuth germanate) is coupled to
four photomultiplier tubes (PMTs). [M E. Casey, et al., "A
multicrystal two dimensional BGO detector system for positron emission
tomography," IEEE Trans. Nucl. Sci., 33, 460-463 (1986) and
S. R. Cherry, et al., "A Comparison of PET Detector Modules
Employing Rectangular and Round Photomultiplier Tubes," IEEE
Trans. Nucl. Sci., 42, 1064-1068 (1995) and U.S. Pat. No. 5,453,623].
In this design, the scintillation crystal is subdivided into semi-discrete
crystals by incomplete cuts which are filled with reflecting material.
The PMTs are not position-sensitive and rely on the different depths
of the cuts in the scintillation crystal to yield a light distribution
on the PMT's which varies linearly with interaction position across
the detector. A problem with the block design of current PET systems
is that the intrinsic spatial resolution and the spatial sampling
of the block is determined by the size of the individual crystals.
In order to improve the intrinsic spatial resolution the size of
the crystals needs to be reduced. However, with the block design
it becomes difficult to decode smaller crystals. Another problem
inherent to the block design PET system is that it is fairly bulky,
because of the large dimensions of most single channel PMTs.
 More recently, high resolution, high sensitivity PET detectors
have been constructed by directly coupling the scintillator material
4 to a compact, low-cost, position-sensitive PMT (PS-PMT). By coupling
small discrete scintillator elements 4 directly onto the active
area of the PS-PMT, one maximizes light transmission from the scintillator
4 to the PS-PMT. [J. J. Vaquero, et al., "Performance Characteristics
of a Compact Position-Sensitive LSO Detector Module," IEEE
Trans. Nucl. Sci.,17, 967-978 (1998) and R. Pani, et al., "Multi-PSPMT
scintillation camera," IEEE Trans. Nucl. Sci., 46, 702708(1998)
and U.S. Pat. No. 5,864,141]. However, these PS-PMT's have a significant
inactive area at the edges. Using the direct coupling method and
tiling many detectors together to form planar arrays, therefore,
produces large gaps between the detector modules 20 because the
effective or active area 10 of the PMT 8 does not span the full
physical dimensions of the face of the tube (FIG. 1a). This reduces
system sensitivity and sampling and causes problems in the reconstruction
of the data. Therefore, it is desirable to develop some sort of
tapered light guide 12 to eliminate these gaps and to form large
continuous arrays (FIG. 1b). [R. Pani, et al., supra.]
 A PET camera based on discrete LSO scintillator elements
and a fixed ring geometry has been reported. [W. Moses, et al.,
"PET camera designs for imaging breast cancer and axillary
node involvement," J. Nucl. Med., 36, 69P (1995) and U.S. Pat.
No. 6,040,580]. However, the flexibility of the planar detector
arrays with variable separation of the present invention offers
advantages in the clinical setting over PET systems in a fixed ring
 Conventional PET imaging devices are designed to image cross
sections of the entire body. Although functional imaging with PET
is a promising technique in conjunction with x-ray mammography for
breast cancer patient management, there are several disadvantages
to employing a whole body PET scanner for the detection of malignant
breast tumors. The first disadvantage is that the whole body PET
system is limited by the spatial resolution and sensitivity. [N.
Avril, et al., supra]. Whole body PET systems typically yield reconstructed
images with a resolution of 8-15 mm, depending on the injected dose,
imaging time, and intrinsic resolution of the scanner. The effect
of this resolution limit is that radioactivity is underestimated.
 The second disadvantage of a conventional whole body PET
is the high cost of the examination. Whole body PET is an expensive
technology, and is generally only available in the larger medical
facilities in the United States.
 A third disadvantage of a conventional whole body PET scanner
is that the PET scanner provides metabolic images of breast cancer
patients with several shortcomings related to the general-purpose
nature of these systems, e.g., in whole body scanners the detectors
are typically 20-30 cm away from the breast or axilla, which reduces
sensitivity. Conventional scanners also have relatively large detector
elements (greater than 4 mm), which limits spatial resolution.
BRIEF SUMMARY OF THE INVENTION
 The present invention is an apparatus and method for examining
a body part. In particular, the present invention is directed to
a dedicated PET system for breast imaging or imaging other body
parts, such as the head, neck, liver, heart, lungs and other extremities,
which overcomes the limitations of prior detectors and improves
the overall diagnostic quality of the images.
 The positron emission tomography imaging apparatus of the
present invention is a dedicated mammary and axillary region PET
imaging system and comprises at least two large-area planar scintillation
detector plates composed of 25, a 5.times.5 array, of modular detectors.
The detectors include an array of scintillation crystals, a plurality
of photomultiplier tubes positioned adjacent the plurality of scintillation
crystals, and a lightguide having an end positioned adjacent to
the array of scintillation crystals and having an opposing end adjacent
to the photomultiplier tubes.
 The planar scintillation detector plates 22 operate in coincidence
and have about a 15.times.15 cm.sup.2 surface area, giving complete
coverage of the breast in a single view. The detector can be mounted
on a flexible gantry, allowing the inter-detector separation to
be varied from about 10 cm up to about 50 cm, and also allowing
the detectors to rotate to collect tomographic information.
 A method for examining a body part is also described and
comprises providing an internal image of the body part including,
a positron emitting radioisotope and a positron recording apparatus
between which the body part is to be disposed; and placing at least
two detector plates 22, each plate comprised of at least one detector,
said detector having a scintillator coupled to one end of a lightguide,
the opposing end of said lightguide coupled to a photomultiplier
tube, said detector is capable of detecting gamma-rays emitted by
the radioisotope infiltrated into the body part in an adjacent relationship
with said recording apparatus for providing the internal image.
 The system of the present invention allows for adjustable
detector separation to accommodate all patients and permits imaging
of the axillary region. The adjustable detector plate separation
also allows closeness to the area being imaged, thereby increasing
the system sensitivity. The detector plates 22 or arrays are large
enough to scan an entire breast in one imaging setup. The flexible
scanner geometry allows planar, limited angle, filtered back projection
or iterative image reconstruction techniques to be implemented.
 The present invention has a number of important advantages
over conventional PET scanners. The present invention brings the
detectors in close to the breast or axilla, resulting in a large
increase in the system sensitivity (the fraction of emitted gamma
ray pairs that are detected) which is due to the increase in solid
angle. This increase in sensitivity allows for an improved image
signal-to-noise and/or image resolution. The increase in sensitivity
also results in a reduced imaging time (increasing patient throughput)
and/or allows for a smaller injected dose of FDG. Furthermore, the
present invention allows for the gamma ray pairs to only pass through
the breast to be detected, and does not require them to pass through
the entire cross-section of the chest. Therefore, tissue attenuation
is reduced and the correction for gamma-ray attenuation can be based
on simple geometric calculations. Lastly, because relatively few
detector modules 20 are needed to construct the scanner of the present
invention (about 50 in the proposed system versus 250-300 in a whole-body
PET scanner), the overall cost of the technology is dramatically
reduced. This, in conjunction with the widespread availability of
FDG from the growing network of PET radio-pharmaceutical distribution
centers, results in PET becoming a viable diagnostic tool for breast
cancer patient management.
 This system offers several advantages when compared with
existing dedicated PET systems for breast imaging and conventional
whole-body PET scanners. Firstly, LSO scintillators provide important
advantages over dedicated systems that use BGO crystals. Lutetium
oxyorthosilicate (LSO) scintillators have a decay time of 40 ns
which provides count-rate performance advantages. Thus, LSO has
similar stopping power to BGO, but produces five times as much scintillation
light and has a seven-fold shorter decay time, which enables the
detector of the present invention to operate successfully in the
high singles count rate environment expected in breast imaging,
due to nearby activity from the heart and liver. The increased light
output allows good timing and energy resolution improving image
quality by reducing the influence of randoms and scatter.
 Secondly, the use of an optical fiber taper allows detector
modules 20 to be tiled together in planar arrays (with no gaps)
which produce detector plates 22 of any desired size. For example,
the 15.times.15 cm.sup.2 detector plates 22 of the present invention
provide a large field of view which provide for better coverage
and a shorter imaging time. Furthermore, these plates 22 are sensitive
and maintain their resolution right to the very edges, allowing
the closest possible imaging of the chest wall and imaging of the
entire body part or breast. [C. J. Thompson, et al., "Positron
emission mammography (PEM): A promising technique for detecting
breast cancer," IEEE Trans. Nucl. Sci., 42, 1012-1017 (1995);
C. J. Thompson, et al., "Feasibility study for positron emission
mammography," Med. Phys., 21, 529-537 (1994); J. L. Robar,
et al., "Construction and calibration of detectors for high
resolution metabolic breast imaging," Nucl. Instrum. Methods
Phys. Res. A, 392, 402-406 (1997); I. Weinberg, et al., "Preliminary
results for positron emission mammography: Real-time functional
breast imaging in a conventional mammography gantry," Eur.
J. Nucl. Med., 23, 804-806 (1996); R. Freifelder, et al., "Dedicated
PET scanners for breast imaging," Phys. Med. Biol., 42, 2463-2480
(1997); and G. Hutchins, et al., "Evaluation of prototype geometries
for breast imaging with PET radiopharmaceutical," J. Nucl.
Med., 36, 69P (1995)].
 The present invention is also surprisingly inexpensive,
as the tapered optical fiber bundles used in this work are a fraction
of the cost of the very high resolution tapers used in conventional
CCD based imaging systems. Lastly, the large-area plate geometry
with variable separation allows unprecedented flexibility for clinical
applications. The detector separation can be adjusted to suit the
patient geometry and planar, limited angle or full tomographic datasets
of the breast can be acquired. Planar images of the axilla may also
be acquired with our proposed system. Bringing the detectors in
close to the object of interest will improve sensitivity relative
to whole-body PET scanners, and the resolution and timing performance
of our detector modules 20 has been demonstrated to be superior
to that measured in whole-body PET detectors. The goal of the maxPET
system is to aid in breast cancer patient management by assisting
in imaging patients with dense, fibro-glandular breasts, detecting
axillary lymph node metastases without surgery and monitoring chemotherapy
 These and other features, aspects, and advantages of the
present invention will become better understood with regard to the
following detailed description, appended claims, and accompanying
BRIEF DESCRIPTION OF THE DRAWINGS
 FIG. 1a shows the direct coupling of scintillator arrays
into the photomultiplier tube 8 (PMT);
 FIG. 1b shows the use of a tapered light guide to couple
the light from the scintillator array into the PMT 8;
 FIG. 2 shows a position-sensitive PMT 8 (PS-PMT) with a
physical surface area measuring about 3.times.3 cm.sup.2 and an
active area of about 2.2.times.2.2 cm.sup.2;
 FIG. 3 shows a plastic grid (manufactured by Stratasys Inc.,
Eden Prairie, U.S.A.) with a 9.times.9 matrix of square holes that
measure about 3.times.3.times.5 mm.sup.3 and a wall thickness of
about 0.3 mm; 9 of the 81 lutetium oxyorthosilicate (LSO) crystals
are shown placed in the holes of the grid;
 FIGS. 4a to 4f are light guides, used to couple the about
3.times.3 cm.sup.2 scintillation array down to the about 2.times.2
cm.sup.2 active area of the PMT 8, showing: tapered LSO crystals
4 (FIG. 4a); individual tapered light guides made from glass (FIG.
4b); a piano-concave lens (PCV) cut into about a 3.times.3 cm square
(FIG. 4c); individual optical fibers (FIG. 4d); a tapered optical
fiber bundle having one end about 5.3 cm in diameter (FIG. 4e);
and a tapered optical fiber bundle having one end of about a 3.times.3
cm square, made by cutting down the 5.3 cm end of FIG. 4e (FIG.
 FIG. 5a is a schematic representation illustrating the measurement
of the energy resolution and the light collection efficiency;
 FIG. 5b is a schematic representation illustrating the measurement
of the flood histogram and the peak-to-valley ratio;
 FIG. 6 shows an assembled detector module, having a 9.times.9
array of about 3.times.3.times.20 mm.sup.3 LSO crystals coupled
through a tapered optical fiber bundle to a Hamamatsu R5900-C8 PS-PMT;
 FIG. 7 shows the flood histogram which was obtained by uniformly
irradiating the detector module with a .sup.22Na point source 14
whereby all 81 crystals from the LSO scintillator array are clearly
 FIGS. 8a to 8c are energy spectra graphs from the 81 crystals
from the 9.times.9 LSO array, the full-width at half maximum (FWHM)
of the 511 keV photopeak was measured to provide the energy resolution,
showing: the highest energy spectrum of about 23.8% (FIG. 8a); the
average energy spectrum of about 19.5% (FIG. 8b); and the lowest
(best) energy spectrum of about 17.1% (FIG. 8c);
 FIG. 9 shows a coincidence timing distribution from two
complete modules using a calibrated time-to-amplitude converter
(TAC), the timing resolution, measured as the FWHM of the distribution
is about 2.38 ns;
 FIG. 10 is a plot of the coincidence point spread function
across a row of crystals in the detector module, showing an average
spatial resolution of about 2.26 mm, the edge crystals have a better
resolution because of the reduced influence of inter-crystal scatter;
 FIG. 11 shows two planar detector plates having 5.times.5
arrays of the modular detectors used in the PET imaging system of
the present invention;
 FIG. 12 shows the maxPET detector assembly comprising two
opposing 15 cm.times.15 cm LSO detector plates mounted about 15
cm apart in an aluminum frame;
 FIG. 13 is the charge division scheme illustrating the multiplexed
readout scheme of the detectors;
 FIGS. 14a-b show the flood histograms of detector plate
1 (FIG. 14a) and 2 (FIG. 14b);
 FIG. 15 shows the coincidence timing spectrum of two detector
plates, indicating a FWHM of 8.1 ns;
 FIGS. 16a-b are images of the line bar phantom reconstructed
using the focal plane tomography algorithm, FIG. 16a is reconstru-cted
with a full acceptance angle and FIG. 16b is reconstructed with
a half angle of acceptance; and
 FIGS. 17a-b are row profiles taken through the line bar
phantom images, FIG. 17a is taken from FIG. 16a and shows a resolution
of about 4 mm, FIG. 17b is taken from FIG. 16b and shows a resolution
of about 5.5 mm.
DETAILED DESCRIPTION OF THE INVENTION
 The maxPET system of the present invention comprises at
least two 15.times.15 cm.sup.2 planar scintillation detector plates
22 operating in coincidence, with each plate composed of a 5.times.5
array (25) modular detectors (FIGS. 11 and 12). Each modular detector
is composed of three individual components: a photomultiplier tube,
a lightguide 12, such as an optical fiber bundle, and a scintillator
array. The scintillator array is comprised of a 9.times.9 array
of about 3.times.3.times.20 mm.sup.3 lutetium oxyorthosilicate (LSO)
scintillator crystals 4 or detector elements, which are coupled
to the lightguide 12, such as the optical fiber bundle, which in
turn is coupled to the position-sensitive photomultiplier tube 8
(PS-PMT), such as the Hamamatsu R5900-C8. The modular detectors
are thus read out by a 5.times.5 array of PS-PMT. A mutliplexing
readout scheme is also utilized to reduce the number of readout
channels from 200 (4X and 4Y readouts per PS-PMT) to 8 channels
 Although LSO crystals 4 are preferred, other light-output
equivalent crystals may be used, such as gadolinium oxyorthosilicate
(GSO), bismuth germanate (BGO), LGSO (a mixture of BGO and LSO),
yttrium aluminum pyrovskite (YAP), and sodium iodide (NaI(TI). In
another embodiment of the invention, the detectors are tiled together,
without gaps, to construct large area detector arrays to form a
dedicated PET cancer imaging system, preferably a breast cancer
imaging system. All detector elements are clearly visualized upon
flood irradiation of the module.
 Thus, the dedicated PET system of the present invention
takes advantage of the high specificity of FDG PET imaging and,
at the same time, improves the sensitivity for breast cancer detection
by improving the image resolution to about 3 mm or better. The use
of smaller detector elements also improves the resolution. It is
also possible to rotate the planar detectors around the breast to
obtain fully sampled datasets to allow tomographic reconstruction.
 By using a LSO scintillator, a novel lightguide 12 coupling
method, and an efficient way to construct the scintillator array,
the PET modular detector of the present invention has a measured
intrinsic spatial resolution (full-width at half maximum) of about
1.8-2.6 mm, typically about 2.26 mm, an average energy resolution
of about 17-24%, typically about 19.5% at 511 keV and a coincidence
timing resolution of about 2.4 ns. The detector efficiency was about
53% for 511 keV gamma rays, using an energy threshold set slightly
above the electronic noise. These measurements equal or exceed those
obtained from conventional whole-body PET detector designs. Over
95% of the 4050 crystals in the system of the present invention
can be identified in flood histograms of the detector plates 22.
The coincidence timing resolution for the entire system is 8.1 ns.
 I. Detector Design
A. Photomultiplier Tube
 The PS-PMT 8 chosen for the detector module 20 was the Hamamatsu
R5900-C8 (manufactured by Hamamatsu Photonics K.K., Japan) which
had a 4+4 cross plate anode arrangement and 11 dynode stages with
an approximate gain of 6.times.10.sup.6 at -800V, shown in FIG.
2. The photocathode's maximum response was at about 420 nm, which
corresponded well to the light emission spectrum of LSO. The physical
surface area of the PS-PMT 8 was about 3.times.3 cm.sup.2, and the
active photocathode area of the tube was about 2.2.times.2.2 cm.sup.2.
Because there was a discrepancy between the active area of the tube
and the total surface area size, a one-to one coupling of the scintillation
crystals 4 to the active area resulted in a large dead space 16,
equal to about 1.6 cm between two adjacent PS-PMTs 8, seen in FIG.
1a. In order to allow close packing of the detector modules 20 without
any gaps (FIG. 1b), a scintillator crystal array 4 that matched
the outer surface area of the PS-PMT 8 was used. The coupling of
the array to the active area of the PS-PMT 8 is described in section
B. Scintillator Array
 Conventionally, scintillator arrays have been formed from
polished crystals that are either hand-wrapped in reflective PTFE
tape and bundled together, or alternatively, glued together using
a white pigment, such as BaSO.sub.4 or TiO.sub.2 mixed with an epoxy
or RTV. The disadvantage of the approach of wrapping in reflective
PTFE tape is that it is extremely labor intensive and difficult
to control. The disadvantage of the latter approach, bonding the
reflective pigment onto the surfaces of the crystal 4, is that light
output is reduced substantially. Also, the mechanical polishing
of large numbers of small crystals is also an expensive process.
 The arrays of the present invention were formed using a
different approach, designed to reduce the cost and labor involved,
while maintaining high light output. The dimensions of the LSO crystals
4 were about 3.times.3.times.20 mm.sup.3. Slabs of raw LSO were
initially cut to about 3.times.3.times.20 mm.sup.3 in size and then
chemically polished, rather than mechanically polished using abrasives.
The chemical polishing technique required bathing the crystals in
phosphoric acid, the concentration of which was 85% by volume, for
about 16 minutes at 190.degree. C. [R. Slates, et al., "Chemical
Polishing of LSO Crystals to Increase Light Output," IEEE Trans.
Nucl. Sci., 47, 1018-1023 (2000); J. S. Huber, et al., "Geometry
and Surface Treatment Dependence of the Light Collection from LSO
Crystals," Nucl. Inst. Meth., 437, 374-380 (1999); and K. Kurashige,
et al., "Surface Polishing of GSO Scintillator Using Chemical
Process," IEEE Trans. Nucl. Sci., 45, 522-524 (1998)]. The
chemical polishing resulted in equivalent or increased light output
compared with mechanical polishing.
 The scintillator array of the present invention included
a 9.times.9 matrix of individually cut LSO crystals 4. To form the
array, a plastic grid 18 was used in order to hold the chemically
polished crystals in place. The grid 18, shown in FIG. 3, consisted
of a matrix of 9.times.9 square holes of a size about 3.times.3
mm.sup.2 with a wall thickness of about 0.3 mm (the gap between
the crystals). The height of the grid 18 was about 5 mm. The grid
18 was fabricated using a 3-D stereolithography system (manufactured
by Stratasys Inc., Eden Prairie, U.S.A.) which used a very fine
extrusion process to build multi-layered objects. Each crystal 4
was encapsulated in white reflective material on the five sides
not coupled to the PS-PMT 8, to enhance the light output from the
side or end from which the scintillation light crystal 4 was coupled
into the PS-PMT. The reflective material was BaSO.sub.4 powder and
methanol in a 1:1 mixture by weight. The thickness of the reflective
material was 300 .mu.m which resulted in overall array dimensions
of about 3.times.3 cm.sup.2 that matched the physical dimension
of the PS-PMT, which enabled detector modules 20 to be tiled together
without gaps. Thus, once the crystals 4 were placed in the grid
18, the 300 .mu.m gap between the crystals 4 was filled with a slurry
of reflective material, BaSO.sub.4 powder and methanol in a 1:1
mixture by weight. BaSO.sub.4 has an extremely high reflectivity
in about the 400-500 nm wavelength range. [W. Budde, "Standards
of Reflectance," J. Opt. Soc. Am., 50, 217-220 (1960)]. The
crystal array 4 was left overnight, during which time the methanol
evaporated and left a uniform coating of BaSO.sub.4 on the crystals.
The outer four sides of the crystal array 4 were then wrapped in
PTFE tape. The top of the array was also covered with PTFE tape
or powder to provide high reflectance.
C. Crystal/PMT Coupling Arrangements
 Conventional crystal/PMT arrangements involve the placement
of the outer edge of a PMT adjacent to and aligned with the outer
edge of an array of scintillation crystals. By constructing a scintillator
array that matched the physical area of the PMT in the present invention,
the dimensions of the crystal array 4 now exceeded the active area
of the PMT, which read out the crystal array 4. Thus, it was necessary
to minify the light distribution from the crystal array so that
it could be read by the PMT. This was accomplished by refocusing
or tapering of the light from the about 3.times.3 cm.sup.2 surface
area down to about a 2.times.2 cm.sup.2 surface area using a lightguide
12, while still maintaining the spatial coherence of the light emitted
by the individual crystals.
 High efficiency of light transmission through the lightguide
12 is of vital importance to preserve energy and timing resolution
in the detector. These directly impact the ability of the PET system
to reject scattered events and to reduce the occurrence of random
coincidence events. A further constraint to overcome was that the
surface area of the lightguide 12 should be no larger than 3.times.3
cm.sup.2, so that multiple individual detector modules 20 can be
tiled together into a larger detector array. In order to determine
the optimal configuration for coupling the scintillation crystals
to the PMT face, five different arrangements were tested.
 Tapered LSO crystals were directly coupled to the PMT face.
Nine rectangular crystals of a size about 3.times.3.times.20 mm.sup.3
were taken and cut with the aid of a diamond saw into a tapered
form to match the active area of the PMT (FIG. 4a). Each individual
crystal had been chemically polished and wrapped in polytetrafluoroethylene
(PTFE) tape (such as TEFLON, manufactured by Dupont, U.S.A). The
unwrapped side or face was coupled to the PMT with the aid of optical
grease (index of refraction of about 1.433). This arrangement served
as a good reference with which to measure the degradation introduced
by the coupling arrangements described below.
Lightguide 12 Arrangements
 A lightguide 12 constructed from B-270 glass (Precision
Glass and Optics, Santa Ana, Calif., U.S.A.) cut into individually
tapered lightguides 12 was used (FIG. 4b). This particular glass
is available in large sizes for machining and has good transmission
in the blue part of the spectrum where LSO emits most of its light.
The dimensions of the lightguide 12 were cut so that one side would
match the scintillation crystal array (about 3 cm) and the other
side would match the active face of the PMT (about 2 cm). Both interfaces,
between the lightguide 12 and PMT 8 and between the lightguide 12
and the crystal array 4, were coupled with optical grease.
 A conventional optical lens (Edmund Scientific, Barrington,
N.J.) was utilized. A central square section of size about 30 cm.sup.2
was cut out of a plano-concave (PCV) lens of diameter about 50.0
mm and having an effective focal length of about -100.0 mm (FIG.
4c). The curved surface of the lens was placed directly on the PMT.
The crystal array was coupled to the planar side of the lens using
 A set of nine, double-clad optical fibers (Kuraray Corp.,
Japan) about 2 mm diameter and about 5.3 cm long were utilized (FIG.
4d). The indices of refraction for this fiber were about 1.59 (core),
about 1.49 (inner cladding) and about 1.42 (outer cladding), giving
a numerical aperture of about 0.72. This configuration was similar
to that used in a detector previously developed for small animal
imaging. [S . R. Cherry, et al., "Optical fiber readout of
scintillator arrays using a multi-channel PMT: A high resolution
PET detector for animal imaging," IEEE Trans. Nucl. Sci., 43,
1932-1937 (1996)]. All interfaces, as in Example 2, were coupled
with the aid of optical grease.
 A tapered fiber bundle (TaperVision Inc., Pomfret, Conn.),
as shown in FIG. 4e, was used. The optical fiber bundle was a coherent
bundle composed of many thousands of micron diameter glass fibers
fused together. To allow detector modules 20 to be tiled together,
the larger end of the fiber taper, FIG. 4e, was cut into about a
3.times.3 cm square to match the physical dimensions of the PMT
and the LSO array, as shown in FIG. 4f. The taper was made from
thousands of 10 micron diameter glass fibers that were vacuum fused
and then drawn out to form the tapered end while maintaining spatial
coherence of the light. [E. Peli, et al., "Fiber-optic reading
magnifiers for the visually impaired," J. Opt. Soc. Amer. A,
12, 2274-2285 (1995)]. The light transmission at about 420 nm, as
measured with a spectrophotometer, was about 30% with a numerical
aperture of about 0.98. The larger diameter of the uncut taper,
FIG. 4e, was about 5.3 cm and the smaller diameter measured about
2.9 cm, therefore the effective minification factor was about 1.8.
RESULTS OF COUPLING ARRANGEMENTS EXAMPLES
 Each of the five coupling methods described above was evaluated.
Energy resolution, light collection efficiency, and the identification
of the individual elements in the scintillator array were measured.
 To measure the energy resolution and light collection efficiency,
a single about 3.times.3.times.20 mm.sup.3 LSO crystal was used
in conjunction with the different coupling arrangements. A .sup.22Na
point source 14 was placed about 10 cm from the proximal face of
the LSO crystal and an energy spectrum was acquired. From the energy
spectrum, the full-width at half maximum (FWHM) of the 511 keV photopeak
was measured to provide the energy resolution. The relative position
of the photopeak, with respect to that obtained with the LSO directly
coupled to the PMT, was used to measure the light collection efficiency.
A schematic diagram of this measurement is shown in FIG. 5a.
 To assess the identification of individual elements in the
scintillator array, a one dimensional array of nine about 3.times.3.times.20
mm.sup.3 in size LSO crystals was coupled to the PS-PMT using the
coupling arrangements described above and compared with direct coupling
of tapered LSO crystals to the PS-PMT (FIG. 5b). The array was flood
irradiated with a .sup.22Na point source 14. For each detected event,
the four outputs of the PS-PMT corresponding to the direction of
the array on the PS-PMT face were digitized and the centroid position
calculated. The histogram of the centroid positions for a large
number of events was examined to see if the individual crystals
can be separated. A profile through the histogram provided a more
quantitative assessment of crystal identification by measuring the
peak to valley ratio. Only the central three crystals were used
to measure the average peak-to-valley ratio, as the nine crystals
could not all be identified with every coupling scheme.
 To minimize variability between measurements, the same LSO
crystals, PMT and electronics were used throughout all of the experiments.
The high voltage bias to the PMT, constant-fraction discriminator
setting, and timing gate width were all kept constant. The measurements
of the various figures-of-merit (energy resolution, light collection
and flood histogram peak-to-valley ratio) for the different experimental
setups are presented in Table I.
1TABLE I Summary results from the various lightguide configuration
experiments Number of Energy Light Average Crystals Resolution Collection
Peak-to- Clearly Coupler (FWHM %) Efficiency (%) Valley Ratio Resolved
Direct LSO* 13.0 100.0 10.0 9 Lightguide* 19.9 40.6 2.5 8 PCV Lens
27.2 28.0 2.5 7 Fiber* 35.0 12.6 6.0 6 Fiber taper 19.5 27.0 7.5
9 *Energy resolution and light collection efficiency were measured
with single lightguide elements.
 Compared to direct coupling, the best combination of energy
resolution and crystal identification was obtained with the tapered
fiber bundle, which was also able to clearly resolve all the individual
crystals in the flood histogram. The measured energy resolution
was about 19.5% with a light collection efficiency of about 27%
compared with direct coupling. The tapered fiber bundle was a relatively
inexpensive solution and was very easy to handle. Individual light
guides had better light collection efficiency, but the energy resolution
was not significantly better than the tapered fiber bundle and the
crystals were poorly resolved, as indicated by the inferior peak-to-valley
E. Construction of the Detector Module
 Based on the results of the coupling examples, two complete
detector modules 20 were constructed. The scintillator array was
formed as described further, in section II.A, and coupled through
the optical fiber taper, shown in FIG. 4e, to the R5900-C8 PS-PMT.
A completed detector module 20 is shown in FIG. 6. The fully assembled
detector module 20, including the PMT socket containing the dynode
resistor chain bias network, was about 3 cm long, about 3 cm wide,
and about 9.75 cm long.
II. Detector Module Characterization
A. Flood Source Histogram
 One of the constructed detector modules 20 was uniformly
irradiated with a .sup.68Ge point source. The position signals from
the PS-PMT (4 X and 4 Y outputs) were multiplexed to give 2 X and
2 Y outputs with the use of a simple resistive chain readout configuration.
The four position signals were integrated for about 0.2 .mu.sec
and fed into an analog-to-digital conversion (ADC) board (Model
PCI-416L manufactured by Datel Inc., Mansfield, U.S.A.) located
inside the data acquisition computer (Model Optiplex GX1P manufactured
by Dell Computers, Inc., Round Rock, U.S.A.). The X and Y coordinates
were calculated for each detected event according to Anger logic
and histogrammed to produce a 2-D position map. [.sup.30H. Anger,
"Scintillation cameras," Rev. Sci. Instr., 29, 27-33 (1958)].
The lower energy threshold was set to about 100 keV with the aid
of the constant fraction discriminator and no upper energy threshold
 An image of the flood histogram from one detector module
20 is shown in FIG. 7. All 81 crystals from the 9.times.9 LSO array
were clearly visible. An average peak-to-valley ratio of 3.5 was
obtained over the central row of 9 crystals. Not all crystals were
uniformly spaced in the flood histogram. This may be a result of
the non-uniform tapering of the optical fiber taper or the non-uniform
packing of the reflectance powder between the crystals. Also, there
were variations in gain, light sharing, and position linearity across
the PMT. However, each crystal was clearly identified and a position
look-up table (LUT) can be easily created from the flood image.
B. Energy Spectra
 Boundaries were drawn on the 2-D position map to define
a look-up table (LUT) which relates position in the 2-D histogram
to the appropriate element in the LSO array. The raw list mode data
was then resorted and a histogram of total pulse amplitudes (sum
of the four position outputs) was generated for each crystal in
the array. These energy spectra were analyzed to determine the FWHM
and the location of the 511 keV photopeak of each crystal, these
two parameters measured the energy resolution and light collection
 Energy spectra showing the worst, best and an average measurement
from the 81 crystal elements are presented in FIG. 8. The average
energy resolution for the entire detector module 20 was about 19.5%.
This compared very favorably with the 20% energy resolution measured
in the detectors used in ECAT EXACT HR+, a clinical whole-body scanner.
[S. R. Cherry, et al., supra].
C. Timing Resolution
 Two detectors were mounted in an aluminum frame and aligned
facing each other in coincidence, a distance of about 15 cm apart
(FIG. 12). A .sup.22Na point source 14 was placed in the center
of the two detectors. For each detected coincidence event, the sum
of the four position signals for each detector was sent to constant
fraction discriminators which generated timing pulses. These two
timing pulses, one for each module, were in turn fed into a calibrated
time-to-amplitude converter (TAC) module. The output from the TAC
was then digitized to produce the timing spectrum.
 The timing spectrum is shown in FIG. 9. The FWHM of the
time response was about 2.4 ns. Typical BGO block detectors, those
used in conventional whole body PET scanners, have a timing resolution
of about 4-6 ns, e.g., the EXACT HR+ has a timing resolution of
5.5 ns. [S. R. Cherry, et al., supra]. Therefore, this new detector
design allows better rejection of random coincidence events, as
the system timing window can be reduced.
D. Coincidence Point Spread Function
 Flood source histograms of both detectors were obtained,
as described above in section II.A, from which the position LUT's
were defined. The detectors were then connected in coincidence,
about 15 cm apart, and list-mode data was acquired by stepping an
about 1 mm diameter .sup.22Na point source 14 between the detectors
in about 0.254 mm steps. For each opposing crystal pair, the counts
were recorded as a function of the point source position. A lower
energy window of about 100 keV was applied. The FWHM of the resulting
distribution for each crystal pair was determined to provide the
intrinsic spatial resolution of the detectors.
 The coincidence point spread function for an entire row
of crystals is shown in FIG. 10. The average FWHM was measured to
be about 2.3 mm, with the worst being about 2.6 mm. The edge crystals
tended to have better intrinsic spatial resolution, most likely
due to reduced inter-crystal scattering from the adjacent crystals,
lowering the probability of mis-positioning events at the edges
and corners of the module.
E. Detector Efficiency
 A measure of the detector efficiency was performed. A .sup.68Ge
point source with known activity was placed about 9.5 cm away from
the face of the detector module 20. The actual photon flux impinging
on the detector face was calculated from the solid angle subtended
by the detector at the source. The constant fraction discriminator
was set to eliminate electronic noise and the number of counts detected
by the module was recorded. The number of counts detected was then
divided by the number of photons impinging on the detector module
20 to obtain the detector efficiency.
 The detector efficiency was calculated to be about 53%.
This result coincided well with calculations for the geometry of
our module. Based on this measurement, a coincidence efficiency
of (0.53).sup.2=0.28, or about 28% is expected, with a wide energy
window. The energy window will ultimately depend on the trade-off
between efficiency and scatter. This efficiency, when combined with
the large solid angle or our proposed system, leads to excellent
III. System Characterization
 Two maxPET detector plates 22 were mounted in an aluminum
frame a distance of about 15 cm apart. Alternatively, the two detector
plates 22 can be mounted on a gantry allowing variable plate separation,
detector plate rotation, and angular motion. The detector plates
22 were connected through NIM pulse shaping electronics to a PC-based
data acquisition system running LabView (National Instruments, Austin,
Tex.), containing a 16-channel PCI-based ADC board (PCI-416L,Datel
Inc., Mansfield, Mass.).
A. Readout Scheme
 The design for the maxPET readout electronics involves using
commercially available modules and boards. In order to reduce the
number of channels to be digitized, the detector plates 22 utilized
a modified resistor chain readout scheme based on segmentation of
the 5.times.5 array. Since each PMT produced 8 anode outputs (4X
and 4Y), a total of 200 channels (8 anodes.times.25 tubes) per plate
needed to be digitized if each tube was handled individually. To
digitize every channel is an impractical and costly approach. Therefore,
a multiplexing scheme to reduce the number of channels was again
implemented. No more than 4 PMTs were used per readout segment in
an attempt to offset multiplexing losses while still significantly
reducing the number of readout channels.
 The multiplexing scheme is seen in FIG. 13. In this approach,
the X anodes from all the PMTs along a row were connected together
along separate bus lines and then fed into a resistor chain. Similarly,
the Y anodes along a column were connected together along separate
bus lines and then fed into another resistor chain. There were four
summing junctions connected to each resistor chain, producing a
total of 8 outputs to be digitized per plate. Each resistor chain
utilized 100 ohm resistors in between the anode outputs and 750
ohm resistors in the operational amplifier feedback circuitry. There
were two operational amplifier stages. The first stage was a current
feedback amplifier and the second stage was a unity gain inverter.
The inverter stage produced a negative polarity pulse which is required
by the pre-amp input. Rather than using two separate operational
amplifiers, a single, surface mount, dual OpAmp (Model AD8015, Analog
Devices, Norwood, Mass.) was used.
 The resistor chain readout scheme effectively segmented
the entire 5.times.5 array into a total of 9 sectors. This scheme
allowed each sector to utilize the full dynamic range of the digitizer
thereby allowing better crystal identification. A slightly modified
Anger logic algorithm was used to position the event using this
readout scheme. Since all 8 channels in the X and Y directions were
digitized, the algorithm first determines which sector registered
the largest signal output. This is accomplished by summing the two
"end" channels per sector and comparing the sum to the
other two sectors in that direction. Once the sector which produced
the highest signal output for a particular event is determined,
conventional Anger logic was only applied to the channels coming
from that sector.
B. Flood Histograms
 A flood source histogram image was obtained separately for
each plate. (FIGS. 14a and b). A 2 cm diameter .sup.68Ge disc source
was used to irradiate each plate independently. The 8 position signals
(4X and 4Y) coming from the readout board were integrated for about
0.2 microseconds and fed into the analog-to-digital conversion (ADC)
board located inside the data acquisition computer. The lower energy
threshold was set to about 100 keV with the aid of the constant
fraction discriminator and no upper threshold was applied. After
collecting the list mode data from an experiment, the X and Y coordinates
were calculated on an event by event basis from the 8 digitized
position signals using the modified Anger logic scheme described
previously. The X and Y coordinates were then histogrammed to create
a 2-D position map. This process was performed for each plate independently.
 A flood image of one of the detector plates 22 is shown
in FIG. 14a. More than 95% of the 2025 crystals were identified
on each detector plate. The variations in the intensity of the different
crystal elements was due to differences in gain, coupling variations,
differences in the LSO light output and variations in the amount
of reflector in between the crystals. The flood images appeared
to be segmented into 9 separate sectors because of the multiplexed
sector readout scheme. The uniform background in the flood image
was probably a combination of noise due to the low threshold and
contributions from inter-crystal scatter.
C. Position Look-Up Tables and Energy Resolution
 Boundaries were drawn on the 2-D histogram maps obtained
from the examples above and were used to generate look-up-tables
(LUTs) for each sector for both the detector plates 22. The LUT
was then used to relate the position in the 2-D histogram to the
appropriate crystal element in the LSO array. In order to generate
an energy histogram plot for each element, the list mode data were
then resorted based on the total pulse amplitudes (sum of the 4
position signals) for a particular crystal. These energy spectra
were then analyzed to determine the FWHM which provided a measure
of the energy resolution.
 The average energy resolution for each detector plate was
calculated as the average of the energy resolutions for the individual
crystal elements for which the photopeak could be identified. The
average energy resolution for detector plate 1 was about 22.9%,
with a range of 14-39%. The average energy resolution for detector
plate 2 was about 20.4%, with a range of about 12-28%. Greater than
90% of the crystals had clear photopeaks from which the energy resolution
was determined and included in the averages quoted above.
D. Timing Resolution
 The two detector plates 22 were connected in coincidence.
A .sup.22Na point source 14 was placed at the center of the two
detector plates 22. The singles rates on each detector were kept
low by using a weak source to minimize random events. For each detected
coincidence event, the sum of the 8 position signals from each detector
plate was sent to constant fraction discriminators which generated
the timing pulses. These two timing pulses, one for each detector
plate, were in turn fed into a calibrated time-to-amplitude converter
(TAC) module. The output from the TAC was then digitized to produce
the timing spectrum. A background measurement, acquired for the
same amount of time, was taken without the source and subtracted
to remove coincidences due to LSO background.
 The timing spectrum measured with a positron source placed
at the center of the two detector plates 22 is shown in FIG. 15.
The FWHM of the time response was 8.1 ns. This was a different response
than that measured with the two single detector modules 20 alone,
which showed a 2.4 ns timing resolution. This result may be due
to lack of time alignment of the 50 detector modules as well as
variable, position dependent delays introduced in the readout board
itself. Thus, a timing window of 16-20 ns could be beneficial.
E. Phantom Image
 An acrylic line source phantom of size 4.78.times.1.08.times.13
cm consisting of 8 drilled channels was filled with FDG and imaged.
Each square channel measured 1.08.times.1.08.times.130 mm producing
a fillable volume of 0.15 ml. The channels were spaced with a variable
center-to-center distance as follows: 10 mm, 8.5 mm, 7 mm, 5.5 mm,
4 mm, 2.5 mm and 1 mm. The total amount of activity in the phantom
was approximately 50 .mu.Ci. The phantom was imaged for approximately
11/2 minutes for a total of 1.1 million coincidence counts at an
average count rate of 10,500 counts per second. No corrections were
made for random coincidences or individual detector efficiencies.
An energy threshold of about 250 keV was applied to the data.
 A line bar phantom was scanned with the system comprising
the two 15.times.15 cm.sup.2 planar scintillation detector plates
22, and the images were reconstructed using focal plane tomography.
The images from the line bar phantom experiments are shown in FIGS.
16a and b and represent the in-focus plane. FIG. 16a was reconstructed
using a full angle of acceptance of +/-45.degree. where each detector
element may be in coincidence with every other detector element
in the opposing plate. The image contains 1.1 million events. FIG.
16b was reconstructed using a half angle of acceptance of +/-22.5.degree.
and contained 550,000 events. Each image was scaled to the maximum
value in the respective image. No corrections were applied to either
F. Projection Image Resolution
 Images were generated using a simple focal plane tomography
algorithm (simple backprojection). Two sets of images were reconstructed,
one using line of responses (LORs) corresponding to the full angle
of acceptance (+/-45.degree.) and the other using LOR's from one
half of the full acceptance angle (+/-22.5.degree.). Profiles through
the line bar images were taken and analyzed to assess the point
at which two adjacent channels were no longer distinguishable as
two separate lines.
 Qualitatively, one is able to visually separate the first
three lines in the FIG. 16a and the first four lines in FIG. 16b.
Profiles taken through the two images are shown in FIGS. 17a and
b. In the profile obtained from the FIG. 16a (full angle of acceptance),
six clear peaks may be identified. The distance between the two
closest peaks represents a physical distance of 4 mm. In the profile
obtained from FIG. 16b (half angle of acceptance), five peaks may
be clearly identified. The distance between the two closest peaks
in this profile represents 5.5 mm. The differences in the images
in this experiment are minimal as seen in the profiles.