++++++ ,+ _ +_'-_,.,+
'+:++,+£,,,_m IUIM
-X,_J_+ +_+ @ .,_\\_ Associationforlnf°rmationandult?:g:Mana" ement
+7////,+:+_+++++-,=, s+r++:+++7:++_++++°++°+,+.+"_"'"+++-+++
Centimeter
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 mm
_'_''i'''_'_''i'_''_'''_i'_''_`'_i_''_'''_i''_'_'_'i'_''_''1''_'_''''i_'''_'_''i'_'_''''i_''_'''_''_'_'_''i''_'_'_''i_''_'_'_'''_''''i'_''_
i,,,,i,,,,i,,,,i,,,,i,,,,i,,,,i,,,,i,,,,i,,,,i,,,,I,,,, I,,,,i
1 2 3 4 5
iiiii+
Inches 1.0 ,= _ II1_
,._lllll_
+ iiiii_
iliil--'.."_
_,=22
illli_
IlllINillligIIIII'_
+++.o+ ++++m
+ + ,<,,+_'+
+_, +.-.N + ..,
:' "_ #, MglNUFglCTUREDTO glTTM STglNDI::IRDS _%-_" _++_x??+
BY RPPLIEO IMRGE, INC. __
Presented at the 1994 Symposium on Radiation Measurements and Applications LBL-35695
PET Detector Modules Based on Novel Detector Technologiest
William W. Moses, Stephen E. Derenzo, and Thomas F. Budinger
Lawrence Berkeley Laboratory, University of California, Berkeley, CA 94720
ABSTRACT
A successful PET detector module must identify 511 keV photons with: high efficiency
• (>85%), high spatial resolution (<5 mm fwhm), low cost (<$600 / in2), low dead time (<4 _ts
in2), good timing resolution (<5 ns fwhm for conventional PET, <200 ps fwhm for time of
flight), and good energy resolution (<100 keV fwhm), where these requirements are listed in
decreasing order of importance. The "high efficiency" requirement also implies that the detector
modules must pack together without inactive gaps. Several novel and emerging radiation detector
technologies could improve the performance of PET detectors. Avalanche photodiodes, PIN
photodiodes, metal channel dynode photomultiplier tubes, and new scintillators all have the poten-
tial to improve PET detectors significantly.
INTRODUCTION
Positron emission tomography (PET) is a nuclear medical imaging technique whereby the
patient is injected with (or inhales) a biologically active tracer compound (i.e. a drug) that is labeled
with a positron emitting isotope. The tracer accumulates in the patient, carrying the radioisotope
with it. When the radioisotope decays, the emitted positron annihilates with an electron to form
back-to-back 511 keV photons, which penetrate the patient and enter the PET detector ring.
Individual positron annihilations are identified by simultaneous detection of these photons, and the
parent radioisotope known to lie along the line connecting the two detected photons (known as a
chord). The mathematical technique of computed tomography then uses these detected chords to
reconstruct the spatial distribution of the radioisotope, and therefore the drug concentration in the
patient. The time dependence of the drug uptake also contains metabolic information. This tech-
nique has been successfully applied to neurological, cardiovascular, and oncological medicine [1].
An individual PET detector module must identify the 511 keV photons exiting the patient,
and for each photon detected, provide a timing pulse and a spatial location to the tomograph coinci-
dence electronics, which then pairs detected photons based on their arrival time and assigns them to
a chord based on the two ir.teraction positions. The requirements for the detector module are, in
approximate order of decreasing importance, are as follows. (1) High detection etficiency (>85%
per 511 keV photon), as the chord efficiency is the square of this individual photon efficiency and
PET scans are generally "starved" for statistics. This requirement also implies that the detector
modules must pack together without inactive gaps. (2) High spatial resolution (<5 mm fwhm), as
the detector spatial resolution is the main factor influencing the spatial resolution in the recon-
structed image. (3) Low cost (parts cost <$600 / in2 of "front" surface area), as virtually all PET
cameras are commercially manufactured. (4) Low dead time (<4 ILtsin2), for the high counting
. rates found during transmission scans and with short half-lived radioisotopes. The figure of merit
given for dead time is the product of the detector dead time and the front surface area of the portion
of the detector that is dead. (5) Good timing resolution (<5 ns fwhm), in order to reduce the acci-
dental coincidence rate, which is proportional to the square of the single photon rate. If very good
" timing resolution is achieved (<200 ps fwhm), the arrival time difference can be used to localize
the radioisotope position along the length of the chord. (6) Good energy resolution (<100 keV
fwhm), in order to reject photons that have Compton scattered in the patient.
When evaluating different materials for stopping the 511 keV photons, one must consider
tThis work was supported in part by the U.S. Department of Energy under contract No. DE-AC03-76SF00098, and
in part by Public Health Service Grant Nos. p0! HL25840, R01 CA48002, and R01 NS29655.
M/STER
DISTRIBUTION OF THIS DOCUMENT IS UNLIMITED
<
q _ ..... i II , , i i
Presented at the 1994 Symposium on Radiation Measurements and Applications LBL-35695
both the attenuation length and the type of interactions the 511 keV photons have in the material.
Photoelectric interactions are greatly preferred over Compton scatter, as a 511 keV photon that
interacts via Compton scatter deposits energy in two (or more) locations in the detector ring,
frequently separated by 1 cm or greater, and thus reduces the spatial resolution of the detector
module. A convenient figure of merit for evaluating materials is the photoelectric attenuation
length, which is the attenuation length.... divided by the photoelectric fraction (i..e _p / (_p+Oc),
where _.p and _c are the cross sections for photoelectric effect interactions and Compton scatter ,
interactions at 511 keV) [2]. Low values of this length are desired, as this implies both short
attenuation length and high photoelectric fraction.
EXISTING DETECTOR MODULES
Current commercial PET cameras generally use detector modules similar to that shown in
Figure 1 [3]. The 511 keV photons interact in the BGO (bismuth germanate, or Bi4Ge3012)
scintillator crystal, and the resulting scintillation light observed by four photomultiplier tubes
(PMTs). BGO is usually the scintillator of choice, as its 2.6 cm photoelectric attenuation length
(1.1 cm attenuation length, 43% photoelectric fraction) is lower than any other commonly avail-
able scintillator. The 30 mm depth of the BGO crystal is nearly 3 attenuation lengths, ensuring
high detection efficiency. Saw cuts in the BGO define "individual" crystal elements by controlling
the light distribution among the four PMTs, and Anger logic (i.e. analog ratios among the four
PMT signals) is used to determine which of the "individual" crystals the interaction occurred in.
The sum of the four PMT signals is used to form a timing pulse (with 3 ns fwhm accuracy) and a
measurement of the photon energy (with 100 keV accuracy). The size of the "individual" crystal
elements determines the position resolution of the detector module, but a limited number of crystals
(typically <64) can be accurately decoded due to the limited light output of BGO. The entire
module is "dead" for approximately 1 Its after a 511 keV photon interaction while the BGO emits
its scintillation light (its decay time is 300 ns), as interaction in any other portion of the module
during this time would confuse the Anger logic. Assuming a price for BGO of $20 per cc and a
price of $200 per PMT (independent of size), the parts cost for this detector module is $600/in 2.
Four 1" Square _ _.._ Fig. 1: "Conventional" PET Detector
Photomultiplier Tubes / /_ Module.
?_ The "conventional" block detector module consists
of a 2" x 2" x 30 mm BGO crystal that is sawed
into 64 segments, each 6 mm x 6 mm x 30 mm
,, deep. When a 511 keV photon interacts in any of
!__ i_ S the segments, the scintillation light is distributed
across the back face of the BGO crystal, where it is
. simultaneously measured by four 1" square PMTs.
I H _ _ _ The sum of the four output signals is used to derive
both a timing signal and a signal proportional to
_ ", _'i__,,,4 _ BGO Crystal Block, the energy deposit. Anger logic (i.e. the ratio of the
_ " _'_.,,,,,_ _s4'_eg_e_t_' light observed in each of the four PMTs)is then
"2" _.*"_''_"'f,_0 mrn each6 mmsquare used to determine the segment of interaction.
To evaluate competitive designs, we set the performance goals of improving the spatial reso-
lution by a factor of two (i.e. reducing the crystal size from 6 mm square to 3 mm square) and
decreasing the dead time figure of merit by a factor of four. Another important goal, albeit difficult
to achieve, is to make the spatial resolution uniform over the field of view by eliminating an artifact
known as radial elongation. Figure 2 shows the origin of this artifact, which is due to 511 keV
photons impinging on the detector crystals at an oblique angle. Because of the 1.1 cm attenuation
length of BGO, these photons can penetrate into adjacent crystals before they interact and are
detected, which causes mis-positioning errors (i.e. events are assigned to chords that do not pass
through the source). This spatial resolution degradation increases for objects placed further away
from the center of the tomograph ring. If the detector module were able to measure the position (in
'i
'.i ',,. r '"
Presented at the 1994 Symposium on Radiation Measurements and Applications LBL-35695
the radial direction) of the 511 keV photon interaction in the BGO with sufficient accuracy (5 mm
fwhm), then the interactions would be assigned to the proper chord and this artifact could be elimi-
nated [4]. No tomograph, commercial or research, yet has the ability to make this measurement,
which is commonly known as depth of interaction measurement.
i
, Fig. 2: Radial Elongation in PET.
511 keV photons impinging on the detector ring at
an oblique angle can penetrate into adjacent crystals
before they interact and are detected. This leads to
• _ mis-positioning errors that increase in severity as
-_1.,._---_'__anII ;rol_ c_i the source moves further away from the center of
the tomograph ring. Due to the detector geometry,
the radial projection (relative to the tomograph ring)
is affected while the tangential projection is not
affected, hence it is known as radial elongation. If
the position of interaction (within the crystal) is
measured with sufficient accuracy (5 mm fwhm),
/I
Radial ]l event pairs can be assigned to the proper chord and
Projection ,/I this artifact removed.
These design goals could be accomplished (without depth of interaction measurement) using
a convention_.l detector module with the linear dimensions on the front surface reduced by a factor
of two (i.e. having a 1" square by 30 mm deep BGO crystal cut into 64 3 mm square "individual"
elements, read out by four 0.5" square PMTs). However, the parts cost for this detector module is
$1200/in 2 due to the increased number of PMTs needed to cover a square inch.
NOVEL PHOTODETECTORS
Numerous novel and emerging photodetector technologies could be incorporated into PET
detectors in ways that could also meet these goals. One promising category of photodetector is
devices that are small (the size of an individual 3 mm square crystal) but have sufficiently high
gain bandwidth product to provide an accurate timing signal and energy measurement. Examples of
such devices are multi-anode PMTs [5], metal channel dynode PMTs [6], avalanche photodiode
(APD) arrays [7], segmented vacuum avalanche phototubes [8], and VLPCs (visible light photon
counters) [9]. These devices would be incorporated into a PET detector module as shown in
Figure 3. The individual crystals are now decoupled, both optically and electronically, and so the
position of the crystal of interaction is determined merely by identifying the photodetector that
fires. The depth of interaction is measured by coating the crystals with a "lossy" reflector, so the
ratio of the light observed in the "front" photodetector (i.e. the one closest to the patient) and the
"back" photodetector depends on the position of the 511 keV photon interaction.
Arrays of 64 Fig. 3: High Gain-Bandwidth
Photodetectors Photodetector Module.
With high gain-bandwidth photodetectors, a module
would consist of 64 optically isolated BGO
" crystals, each 3 mm x 3 mm x 30 mm deep and
coated with a "lossy" reflector. When a 511 keV
photon interacts in any of the elements, the scintil-
, 1" lation light is detected by photodetectors at either
end of the crystal. The sum of the two output
BGO Crystals signals is used to derive a timing signal and a
3 mm square signal proportional to the energy deposit, and the
• ratio used to determine the depth of interaction.
While this type of detector module has the desired spatial resolution improvement due to the
smaller (3 mm square) crystals, its main advantage is that the crystals are completely decoupled,
and so a much smaller area is inactivated for the 1 ILtSthat the BGO is emitting light. This