Professional Documents
Culture Documents
14
The Gamma Camera:
Performance
Characteristics
The performance of a gamma camera system blurring caused by the collimator, is called the
is defined by the sharpness and detail of the intrinsic spatial resolution of the camera.
images it produces, the efficiency with which Intrinsic resolution is limited primarily by
it detects incident radiation, its ability to two factors. The first is multiple scattering of
measure the energy of the incident rays (to -ray photons within the detector. If a photon
minimize scatter), and the counting rate it can undergoes Compton scattering within the
handle without significant dead time losses. A detector crystal and the residual scattered
gamma camera is not capable of producing photon also is detected, but at some distance
perfect images of the radionuclide distribu- away, the two events are recorded as a single
tion. Certain inherent imperfections arise event occurring at a location along the line
from the performance characteristics of the joining the two interaction sites. This is not
detector, its associated electronic circuitry, a serious cause of degraded resolution for
and the collimator. Image artifacts also can be photon energies 300keV in which multiple
caused by malfunctions of various camera scatter Compton interactions in NaI(Tl) are
components. In this chapter, we describe the almost negligible. Even at 662keV, Anger
major factors that determine gamma camera calculated that for a detector thickness of
performance and examine the limitations that 6.4mm, less than 10% of photons are mis-
can lead to artifacts in gamma camera images placed by more than 2.5mm as a result of
and their correction. Standard tests of gamma multiple scattering events.1
camera performance also are summarized. The second, and primary, cause of limited
intrinsic resolution is statistical fluctuation
in the distribution of light photons among
A. BASIC PERFORMANCE photomultiplier (PM) tubes from one scintil-
lation event to the next. The problem is
CHARACTERISTICS
exactly analogous to the statistical fluctua-
tions observed in radioactive decay, discussed
1. Intrinsic Spatial Resolution in Chapter 9. If a certain PM tube records, on
Spatial resolution is a measure of the sharp- average, N light photons from scintillation
ness and detail of a gamma camera image. events occurring at a certain location in the
Sharp edges or small, pointed objects produce detector crystal, the actual number recorded
blurred rather than sharply defined images. from one event to the next varies with a stan-
Part of the blurring arises from collimator dard deviation given by N . Thus if a very
characteristics discussed in Sections C and D narrow beam of rays is directed at a point
and part arises in the sodium iodide [NaI(Tl)] on the detector, the position of each event as
detector and positioning electronics. The limit determined by the positioning circuitry or
of spatial resolution achievable by the detec- computer algorithm is not precisely the same.
tor and the electronics, ignoring additional Rather, they are distributed over a certain
209
210 Physics in Nuclear Medicine
area, the size of which depends on the mag- their distribution are therefore proportional
nitude of these statistical fluctuations. to 1/ N . This causes noticeably greater blur-
A detailed method for measuring and char- ring at lower -ray energies. An example of
acterizing intrinsic spatial resolution is dis- the change of intrinsic spatial resolution as
cussed in Section E.1. Typically, a lead mask a function of -ray energy is shown in Figure
containing a number of narrow (~1mm) slits 14-1.
is placed on the face of the gamma camera Intrinsic resolution also depends on detec-
(without the collimator) and the camera is tor crystal thickness. Thicker detectors result
irradiated using a 99mTc (140-keV) point in greater spreading of scintillation light
source. The resulting image is a series of lines before it reaches the PM tubes. Furthermore,
corresponding to the locations of the slits there is a greater likelihood of detecting
(e.g., see Fig. 14-10A). The resolution is cal- multiple Compton-scattered events in thicker
culated as the full width at half maximum detectors, particularly with higher-energy
(FWHM) of a profile drawn perpendicular to radionuclides. These are the primary reasons
the image of the lines at various locations in why gamma cameras use relatively thin
the field of view. The intrinsic spatial resolu- detectors in comparison with NaI(Tl) systems
tion of modern large field-of-view gamma that are used for counting applications.
cameras measured with 99mTc in this manner Figure 14-2 shows an example of the intrin-
is in the range of 2.9- to 4.5-mm FWHM. sic spatial resolution versus crystal thickness
Because the resolution is considerably worse for 140-keV rays.
than the width of the slits, the contribution of Intrinsic resolution improves with increased
the slits themselves to the measured resolu- efficiency of collection of scintillation photons.
tion is very small (10% for measured resolu- Modern cameras are substantially improved
tion 2.5mm). over earlier versions in this regard because of
Intrinsic resolution becomes worse with the use of more efficient PM tubes and of
decreasing -ray energy because lower-energy better techniques for optical coupling between
rays produce fewer light photons per scin the detector crystal and the PM tubes. The
tillation event, and smaller numbers of light use of greater numbers of smaller PM tubes
photons result in larger relative statistical (5-cm-diameter tubes have become the stan-
fluctuations in their distribution (Chapter 9, dard, and some gamma cameras have as
Section B.1). As a rule of thumb, intrinsic many as 110 PM tubes per head) and improved
resolution is proportional to 1/ E, in which electronics also have contributed to this
E is the -ray energy. This follows because improvement. Accurate corrections for non-
the number of scintillation light photons linearity (see Section B.1) and nonuniformity
produced, N, is roughly proportional to E (see Section B.2) have also resulted directly
and the relative statistical fluctuations in in improvements in intrinsic resolution, as
4.5
Intrinsic resolution (mm FWHM)
4
FIGURE 14-1 Intrinsic spatial resolution
of a gamma camera as a function of -ray
energy for a 6.3-mm-thick NaI(Tl) crystal.
(Compiled with data from Sano RM,
3.5 Tinkel JB, LaVallee CA, Freedman GS:
Consequences of crystal thickness reduc-
tion on gamma camera resolution and sen-
sitivity. J Nucl Med 19:712-713, 1978; and
Muehllehner G: Effect of crystal thickness
3 on scintillation camera performance. J
Nucl Med 20:992-993, 1979.)
2.5
0 50 100 150 200 250 300 350 400
-ray energy (keV)
14 The Gamma Camera: Performance Characteristics 211
5.5
FIGURE 14-2 Intrinsic spatial resolution
of a gamma camera at 140keV as a function
of crystal thickness. (Compiled with data 5
discussed in the following sections. The best 100% efficient for energies up to approxi-
reported intrinsic resolution for a large field- mately 100keV for all crystal thicknesses,
of-view gamma camera is just below 3mm but then shows a rather marked decrease in
FWHM at 140keV ( 99mTc). Significant efficiency at higher energies, depending on
improvements beyond approximately 2mm crystal thickness. At 140keV (-ray energy of
99m
FWHM will be difficult to achieve, owing to Tc), the difference in efficiency between
the ultimate limitation of the light photon 6.4-mm and 12.7-mm-thick crystals is approx-
yield of NaI(Tl). In most practical situations, imately 20% and the photopeak detection
however, the intrinsic spatial resolution efficiency is in the 70% to 90% range. At
makes a negligible contribution to the overall approximately 500keV, the standard gamma
system resolution of the gamma camera, camera (detectors 0.64-0.95-cm-thick) is less
which is largely determined by the resolution than 20% efficient at converting incident
of the collimator (see Sections C and D). rays into photopeak pulses.
At high energies, the performance of gamma
2. Detection Efficiency cameras with 0.64- to 1.27-cm-thick crystals
The gamma camera employs a sodium iodide is limited by decreasing detection efficiency
crystal that is relatively thin in comparison (as well as increasing collimator septal
with most other sodium iodide detectors used penetrationsee Section C.2). Deteriorating
in nuclear medicine: 6.4 to 12.7mm versus 2 intrinsic spatial resolution becomes the limit-
to 5cm for probe counting systems, scanners, ing factor at lower energies. Because of these
and so on. The trade-off in gamma cameras is tradeoffs, the optimal -ray energy range is
between detection efficiency (which improves approximately 100 to 200keV for most gamma
with thicker crystals) and intrinsic spatial cameras. Some gamma cameras are now fitted
resolution (which improves with thinner with thicker crystals (12.7-25.4mm), enabling
crystalssee Fig. 14-2). The gamma camera is them to achieve improved efficiency for
designed to provide acceptable detection effi- imaging positron-emitting radionuclides at
ciency while maintaining high intrinsic spatial 511keV (Chapter 18, Section B.4). This comes
resolution in the energy range of 100-200keV. at the expense of some loss of intrinsic spatial
As a result, the detection efficiency of the resolution (see Fig. 14-2) when these systems
gamma camera detector is somewhat less than are used in the 100-200-keV energy range.
would be desirable at higher -ray energies.
Figure 14-3 shows photopeak detection 3. Energy Resolution
efficiency versus -ray energy for the gamma It is not unusual in a typical patient study for
camera detector for a range of NaI(Tl) crystal there to be more Compton-scattered than
thicknesses. The gamma camera is nearly unscattered rays striking the detector (see
212 Physics in Nuclear Medicine
5.08
2.54
10 1.27
0.64
1
0 100 200 300 400 500 600 700
-ray energy (keV)
FIGURE 14-3 Photopeak detection efficiency versus -ray energy for NaI(Tl) detectors of different thicknesses.
(Adapted from Anger HO: Radioisotope cameras. In Hine GJ [ed]: Instrumentation in Nuclear Medicine, Vol 1. New
York, 1967, Academic Press, p 506.)
and those scattered through smaller angles cardiac studies. Dead time corrections can be
are rejected less efficiently. applied; however, these corrections generally
Two advantages are obtained with improved become increasingly inaccurate as counting
energy resolution. First, the photopeak losses increase.
becomes narrower, resulting in more efficient Because pulse pile-up can occur between
detection of unscattered photons within the any two events in the pulse-height spectrum,
chosen energy window. This increases the system counting losses are determined by
number of valid events recorded and improves total-spectrum counting rates. Most gamma
the statistical quality of the image. Second, cameras behave as paralyzable systems. The
rays scattered through large angles are apparent dead time for a selected energy
rejected more efficiently, because their energy window depends on the window fraction, that
spread within the pulse-height spectrum is is, the fraction of the total spectrum counting
also smaller. Thus image contrast is improved. rate occurring within that window. The smaller
It also is true that rays scattered through the window fraction, the larger the apparent
smaller angles are detected somewhat more dead time. Thus the apparent dead time is
efficiently, because of the narrowing of their longer when a photopeak window is used than
distribution as well. However, the increased when a full-spectrum window is used. The
efficiency for recording photopeak events apparent dead time also is longer when
more than offsets this effect, in terms of scattered radiation is present, because this
contrast-to-noise ratio (Chapter 15, Section also adds to the counting rate outside the
D.2). Alternatively, one can take advantage of photopeak window (Fig. 14-5). Therefore,
the improved energy resolution to use a nar- when specifying gamma camera dead time, it
rower PHA window, trading back some of the is important to note the conditions of mea
increased efficiency for recording photopeak surement. Dead time values as short as 1 to 2
events for improved rejection of small-angle sec can be obtained in the absence of scatter-
scatter. Either way, improved energy resolu- ing material with a full-spectrum window;
tion results in better image quality. however, under clinically realistic conditions
(99mTc source in scattering material, 15% pho-
4. Performance at High topeak window), system dead times of 4 to 8
Counting Rates sec are more typical. For a dead time of 5 sec,
At high counting rates, there is increased like- counting losses are approximately 20% for a
lihood of recording two events at the same counting rate of 4 104 counts per second (cps).
time. The most troublesome effect is known as Dead time losses are not serious in most
pulse pile-up (Chapter 8, Section B.3). Pulse static imaging studies, but they can be impor-
pile-up has two undesirable effects on gamma tant in certain high-counting-rate applica-
camera performance: counting losses and tions (e.g., first-pass cardiac studies) in which
image distortion. counting rates as high as 105 cps may occur.
Counting losses cause inaccurate counting Pile-up rejection circuitry (see Chapter 8,
rates to be recorded at higher counting rates. Section B.3) is used to achieve higher usable
The inaccuracies are described by conven- counting rates in such situations. Another
tional dead time models (Chapter 11, Section approach for shortening camera dead time is
C) and may be significant in some high-count- by the use of analog buffers, or derandomiz-
rate quantitative studies, such as first-pass ers. These are electronic circuits that hold a
Observed counting rate
Source Source
Source behind in air No losses in air
Counts
10 cm water
voltage level or pulse from one circuit compo- (see Chapter 13, Section B.2), two events
nent (e.g., an amplifier) until the next circuit detected simultaneously at different locations
in the pulse-processing sequence (e.g., the in the detector are recorded as a single event
PHA) is ready to receive it. with energy equal to the sum of the two
Similarly, in digital gamma cameras, data events, at a location somewhere between
can be buffered in memory until the computer them (Fig. 14-6). If both are valid photopeak
is ready to process them. Both these approaches events, their total energy exceeds the value
result in a decrease in the apparent dead that would be accepted by the PHA window
time of the camera by effectively changing the and both events are rejected, resulting in
arrival times of the pulses. This, however, counting losses. On the other hand, it is pos-
means that the simple dead time models and sible for two Compton-scattered rays to
corrections presented in Chapter 11, Section have a total energy that falls within the
C can no longer be used, and more complex selected energy window, so that two invalid
modeling of system dead time must be carried events are accepted as a single valid event.
out to produce accurate correction at high The visible result at very high counting rates
counting rates. is to add a diffuse background to the image,
It also is possible to physically shorten the as illustrated in Figure 14-7. Note as well the
dead time of a camera by shortening the image in the upper right-hand corner of this
charge integration time from the PM tubes figure, showing how contrast can be restored
and using electronic circuitry that returns the by shielding high-activity areas outside the
signal to baseline after the chosen integration imaging area of interest (e.g., with a thin
time.2 Clearly, this also decreases the amount sheet of lead).
of signal used for determining event location. Early pile-up rejection methods were based
For example, with a charge integration time on measuring the length of a pulse. If the
of 0.4 sec, only 81% of the scintillation light pulse did not return close to baseline level
is collected, compared with 98% for a 1-sec within the time expected given the decay time
integration time. This causes a degradation of of NaI(Tl), it was assumed that pile-up of
intrinsic spatial resolution and energy resolu- two pulses had occurred and the event was
tion. Some gamma cameras have a variable rejected, resulting in the loss of both rays.
integration time, in which the charge integra- This improved image quality but resulted in
tion is automatically shortened as the count- an effective increase in system dead time,
ing rate increases. because many events were rejected at high
Other means for shortening dead time are counting rates.
to bypass altogether the pile-up rejection cir- Many gamma cameras now incorporate cir-
cuits and nonuniformity correction circuitry cuits that continuously monitor the decay of
(see discussion on pile-up correction later in a pulse and use a method based on pulse-tail
this section and on nonuniformity and its cor- extrapolation for pile-up correction. Consider
rection in Section B.3). The signal processing two -ray interactions that occur close together
that occurs in these circuits slows down the in time and create overlapping pulses. When
rate at which the camera can handle indi- the second ray arrives, the decay of the
vidual events, and bypassing them can pulse created by the first ray immediately
shorten system dead time from typical values
of 4 to 8 sec down to 1 to 3 sec. Some
cameras provide an optional high count rate
mode of operation in which some or all of
these corrections are turned off by software Source
control. This mode is intended specifically for Pile-up
applications requiring high counting rates, Source
such as first-pass cardiac studies. Normal
mode, in which all corrections are employed,
is used for routine imaging to obtain the
desired high-quality images. Obviously speci-
fications for gamma camera dead time should
indicate whether any circuits were bypassed
to achieve the reported value.
FIGURE 14-6 Images of two 99mTc point sources of rela-
The second undesirable effect of pulse tively high activities (~370 MBq each). Events appearing
pile-up is image distortion. Using standard in the band between the two point-source locations are
pulse-positioning logic for gamma cameras mispositioned events caused by pulse pile-up.
14 The Gamma Camera: Performance Characteristics 215
105
3.2 min
103
1.0 10 100
Activity in phantom (mCi)
deviates from the expected exponential decay of NaI(Tl). At the same time, this extrapo-
and the gamma camera signal is switched to lated tail is also sent to the second amplifier
a second amplifier circuit. Estimator circuitry circuit and subtracted from the second pulse.
in the first amplifier circuit completes the This removes the contribution of the pulse
signal from the first ray by extrapolating the generated by the first ray from that of the
remainder of the tail of the pulse with an second ray. This process is summarized
exponential function based on the decay time in Figure 14-8. The pulse-tail extrapolation
P2 P1 + P2
P1
Amplifier
Estimator
P2
Subtractor
Amplifier 2
P2 + P1 Tail
216 Physics in Nuclear Medicine
B. DETECTOR LIMITATIONS:
NONUNIFORMITY AND
NONLINEARITY
1. Image Nonlinearity
A basic problem arising in the detector and
electronics is image nonlinearity. Straight-line
objects appear as curved-line images. An
inward bowing of line images is called pin-
cushion distortion; an outward bowing is Pincushion Barrel
called barrel distortion (Fig. 14-9). Non distortion distortion
linearities result when the X- and Y-position FIGURE 14-9 Appearance of straight-line objects with
signals do not change linearly with pincushion and barrel distortions.
A B
FIGURE 14-10 A, Illustration of nonlinearities in images of a straight-line test pattern obtained with a gamma camera.
Image demonstrates subtle waviness in the lines. B, Flood-field image obtained by exposing the same camera to a
uniform radiation field. This is the image obtained in the absence of any corrections for nonuniformity. Notice that the
photomultiplier tube pattern can be seen. The gray levels in this image are confined to a narrow display window to
improve visualization of the artifacts. (Images courtesy Dr. Magnus Dahlbom, UCLA School of Medicine, Los Angeles,
CA.)
14 The Gamma Camera: Performance Characteristics 217
modern gamma camera to demonstrate the in areas of barrel distortion events are pushed
general appearance of nonlinearities. On outward from the center, causing an apparent
close inspection, some waviness of the lines is cold spot. Because of the characteristic pin-
apparent. On properly functioning cameras, cushion distortions occurring in front of PM
including the one illustrated, the nonlineari- tubes, it is common to see a pattern of hot
ties themselves (including the pincushion spots at the locations of the PM tubes on an
distortions in front of PM tubes) are barely otherwise uniform gamma camera image.
perceptible and rarely interfere directly with Other causes of nonlinearities (e.g., PM tube
image interpretation; however, they can have failure, crystal cracking, and collimator
significant effects on image nonuniformities, defects) also can result in nonuniformities.
as discussed in the following section. Another characteristic nonuniformity is a
bright ring around the edge of the image. This
2. Image Nonuniformity artifact, called edge packing, results from a
A more noticeable problem is image non somewhat greater light collection efficiency
uniformity. Exposing the detector crystal to a for events near the edge versus central regions
uniform flux of radiation produces a flood- of the detector crystal. This is the result of
field image with small but noticeable nonuni- internal reflections of scintillation light from
formities in intensity, even with a properly the sides of the detector crystal back into the
functioning camera. These variations may be PM tubes near the edge. Also, for events
equivalent to counting rate variations of 10% occurring toward the center of the crystal,
or more. A flood-field image from a gamma there are always PM tubes on either side of
camera demonstrating image nonuniformity the event location, whereas at the edges of the
is shown in Figure 14-10B. Intrinsic flood- crystal there are PM tubes only to one side.
field images are acquired with the collimator Thus events at the very edges are not distrib-
removed, using a point source placed far uted uniformly across the edge, but are
enough from the surface of the gamma camera pulled toward the center, compounding the
to give uniform irradiation of the surface (dis- edge-packing artifact. The portion of the
tance equal to 4-5 times the camera diame- image demonstrating this artifact usually is
ter). Extrinsic flood-field images are acquired masked on the image display and therefore is
with the collimator in place using a disk or not a part of the useful field of view (UFOV ).
thin flood phantom that covers the area of the Typically, 5cm or more of the detector width
detector. 99mTc or 57Co are the two most is eliminated by the mask. When specifying
commonly used radionuclides for flood-field gamma camera detector dimensions, it is
measurements. important to distinguish between the physi-
There are two primary causes of gamma cal dimensions of the crystal and the dimen-
camera nonuniformities. The first is non- sions of the useful imaging area.
uniform detection efficiency arising from (1) Both nonuniformity and edge-packing
small differences in the pulse-height spec- artifacts are related to the pattern of the
trum for different PM tubes and (2) position- distribution of scintillation light falling on
dependent collection efficiency of scintillation the PM tubes. For this reason, they also have
light, particularly for events located over the an energy-dependent component. When the
gaps and dead areas between the PM tubes gamma camera is used to image higher-
compared with events located directly over energy radionuclides, interactions, on average,
the center of a PM tube. The differences in occur deeper in the crystal, closer to the PM
PM tube response can be minimized by careful tubes. This produces a more narrow light
selection and tuning of all of the PM tubes of spread distribution on the PM tubes and gen-
a gamma camera; however, position-dependent erally results in a worsening of detector
effects on the pulse-height spectrum remain. nonuniformity.
If a fixed pulse-height window is used for all
output pulses, the result is an apparent dif- 3. Nonuniformity Correction
ference in detection efficiency owing to differ- Techniques
ences in the window fraction for different All modern gamma cameras incorporate tech-
areas of the crystal (see Fig. 13-6, top). niques that attempt to correct the causes of
The second cause of nonuniformities is nonuniformity described in the preceding
image nonlinearities described in Section section. All of these techniques begin with
B.1. In areas of pincushion distortion events spatially varying energy corrections, normally
are crowded toward the center of the distor- derived from an intrinsic flood-field image.
tion, causing an apparent hot spot, whereas The flood-field image is divided into a matrix
218 Physics in Nuclear Medicine
of small, square elements, typically 128 128 counts recorded in each pixel are thrown out,
elements (or pixels). Using the PHA, the depending on the relative value for that pixel
channel number (pulse amplitude) of the photo in the energy-corrected flood-field image.
peak in the pulse-height spectrum is deter- For example, if the value in the normalized
mined for each element. This information is intensity-correction matrix is 110, then 1 of
stored in a 128 128 look-up table and used every 11 counts is subtracted from the patient
to set regionally varying PHA windows for image at that location. This process is some-
subsequent studies on patients. For example, times called count skimming.
if a 20% window is chosen for a patient study, Most modern digital gamma cameras
and the center of the photopeak is found in replace the second step described in the pre-
PHA channel 100 in a particular pixel in ceding paragraph with a correction for image
the flood-field image, then events at that nonlinearity, which more directly attacks the
location having Z-signal amplitudes between major underlying cause of image nonunifor-
PHA channels 90 and 110 are accepted in mity. For nonlinearity corrections, another
patient imaging studies. If the center of the flood-field image is obtained, this time with a
photopeak is in channel 110 at another loca- sheet of lead having a uniformly spaced array
tion, events for which the Z-signal falls within of small holes (~1-mm diameter, ~4-mm sepa-
the range of 99 to 121 are accepted at that ration) placed directly on the gamma camera
location. The position-dependent PHA window face (no collimator). The locations of the images
corrects for variations in the pulse-height of these holes are compared with their known
spectrum across the face of the camera detec- locations in the lead sheet to derive a matrix
tor. It also provides a partial correction for of offsets, x and y, for each (X,Y ) location on
image nonuniformity. the detectors, which is stored as another
The second step in the nonuniformity cor- look-up table. When an event is detected, its X
rection is to account for the remaining regional and Y coordinates are computed using conven-
variations in image intensity, largely caused tional positional circuitry or algorithms. These
by detector nonlinearity. In one older method, values then are corrected using the positional
the correction is based directly on variations offsets for that location stored in the look-up
in intensity of the energy-corrected flood-field table. The offsets and the corresponding
image. The number of counts recorded within look-up table usually are measured and gener-
each pixel in that image is stored in a matrix ated at the factory prior to shipment.
and compared to the smallest number Figure 14-11 shows the same data as Figure
recorded in the pixel array. This is used to 14-10, after the corrections for nonuniformity
derive a matrix of normalized intensity values, and nonlinearity described in the preceding
which range from 100 for the coldest pixel paragraphs have been applied. Figure 14-12,
to higher values for other pixels. In subse- showing intensity profiles across the flood-
quent patient studies, a certain fraction of the field images in Figures 14-10 and 14-11,
A B
FIGURE 14-11 Straight-line test pattern (A) and uniform flood-field (B) images after nonuniformity corrections are
applied. Compare with Figure 14-10. (Images courtesy Dr. Magnus Dahlbom, UCLA School of Medicine, Los Angeles,
CA.)
14 The Gamma Camera: Performance Characteristics 219
1.2 1.2
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 50 100 150 200 250 0 50 100 150 200 250
Pixel Pixel
No correction Corrected
FIGURE 14-12 Profiles through the uniform flood-field images in Figures 14-10 and 14-11 showing relative uniformity
of flood-field image with and without nonuniformity correction. The standard deviation is improved from 3.4% to 1.9%
after correction.
clearly illustrates the improvements. The the tube output has changed by more than 1%
examples in these figures are for 99mTc. Note from the original reference value.
that different correction matrices must be In newer gamma cameras, the large
obtained for each radionuclide used, because number of PM tubes makes the manual
the effects corrected for generally vary with method impractical. Many digital gamma
-ray energy, for example, because of different cameras therefore contain tuning circuitry
average depths of interaction in the NaI(Tl) that allows the output of each individual PM
crystal. tube to be automatically adjusted to a set of
Improvements in camera uniformity also reference outputs. One automated approach
have contributed to improvements in intrinsic involves the use of light-emitting diodes
resolution. Earlier cameras used thicker light (LEDs) that are coupled to the neck of each
guides and large-diameter PM tubes, in part PM tube. These LEDs are pulsed to produce
to achieve satisfactory uniformity, at the a light signal on the photocathode of the PM
expense of somewhat degraded spatial resolu- tube that does not vary with time. The PM
tion. Because of effective uniformity correc- tube signals are then monitored and the pre-
tions, newer gamma cameras can use thinner amplifier adjusted electronically if the PM
light guides (or eliminate the light guide tube signal has drifted.
entirely) and smaller PM tubes, both of which A second approach uses two narrow energy
contribute to more accurate event localization windows, placed just above the photopeak
and improved intrinsic spatial resolution. position to minimize the influence of scatter
(Fig. 14-13). The count ratio between the two
4. Gamma Camera Tuning energy windows during flood-field irradiation
The nonuniformity corrections described pre- by the radionuclide of interest is measured for
viously require that the gamma camera each PM tube. This ratio remains constant,
remain very stable over time. However, the unless the PM tube signal drifts over time. If
gain of PM tubes invariably changes as the the count ratio changes, the PM tube pream-
tubes age. The high-voltage supply and ampli- plifier is adjusted electronically to restore the
fier gain can also drift over time. A method to ratio to its original value.
tune the PM tubes to ensure consistent per- Some of these tuning methods also can be
formance over time is therefore necessary. adapted so that they are continuous, in the
On many older systems, the tuning is done sense that the camera is tuned dynamically
manually. One method involves irradiation of every few seconds during a patient study.
the gamma camera detector through a lead This can be used to adjust the energy
mask with holes centered over each of the PM windows in real time, compensating for any
tubes. The output of each PM tube is exam- drift that occurs during the course of a study.
ined and the preamplifier gain is adjusted if The major cause of drift on such short
220 Physics in Nuclear Medicine
500 Window 1
400
Window 2
Counts
300
200
100
0 50 100 150
Energy (keV)
FIGURE 14-13 The ratio of counts detected in two narrow energy windows can be used to determine whether a pho-
tomultiplier tube is drifting. The windows are defined on the high side of the photopeak to avoid any contribution from
scatter. The ratio is virtually independent of source distribution and the amount of scattering material present.
Collimator efficiency g, defined as the frac- camera systems also provide a selection of col-
tion of rays passing through the collimator limators with different combinations of resolu-
per ray emitted by the source is given by tion and efficiency. Those with good resolution
but poor efficiency generally are described as
g K 2 (d/leff )2 [ d2 / (d + t)2 ] (14-7) high-resolution collimators, whereas those
with the opposite characteristics are described
where t is septal thickness and K is a constant as high-sensitivity collimators. Those with
that depends on hole shape (~0.24 for round characteristics intermediate to the extremes
holes in a hexagonal array, ~0.26 for hexago- are referred to as general purpose, all
nal holes in a hexagonal array, ~0.28 for purpose, or by other similar names.
square holes in a square array1). Equation Equation 14-6 indicates that collimator
14-7 applies to a source in air and assumes resolution becomes poorer as source-to-
no attenuation of radiation by intervening collimator distance b increases. Thus struc-
body tissues. tures closest to the collimator are imaged
Several aspects of Equations 14-6 and 14-7 with sharpest detail. Figure 14-16 shows
should be noted. First, resolution improves as graphically the relationship between colli-
the ratio of hole diameter to effective length mator resolution and source-to-collimator
(d/leff) is made smaller. Long, narrow holes distance for three different collimators pro-
provide images with the best resolution; vided by one commercial manufacturer. Typi-
however, collimator efficiency decreases cally, collimator resolution deteriorates by a
approximately as the square of the ratio of factor of 2 at a distance of 4-6cm from the
hole diameter to length (d/leff)2. Thus an collimator.
approximate relationship between collimator On the other hand, according to Equation
efficiency, g, and spatial resolution, Rcoll, is 14-7, collimator efficiency for a source in air
is independent of source-to-collimator dis-
g ( Rcoll )2 (14-8) tance b. This rather surprising result is
obtained provided the counting rate for the
Therefore for a given septal thickness, colli- entire detector area is measured. The reason
mator resolution is improved only at the for this is illustrated by Figure 14-17. As the
expense of decreased collimator efficiency, source is moved farther away from the colli-
and vice versa. mator, the efficiency with which radiation is
transmitted through any one collimator hole
EXAMPLE 14-3 decreases in proportion to 1/b2 (inverse-square
Calculate the efficiency g of the collimator law), but the number of holes through which
described in Examples 14-1 and 14-2, assum- radiation can pass to reach the detector
ing it has hexagonal holes in a hexagonal increases in proportion to b2. The two effects
array. cancel each other, with the result that total
counting rateand thus collimator efficiency
Answer does not change with source-to-collimator dis-
For hexagonal holes in a hexagonal array, tance. Another illustration of this effect is
K = 0.26. Thus, shown in Figure 14-18. As source-to-collima-
tor distance increases, the maximum height
g (0.26)2 (0.25 / 2.4)2 [(0.25)2 / (0.25 + 0.03)2 ] of the PSF or LSF decreases, but the width
(0.0676) (0.0109) (0.797) increases (and resolution becomes poorer), so
that the total area under the curve (total
5.85 104
detector counting rate) does not change.
(photons transmitted/photons emitted))
Invariance of collimator efficiency with
This example illustrates the relatively small source-to-collimator distance applies to point
fraction of emitted rays that are transmitted sources, line sources, and uniform sheet
by a typical gamma camera collimator. sources in air with parallel-hole collimators;
however, it applies only to uniform sheet
Equation 14-7 also demonstrates the effect sources with converging, diverging, or pinhole
of septal thickness on efficiency. Medium- collimators (Section D). When the source is
energy collimators have lower efficiencies embedded at different depths in the patient,
than low-energy collimators because of their attenuation effects also must be considered.
greater septal thicknesses. Septal penetration and scatter of photons
In addition to providing low- and medium- from the walls of the collimator holes also are
energy collimators, manufacturers of gamma not considered in the earlier analysis.
224 Physics in Nuclear Medicine
2.0
1.8
1.6
Collimator resolution FWHM (cm)
1.4
ity
1.2 itiv
ens
hs
1.0 Hig e
pos
al pur
ner
0.8 Ge
0.6 tion
solu
igh re
H
0.4
0.2
0.0
0 2 4 6 8 10 12 14 16
Source-to-collimator distance (cm)
FIGURE 14-16 Collimator resolution versus source-to-collimator distance for three different collimators. (Adapted
from Hine GJ, Paras D, Warr CP: Recent advances in gamma-camera imaging. Proc SPIE 152:123, 1978.)
Single-hole
efficiency 1/b2
Collimator
Detector
Counting rate
Distance
14 The Gamma Camera: Performance Characteristics 225
TABLE 14-1
PERFORMANCE CHARACTERISTICS OF SOME TYPICAL COMMERCIALLY MANUFACTURED
PARALLEL-HOLE COLLIMATORS
Resolution Rcoll
Recommended Max. (FWHM at
Collimator Type Energy (keV) Efficiency, g 10cm)
Low-energy, high-resolution 150 1.84 104 7.4mm
Low-energy, general-purpose 150 2.68 10 4
9.1mm
Low-energy, high-sensitivity 150 5.74 10 4
13.2mm
Medium-energy, high-sensitivity 400 1.72 104 13.4mm
Adapted from Hine GJ, Erickson JJ: Advances in scintigraphic instruments. In Hine GJ, Sorenson JA (eds):
Instrumentation in Nuclear Medicine, Vol 2. New York, 1974, Academic Press.
FWHM, full width at half maximum.
Table 14-1 summarizes the physical con- differences between system resolutions for
struction and typical performance character- cameras having substantially different intrin-
istics of a number of collimators. Collimator sic resolutions (e.g., 4mm vs. 8mm), but the
resolution is the FWHM for a source at 10cm difference in system resolutions for cameras
from the face of the collimator. Collimator effi- having small differences in intrinsic resolu-
ciency g refers to the relative number of rays tions (e.g., 4mm vs. 5mm) is minor and not
transmitted by the collimator and reaching clinically significant. Small differences in
the detector per ray emitted by the source. intrinsic resolution may be apparent on bar-
Note that the approximate relationship pattern images or on images of very superfi-
between collimator efficiency and resolution cial structures in the patient, but they usually
given by Equation 14-8 is verified by these are not apparent on images of deeper-lying
data. Note also the relatively small values for structures.
collimator efficiency. System resolution also is degraded by scat-
tered radiation. This is discussed in Chapter
4. System Resolution 15, Section C. The method for combining com-
The sharpness of images recorded with a ponent resolutions to determine system reso-
gamma camera is limited by several factors, lution also is discussed in Appendix G.
including intrinsic resolution, collimator reso-
lution, scattered radiation, and septal pene-
tration. In terms of the FWHM of a PSF or D. PERFORMANCE CHARACTERISTICS
LSF, the most important factors are the OF CONVERGING, DIVERGING, AND
intrinsic resolution Rint of the detector and PINHOLE COLLIMATORS
electronics, and the collimator resolution Rcoll.
The combined effect of these two factors is to Figure 14-20 illustrates the important design
produce a system resolution Rsys that is some- parameters for converging, diverging, and
what worse than either one alone. System pinhole collimators. Equations for collimator
resolution Rsys (FWHM) is given by resolution, Rcoll, and efficiency, g, for these col-
limators are as follows:
Rsys = 2
Rint + Rcoll
2 (14-9)
Converging Collimator:
Because collimator resolution depends on
source-to-collimator distance, system resolu- Rcoll [ d(leff
+ b) / leff
][1 / cos ]
tion also depends on this parameter. Figure (14-10)
[1 (leff
/ 2) /( f + leff
)]
14-19 shows system resolution versus source-
to-collimator distance for a typical parallel-
hole collimator and different values of intrinsic g K 2 (d / leff
)2 [ d2 /(d + t)2 ][ f 2 /( f b)2 ]
resolution. At a distance of 5-10cm (typical (14-11)
depth of organs inside the body), system
resolution is much poorer than intrinsic where
resolution and is determined primarily by
collimator resolution. There are significant (l 2 1 ) / cos leff / cos (14-12)
leff
226 Physics in Nuclear Medicine
1.8
m
1.6 8m
t
R in m
4m
1.4 t
R in
0
System resolution, FWHM (cm)
t
1.2 R in
0.4 Typical
organ
depths
0.2
0.0
0 2 4 6 8 10 12 14 16
Source-to-collimator distance (cm)
f
d d detector
l l l
d
b b
b
f
Rcoll [ d(leff
+ b) / leff
][1 / cos ][1 + (leff
/ 2 f )] Rcoll deff ,R (l + b) /l (14-15)
(14-13)
g K 2 (d / leff
)2 [ d2 /(d + t)2 ][( f + l) /( f + l + b)]
(14-14) g deff ,g 2 cos3 / (16b2 ) (14-16)
14 The Gamma Camera: Performance Characteristics 227
20 250
Diverging
16 200 Converging
Relative geometric efficiency
System resolution (mm)
Parallel-hole
12 150
Parallel-hole
8 100
Pinhole
4 Converging
50
Diverging
Pinhole
0 0
0 5 10 15 20 0 5 10 15 20
A B
Source-to-collimator distance (cm)
FIGURE 14-21 Performance characteristics ( A, system resolution; B, point-source geometric efficiency in air) versus
source-to-collimator distance for four different types of gamma camera collimators. (Adapted with permission from
Society of Nuclear Medicine from Moyer RA: A low-energy multihole converging collimator compared with a pinhole
collimator. J Nucl Med 15:59-64, 1974.)
228 Physics in Nuclear Medicine
(~35cm), where rays are transmitted distance; they also have a quite limited field
through all of the collimator holes, and then of view because of magnification effects at
decreases beyond that point. Point-source typical imaging distances (Equation 13-3).
collimator efficiency decreases with distance Generally they are used for imaging smaller
for the diverging and pinhole collimators, organs, such as the thyroid and heart, which
more severely for the latter. For an extended, can be positioned close to the collimator. They
large-area sheet source, sufficiently large to also are useful with high-energy -ray emit-
cover the entire field of view of the collimator, ters because they can be designed to reduce
efficiency does not change with source-to- septal penetration problems.
collimator distance for all of these collima- Differences between the resolution and
tors. Again, for sources embedded within a field-of-view obtained at different source-to-
patient, attenuation effects also must be collimator distances with parallel-hole, con-
accounted for. verging, diverging, and pinhole collimators are
Figure 14-21 illustrates that the converg- further illustrated by Figure 14-22. The distor-
ing collimator offers the best combination of tions caused by changing magnification with
resolution and efficiency at typical imaging depth for different structures inside the body
distances (5 to 10cm); however, the field-of- sometimes make images obtained with the
view is also somewhat limited at these dis- converging, diverging, and pinhole collimators
tances (Equation 13-6 and Example 13-2), difficult to interpret (see Fig. 13-9).
and for this reason converging collimators are
most useful with cameras having relatively
large-area detectors. Diverging collimators E. MEASUREMENTS OF GAMMA
offer a larger imaging area (Example 13-1) CAMERA PERFORMANCE
but at the cost of both resolution and effi-
ciency. Pinhole collimators offer very good It is important to define standardized experi-
resolution and reasonable efficiency at close mental protocols for measuring gamma
distances but lose efficiency very rapidly with camera performance that produce consistent
At
collimator
face
10 cm
distance
20 cm
distance
results, are easily reproducible, and that do (<10,000 cps for a small field-of-view gamma
not require specialized equipment. Such pro- camera) to avoid pile-up-related mis
tocols can then be used in comparing one positioning. Two sets of images are taken,
gamma camera with another, in performing with a 90-degree rotation of the test pattern
acceptance testing on a newly installed between acquisitions so that both X and Y
gamma camera, and as part of a quality resolution are measured. Data are acquired
assurance program to ensure that the camera until the peak channel has at least 1000
is always performing to its specifications. counts. Images are acquired in a matrix with
The exact regulations for gamma camera pixel sizes less than 1 10 of the expected reso-
quality assurance and the guidelines for lution (typically <0.35mm). Profiles through
acceptance testing vary with locality. For the images of the line sources are taken at
example, in the United States, the Joint different locations across the gamma camera
Commission requires that instruments be face and fitted to a Gaussian function
tested prior to initial use and that the per (Chapter 9, Section B.3). The FWHM (Fig.
formance of a gamma camera be tested at 14-15) and full width at tenth maximum
least once a year. Because of the rapidly (FWTM) of the profiles are measured in both
changing regulatory environment, and differ- X and Y directions. The reported measure-
ences between states and countries, a detailed ments usually are average measurements
review of the requirements of regulatory across the UFOV, and the average across the
agencies is beyond the scope of this book. In central field-of-view (CFOV ) that has linear
this section, we therefore briefly summarize dimensions scaled by 75% with respect to the
the more common measurements that are dimensions of the UFOV. Typical values of
performed to assess gamma camera perfor- intrinsic spatial resolution are 2.5 to 3.5mm.
mance (whether they be for acceptance
testing or for quality assurance). The proto- 2. System Resolution
cols presented here are largely based on the This measurement is made with the collima-
recommendations of the National Electrical tor in place and should be repeated for each
Manufacturers Association.6 A typical quality collimator of interest. The source consists of
assurance program might involve daily mea- two 1-mm-diameter line sources, placed 5cm
sures of flood-field uniformity, weekly checks apart at a distance of 10cm from the front
of spatial resolution and spatial linearity, face of the collimator. The measurement also
and semiannual checks of other performance can be performed with the addition of a scat-
parameters. It is important that all measure- tering medium by placing 10cm of plastic
ments be taken under the same conditions between the sources and the collimator, and
(pulse-height window width, correction algo- 5cm of the same material behind the sources.
rithms, and correction circuitry on or off ) as Images are acquired (typically several million
are used for routine clinical studies. More events, at a rate of <30,000 cps to avoid pile-
detailed information on performance mea- up) and profiles taken through the image of
surements, quality assurance, and accep- the line sources are fitted to Gaussian func-
tance testing can be found in references tions to determine FWHM and FWTM as
7 to 9. described for intrinsic resolution. The results
vary widely depending on the exact type of
1. Intrinsic Resolution collimator used but are typically in the range
Intrinsic resolution is determined without a of 8 to 14mm for 99mTc.
collimator using a linearity test pattern, such
as the one shown in Figure 14-10 (left), placed 3. Spatial Linearity
directly on the surface of the NaI(Tl) crystal This measurement uses the same slit pattern
housing. The width of the strips in the pattern (Fig. 14-10A) and conditions as for the intrin-
is approximately 1mm, which is significantly sic resolution measurement. Once again,
smaller than the resolution expected in the measurements are taken with two orienta-
measurement. A point source (usually 99mTc tions of the test pattern, rotated by 90 degrees,
or 57Co) is placed at a distance equal to five to provide linearity measurements in both X
times the UFOV from the gamma camera and Y directions. Two measurements can be
face. The UFOV corresponds to the field of made from the resulting images. The differ-
view of the gamma camera after masking off ential spatial linearity is the deviation of the
the portion of the camera face affected by measured distance di between two slits from
edge-packing effects. Data are acquired with the actual distance D between them calcu-
the system count rate below 30,000 cps lated for each row i in the image. The means
230 Physics in Nuclear Medicine
and the standard deviations are reported for across all rows and columns of the image. It
the X and Y directions across the UFOV and is defined as
the CFOV and are defined as:
(high low )
Differential Uniformity (%) = 100
n (high + low )
(d i D)
(14-19) (14-22)
i =1
Mean =
n where high refers to the maximum count
n difference for any five consecutive pixels (row
(d i D)2 (14-20) or column) in the image and low refers to
Standard Deviation = i =1 the minimum count different for any five con-
n 1 secutive pixels. This usually is reported for
the UFOV.
In addition, the absolute spatial linearity is For convenience, uniformity measurements
defined as the maximum deviation of the often are made with the collimator in place
location of the slits from their true location. (extrinsic uniformity). A thin flood-field source
Once again this is assessed for the UFOV of 99mTc or a disk source of 57Co that covers the
and the CFOV. It is not easy to detect small active area of the gamma camera is placed on
nonlinearities using these techniques, and top of the collimator to provide uniform irra-
tests of uniformity (discussed in the next diation. This protocol is more practical for
section) usually are better at revealing the routine quality assurance because the mea-
effects of small nonlinearities. surement can be done without removing the
collimator. Extrinsic uniformity measure-
4. Uniformity ments also have the advantage that they
Intrinsic uniformity is determined from flood- reveal any defects or problems caused by the
field images acquired without a collimator. A collimator itself.
99m
Tc source is placed at a distance of approxi-
mately 5 the UFOV from the front face of 5. Counting Rate Performance
the gamma camera. The source activity is As described in Section A.4, most gamma
such that the counting rate on the gamma cameras behave as paralyzable counting
camera is less than approximately 30,000 cps. systems with the observed count rate described
Flood-field images are acquired so that there as a function of the true count rate by Equa-
are a minimum of 4000 counts in each pixel tion 11-18. The basis for measurement of the
of the image and then smoothed with a 9-point dead time, , is the two-source method
(3 3) smoothing filter with the following described in Chapter 11, Section C.4. Two
99m
weightings: Tc sources are placed approximately 1.5m
away from the camera face. The total activity
1 2 1 should be sufficient to cause approximately a
2 4 2 20% loss in the observed counting rate rela-
tive to the true counting rate. Counting rates
1 2 1 then are measured with both sources present,
and then with each individual source present.
Integral uniformity is based on the Care must be taken that all measurements
maximum and minimum pixel counts in the are performed with exactly the same source
image and is defined as geometry, that pile-up rejection electronics or
any other high counting rate correction cir-
Integral Uniformity (%) cuitry is turned on, and that source decay is
negligible (<1%) during the course of the mea-
max. pixel count min. pixel count
= 100 surement. The dead time can then be calcu-
max. pixel count + min. pixel count lated from Equation 11-25. The observed
(14-21) count rate at which a 20% counting rate loss
occurs, R20%, is also often quoted, and this can
This is calculated for the UFOV and CFOV. be computed from Equation 11-18 using the
Integral uniformity values are typically 2% fact that Ro = 0.8Rt as
to 4%.
Differential uniformity is based on the 0 .8
R20% = ln(0.8) (14-23)
change in counts of five consecutive pixels
14 The Gamma Camera: Performance Characteristics 231