Conference PaperPDF Available

Depth-encoding PET detector module with improved spatial sampling

Authors:

Abstract

Detector modules in small ring diameter PET scanners must possess depth-of-interaction (DOI) encoding, increased spatial sampling, high sensitivity and the ability to handle high photon input rates without excessive pulse pileup or random coincidences. We created such a module by optically coupling an entrance array of individual LGSO crystals to an exit array of individual GSO (and other) crystals that was, in turn, optically and directly coupled to a miniature PSPMT. DOI was determined for each event by delayed charge integration (DCI), a technique that exploits differences in light decay time between GSO and LGSO. Spatial sampling in 3D was increased by introducing a half crystal pitch spatial offset between the entrance and exit arrays in both the X and Y directions. Position detection accuracy in both the LGSO and GSO layers, and the accuracy of DOI assignment of events to either layer was high. These results suggest that this combination of scintillators and acquisition/processing methods may be particularly useful in the design of high performance, small ring diameter PET scanners for small animal imaging
© 1998 IEEE. Personal use of this material is permitted. Permission from IEEE must be
obtained for all other uses, in any current or future media, including reprinting/republishing
this material for advertising or promotional purposes, creating new collective works, for
resale or redistribution to servers or lists, or reuse of any copyrighted component of this
work in other works.
A
Depth-Encoding
PET
Detector Module with Improved Spatial Sampling
Lox)
Lox)
Lox)
Lox)
J.J.
Vaquero,
J.
Seidel,
S.
Siegel',
M.V.
Green
Nuclear Medicine Department
National Institutes of Health, Bethesda MD 20892
Lox)
Lox)
Lox)
Lo60
Lox)
Lo60
Lox)
uiso
LDDO
Lox) Lox)
Lox)
Abstract
Detector modules
in
small ring diameter PET scanners
must possess depth-of-interaction (DOI) encoding, increased
spatial sampling, high sensitivity and the ability to handle high
photon input rates without excessive pulse pileup or random
coincidences. We created such a module by optically
coupling an entrance array of individual LGSO crystals to an
exit array of individual
GSO
(and other) crystals that was, in
turn, optically and directly coupled to a miniature PSPMT.
DO1 was determined for each event by delayed charge
integration (DCI), a technique that exploits differences
in
light
decay time between
GSO
and LGSO.
Spatial sampling
in
3D was increased by introducing a half
crystal pitch spatial offset between the entrance and exit arrays
in
both the
X
and
Y
directions. Position detection accuracy in
both the LGSO and
GSO
layers, and the accuracy of DO1
assignment of events to either layer was high.
These results
suggest that this combination of scintillators and
acquisition/processing methods may be particularly useful in
the design of high performance, small ring diameter PET
scanners for small animal imaging.
I. INTRODUCTION
As the diameter of a PET ring scanner shrinks, the need for
depth-of-interaction (DOI) information, finer spatial sampling
and improved count rate performance increases. Numerous
methods have, in fact, been advanced that address one or
more of these issues. We propose here a module that, in
principle, meets all or most of these conditions
simultaneously, while possessing good position detection
accuracy.
First, the
module is comprised of two layers of optically coupled
scintillators that differ from each other in light decay time (for
DO1 determination). Second, all of the crystals
in
both layers
are geometrically identical and possess small entrance cross-
sections (potentially good position detection accuracy). Third,
the entrance layer is spatially offset with respect
to
the exit
layer by half a crystal pitch
in
both spatial directions
(increased spatial sampling in 3D). Fourth, both scintillators
have short light decay times, high densities, good light outputs
and an extended combined length (potentially high count rates
with low pulse pileup, good stopping power, good position
detection accuracy, good sensitivity).
The module combines five different features.
Finally, these layers are directly coupled to a thin window,
miniature position-sensitive photomultiplier tube with
favorable performance characteristics (good timing properties
'
Current address: Concorde Microsystems, 10427 Cogdill
Rd,
Suite
500,
Knoxville, Tennessee 37932
for DO1 discrimination, efficient conversion of scintillation
light to photoelectrons for good position detection accuracy).
In the work described below, we evaluated a test module
with these features for the purpose of determining whether this
module might be suitable as the detection element of a high
resolution, small diameter rodent PET scanner.
11.
MATERIALS
AND
METHODS
Exit
Array
Viewed
From
RI
ar
Figure
1.
Geometrical layout of the crystal arrays. The entrance array
is centered on,
and
coupled to, the exit array
with
a
half
pitch offset
in
both the
X
and
Y
directions.
All crystal surfaces were finished by chemical etching and
all crystals were wrapped on their long sides with a double
layer of Teflon tape. The entrance end of each crystal in the
entrance layer was also covered with Teflon tape.
The entrance layers consisted only of LGSO crystals while
the exit layer consisted of GSO and other scintillator types.
(These other crystal types were included to evaluate different
scintillator combinations and DO1 methods, but here we shall
focus only on the properties of the LGSOIGSO portion
of
the
combined arrays).
Note that the entrance layer is offset by half a crystal pitch
(2.2 md2
=
1.1 mm) with respect to the exit array
in
both the
X
and
Y
directions. The entrance layer was optically coupled
to the exit layer with optical grease and the exit layer coupled
with optical grease to a Hamamatsu R5900-C8 position-
sensitive photomultiplier tube (PSPMT). The entire assembly
was placed
in
a light-tight box.
1
B.
Data AcquisitiodProcessing
For each scintillation event within its field-of-view, the
PSPMT generates
9
signals, one for each of the four X-anode
plates, one for each of the four Y-anode plates and a ninth
signal from the last dynode for triggering and DO1
discrimination (Figure
2).
The dynode signal is first used to
generate a
200
ns-wide trigger signal for the charge-
integrating ADCs (LeCroy FERA). It is further split into two
halves that are also integrated.
R59W
Anode
Sianals
Dvnode
Sianal
tirigger
I
L&oy
Rtlrsarch
43008
FERA
ADC
;
Computer
Figure
2.
Data
acquisition system.
All anode signals and one of the dynode signals are
delayed for 100 ns to ensure that they arrive at the ADCs
immediately after the trigger signal (Full Charge Integration,
Figure
3).
The other dynode signal is not delayed
so
that the
ADC can only capture the charge that is still present after 100
ns (Delayed Charge Integration, Figure
3).
If the full charge
integral is plotted against the delayed charge integral (Figure
4A), two radial “spokes” appear in the resulting diagram, one
for each scintillator.
Events occurring
in
the spoke with the greater slope
correspond to events
in
the fast scintillators LGSOLSO, while
the spoke with the smaller slope corresponds to events in the
slower
GSO
scintillator. If ROIs are defined that encompass
the LGSO and
GSO
regions individually (Figure 4B),
assignment of an event to one or the other of these regions
also identifies the layer of interaction.
Thus, when an event occurs, depth is determined for that
event by delayed charge integration (DCI) and the crystal
of
interaction
in
that layer determined by centroid calculation
using the anode signals. (Subsequent refinement of the DCI
technique has improved the separation between these two
scintillators substantially as shown in Figure 4C).
C.
Experiments to Determine DOI and Event
Positioning Accuracy
The ability of these methods to place a scintillation event
in the correct crystal were evaluated qualitatively by flood-
field illumination of the module from the front. Because other
scintillators were included
in
the exit array, the calculated
pattern of events assigned to each layer can be compared to
the known pattern of crystals
in
both layers.
0
100
200
300
time
(ns)
Delayed Charge Integration
0
100
ZW
30
time
(ns)
Figure
3.
Illustration
of
the Delayed Charge Integration method.
0
DO1 accuracy was determined by scanning a 1.1 mm wide
F-18 slit source along the depth direction of the scintillator
arrays. At each point, the number of events occurring
in
both
scintillation layers was determined.
Similarly, position detection accuracy was determined for
each layer by scanning a highly collimated beam of F-18
photons in
0.25
mm steps down the LGSO column
in
the
entrance array that overlay the two columns
of
GSO
crystals
in the exit array. At each spatial position, the counts occumng
in each crystal in both arrays were determined.
111.
RESULTS
The results of field flood illumination of the module are
shown
in
Figure
5.
If the DO1 selection conditions are not
applied, an image of the field-of-view appears as
in
Figure 5A.
When the DO1 conditions are applied, however, this image
separates into two images, one corresponding to the fast
scintillators LGSOLSO (Figure 5B) and the other to the slow
scintillator
GSO
(Figure 5C). Note that the pattern of
GSO
in
Figure 5B and the combined pattern
of
LGSOLSO
crystals
in
2
the entrance and exit arrays shown in these figures correspond
to the known pattern of crystals in the actual arrays (Figure
1).
DO1 and position detection accuracy are illustrated in
Figures
6
and
7.
In each of these graphs the fraction of events
assigned to each crystal is plotted against the position of the
illuminating source. No additional event rejection conditions
were applied in the creation of the DO1 graph. However,
spatial rejection conditions were applied before creating the
position detection accuracy graph in Figure
7.
These results were obtained by placing small spatial
ROIs
around each crystal center and rejecting all events that
occurred outside these regions. Under these conditions,
position detection accuracy is high
(>85%
in both layers) but
sensitivity
low
(<40%
of all events
are
accepted). When large
ROIs are used, position detection accuracy decreases
(<75%)
but sensitivity is increased
(>85%
of all events
are
accepted).
Figure
4.
Phoswich diagram
(A
and
C)
created by plotting the full
charge integral (vertical axis) against the delayed charge integral
(horizontal axis). Definition of
LUTs
for
DO1 assigment
(B).
Separation is improved in
(C)
compared to
(A)
by reducing the
dynode time constant.
Iv.
DISCUSSION
AND CONCLUSION
A
phoswich/PSPMT detector module comprised of two
fast, high stopping power scintillators with offset entrance and
exit arrays can accurately locate the endpoints of annihilation
events in space, albeit coarsely in the depth dimension. The
spatial offset between entrance and exit arrays, moreover, does
not appear to unduly degrade the accuracy of event positioning
across the field-of-view. In addition, since both
LGSO
and
GSO
have short light decay times, pulse pileup can be
minimized by utilizing short pulse integration times
(200
ns or
less).
The ability to acquire accurate DO1 and position
information at high rates and with relatively fine
3D
spatial
sampling suggests that this module may be useful in
applications where these advantages are particularly
significant, i.e. in small ring diameter PET scanners intended
for small animal imaging.
Figure
5.
Field flood illumination of the detector module ignoring
DO1 information
(A).
This image separates into an
LGSOLSO
image
(B)
and a
GSO
image (C) when the Delayed Charge Integration
method is applied.
Phoswich layer accuracy
distance
from
PMT
(mm)
Figure
6.
Fraction
of
events occurring
in
each phoswich layer when
scanned
by
a
narrow
slit source.
3
A
LGSO
accuracy
GSO
acEuracy
millimeters millirneteis
Figure
7.
Fraction
of
events occumng
in
four
adjacent
LGSO
crystals
(A)
and
in
four
adjacent
GSO
crystals
(B)
when scanned
by
a
narrow
beam
of
5
1
1
keV photons.
V.
ACKNOWLEDGMENTS
J.J.
Vaquero is supported, in part,
by
a grant from CICYT
(Spanish Government).
S.
Siegel was supported
by
a grant
from the National Research Council.
VI.
REFERENCES
[
13
M. Schmand,
L.
Eriksson, M.E. Casey, M.S. Andreaco, C.
Melcher,
K.
Wienhard, G. Flagge, R. Nutt. “Performance
Results of a New DO1 Detector Block for a High
Resolution PET-LSO Research Tomograph HRRT”.
1997
IEEE Nuclear Science Symposium and Medical Imaging
Conference Record.
[2]
J.J.
Vaquero,
J.
Seidel,
S.
Siegel, W.R. Candler, M.V.
Green. “Performance Characteristics of a Compact
Position-Sensitive
LSO
Detector Module”.
IEEE
Trans
Med
hag,
in press.
4
... At present high resolution scintillator arrays are usually constructed from individual elements [8,9,10,11,12]. For high spatial resolution, smaller scintillator elements are needed. A major difficulty here is that scintillator elements are very hard to cut and handle once they become 1 mm or less in crosssection. ...
Conference Paper
Full-text available
We report on the development of a new high efficiency detector for small animal PET. The detector is based on a monolithic block of LSO pixelated using laser ablation technique. The laser processing allows pixelation with very narrow, 70 μm wide, interpixel gaps resulting in a substantially enhanced sensitivity when the detectors are operated in coincidence mode. This work presents the first results of a detector module fabricated using this approach. Preliminary imaging data at 511 keV obtained by coupling the pixelated LSO to a position sensitive photomultiplier tube (PSPMT) and a position sensitive avalanche photodiode (PSAPD) are presented.
Article
Full-text available
Super Bialkali (SBA) photocathode is a new technology that improves the spectral response characteristics of position sensitive PMTs, boosting their quantum efficiency up to 35%. In this experiment, two SBA tubes were introduced into a production line of PET detectors mixed with the regular tubes. The detectors were assembled using the standard factory protocols for detector mounting, calibration and testing. We report an evaluation of the improvement introduced by the SBA photocathode comparing the spatial and energy resolution and the depth-of-interaction (DOI) performance of PET detector modules with DOI capabilities. We conclude that the superior performance of the SBA tube may enable the use of arrays with a larger number of crystals of smaller footprint, thus potentially improving the detector intrinsic spatial resolution without degrading the energy resolution or the phoswich (DOI) discrimination capability.
Article
Full-text available
We assembled a compact detector module comprised of an array of small, individual crystals of lutetium oxyorthosilicate:Ce (LSO) coupled directly to a miniature, metal-can, position-sensitive photomultiplier tube (PSPMT). We exposed this module to sources of 511-keV annihilation radiation and beams of 30- and 140-keV photons and measured spatial linearity; spatial variations in module gain, energy resolution, and event positioning; coincidence timing; the accuracy and sensitivity of identifying the crystal-of-first-interaction at 511 keV; and the effects of intercrystal scatter and LSO background radioactivity. The results suggest that this scintillator/phototube combination should be highly effective in the coincidence mode and can be used, with some limitations, to image relatively low-energy single photon emitters. Photons that are completely absorbed on their first interaction at 511 keV are positioned by the module at the center of a crystal. Intercrystal scatter events, even those that lead to total absorption of the incident photon, are placed by the module in a regular "connect-the-dot" pattern that joins crystal centers. As a result, the accuracy of event positioning can be made to exceed 90%, though at significantly reduced sensitivity, by retaining only events that occur within small regions-of-interest around each crystal center and rejecting events that occur outside these regions in the connect-the-dot pattern.
Article
To improve the spatial resolution and uniformity in modern high resolution brain PET systems over the entire field of view (FOV), it is necessary to archive the depth of interaction (DOI) information and correct for spatial resolution degradation. In this work we present the performance results of a high resolution LSO/GSO phoswich block detector with DOI capability. This detector design will be used in the new CTI High Resolution Research Tomograph, ECAT HRRT. The two crystal layers (19×19×7.5 mm<sup>3</sup>) and a light guide are stacked on each other and mounted on a (2×2) PMT set, so that the corners of the phoswich are positioned over the PMT centers. The crystal phoswich is cut into an 8×8 matrix of discrete crystals. The separation of the LSO and the GSO layer by pulse shape discrimination allows discrete DOI information to be obtained. The high light output and the light guide design results in an accurate identification of the 128 single crystals per block. Flood source measurements document a very good homogeneity of events, energy centroid stability and energy resolution (14-20% FWHM) per single crystal. An intrinsic resolution of ~1.3 mm and the DOI feasibility is extracted by coincidence measurements with a single GSO crystal
Performance Characteristics of a Compact Position-Sensitive LSO Detector Module
  • J J Vaquero
  • J Seidel
  • S Siegel
  • W R Candler
  • M V Green
J.J. Vaquero, J. Seidel, S. Siegel, W.R. Candler, M.V. Green. "Performance Characteristics of a Compact Position-Sensitive LSO Detector Module". IEEE Trans Med h a g, in press.