J Neural Transm (1992) [Suppl] 36: 3-12
© Springer-Verlag 1992
Critical review of PET, SPECT and neuroreceptor studies in schizophrenia T. F. Budinger Lawrence Berkeley Laboratory, University of California, Berkeley, CA, U.S.A.
Hypotheses having to do with the neurochemical basis for mental disease can be tested by noninvasive measurement of flow, metabolism and neuroreceptor concentrations in various parts of the brain. The capabilities of these technologies for accurate absolute quantitation are limited by the instrument resolution and the limited statistics associated with allowable radiation doses. Nevertheless, very important new information regarding the brain function in normal and diseased states can be gleaned from the techniques of positron emission tomography (PET) and single photon emission tomography (SPECT). This chapter presents the critical problems which face the clinical researcher who endeavors to test hypotheses regarding mental illnesses using PET or SPECT. First, the quantitative capabilities of emission tomography will be discussed; then the special requirements of SPECT for attenuation correction will be reviewed. Next, the physiological principles required for inter-institutional comparisons are presented.
Summary.
Resolution and quantitation
Commercial PET and SPECT instruments have the ability to give valid numbers for flow, metabolism, and receptor densities only in 10 mm sized regions of the brain. Whereas the stated resolution of a PET instrument might be 6 mm, the quantitative recovery of data in regions smaller than the thickness of the cortex (e.g. 4mm) will be less than 50% due to the fact that the limited resolution will result in the activity being spread over a region larger than the cerebral cortex, for example (Budinger et al., 1984). This spreading or blurring of activity results in a lowering of the concentration giving rise to erroneous values for flow, metabolism, and receptor density (Fig. 1). An accurate estimate of the concentration of fluorodeoxyglucose in the cortex cannot be expected unless the instrument resolution is better than the thickness of the cerebral cortex. For example. the error in glucose metabolism can be greater that 30% if the measurements are made with an instrument with a resolution of 8 mm as shown in Fig. 2. A region of
A
10 5 Events
10 7 Events
Low Resolut ion
M
Hi gh
1llt
Reso l ution
Fig. 1. Theoretical expectation of the spread of activity from an system which have a resolution comparable to the size of the actual activity distribution
RELATION BETWEEN RESOLUTION AND GMRglu OR FLOW ACCURACY
GMRgl u fx
10.0 x 0.7 3.0 x 0.3
High Resolution
True Observed
7.0
2....2 7.9
= 1.O..Jl 7. 9
= 1.3
Low Resolution
Fig. 2. Comparison between the imaging capabilities and the quantitative information for two PET instruments. Upper shows the results on the same patient as the lower image
T. F. Budinger: PET, SPECT and neuroreceptor studies in schizophrenia
5
interest over the low resolution image is shown to include both white and gray matter. Though these facts severely limit the absolute quantitative ability of PET or SPECT, the limitations do not invalidate the techniques for measurements of relative concentrations - metabolism, receptor concentrations, and flow. When comparing one area of the cortex to another, the relative concentration is generally a reliable estimate. Indeed, if the activity adjacent to the gray matter cortical ribbon is low and the tracer is distributed uniformly in a cerebral cortical area, it is possible to correct the PET or SPECT data based on a priori anatomical information provided by NMR. In general, when examining large structures such as the striatum, the absolute quantitation will be reliable. If the resolution allows, the concentration within the putamen, for example, can be quantitated but it is not possible to achieve accurate ratios of concentration between portions of the putamen or hippocampus with the resolution of available commercial instruments. Two areas of contemporary interest provide examples of the problem. It is of interest to evaluate the metabolism of the hippocampus in schizophrenia. Due to the limited size of the hippocampus relative to the resolution of available instruments, one must view with extreme caution even the relative concentrations of activity left to right and anterior to posterior because we can expect errors not only due to the blurring of activity but also due to partial volume contributions which are related to the slice thickness and position of the tomographic slices relative to the hippocampus. A second aspect of research in schizophrenia which cannot be pursued easily with the present instruments is the evaluation of the balance of neuroreceptors between different parts of the cerebral cortex and between the cerebral cortex and striatum. Both the dopamine and serotonin systems exist in the cerebral cortex and hippocampal gyri. These neuroreceptors are found within thin layers of the cortex less than 1 mm thick. Considerinb the specific to nonspecific accumulation of ligands, the limitations of 6 mm resolution instruments in detecting concentrations in 1 mm regions and the statistical noise associated with emission tomography, it is unlikely that we will be able to explore the question of neuroreceptor relative concentration with instruments having the present resolution. Let us examine the determinants of resolution for a PET device. The general concept of PET detection shown in Fig. 3 shows that the first determinant of resolution is the detector size. Because of the geometry of the detection of annihilation photons at 180 degrees, the geometric resolution is 1/2 the detector thickness. Thus, for a detector of 3 mm width, we expect 1.5 mm resolution. But there are two other physical factors which limit resolution (Fig. 3). The angulation error of 0.25 degrees at the time of photon production through the annihilation of a positron with an electron can cause a error of a few mm in the reconstructed image. The positron leaves the nucleus with kinetic energy which means it travels a short distance in tissue and this range effect can seriously affect resolution if the
6 POSItron annihilation photons 1800
_
0.25 0
v
Fig. 3. Positron emission tomography concept
20 r - - - - - - - - -
16
E
12
.a
8
.§. .~
~ a:
4
The evolution of resolution.
Year
Fig. 4. Factors affecting resolution
kinetic energy is high as is the case for some positron emitters (e.g. iodine122, iodine-124, rubidium-82) ; however, this effect is not as serious as some have speculated as it results in a resolution deterioration of less than 1 mm (Fig. 4). Important image resolution over the last 15 years is shown in Fig. 5. (Budinger, 1990). The Donner 600-crystal tomograph was built to demonstrate the possibilities of achieving 2.6 mm resolution using 3 mm crystals. Results from this instrument in the measurement of deoxygiucose concentration in Alzheimer's are shown in Fig. 6.
PET, SPECT and neuroreceptor studies in schizophrenia FACTOR
a.
7
CONTRIBUTION
~ ------=--------""-I:=Jt 3.0 mm
~:,
D e tector G ometry
A
1.5 mm
I O·±.2S·
b.
1.3 mm
D
0.5 mm Range Theoretical R
(a 2 +b2
solution C2)~
2.0 mm
Resolution after Recon s truction
Calculated as 2.35 x rms deviations for
2.6 mm
18F
Fig. 5. Improvement in resolution over the past 15 years. Arrow indicates point sources 4 mm apart PET 600 Control
AD
Fig. 6. PET images of a control and Alzheimer's patient
It is now believed to be possible to achieve 2 mm resolution using a new detector design (Moses et aI., 1992). However, even with an instrument with a resolution of 2 mm, we would not expect an accurate concentration value for neuroreceptors because the neuroreceptors might reside in layers
8
T. F. Budinger
thinner than 1 mm. The requirement for increased sensitivity in order to achieve sufficient statistics to give reliable data with 2 mm resolution can be satisfied by a new design approach which involves new detectors just now under development and time-of-flight implementation to improve the signal-to-noise. Sensitivity
Sensitivity is defined as the number of detected good events of the tomographic system (resolution) per a given source activity in an attenuating medium. The sensitivity of a PET system is dependent on the detector size and efficiency. A typical sensitivity of a commercial PET system is 15,000 events per second per J..I.Ci in a 20 cm diameter attenuator for 1 cm axial resolution. This rather lengthy characterization of sensitivity is needed in order to be able to compare one system to another and to be able to predict feasibility of new types of quantitative studies. The major determinant of sensitivity of SPECT systems is the amount of detector material surrounding (usually how many cameras) the subject and the resolution which is provided by the collimator. The better the resolution, the lower the sensitivity. SPECT systems have less sensitivity than PET by the factor given in the equation below: PET _ 150 (mm) d(mm) SPECT
(1)
where d is the resolution in millimeters. This equation assumes an axial section 1 cm thick. The major reason for this difference in sensitivity lies in the need to collimate the photons for the SPECT single photon radionuclides. The collimation for PET radionuclides in given by the electronic acceptance of only those photons which arrive in time coincidence - the assumption is that they arose from a source located on a line between the detectors which registered the detection (Fig. 4). Not only is SPECT convenient for blood flow studies wherein the resolution is adequate to give numbers reflecting blood flow in one part of the cortex relative to another, but SPECT does provide an opportunity to evaluate the receptor activities in the striatum because of the availability of 123I-Iabeled ligands specific for the D2 and 5-HT2 systems. Attenuation compensation
A critical aspect of quantitative SPECT which distinguishes this technique as potentially less quantitative than PET is the need for special manipulations of the data during reconstruction. Both in PET and SPECT, the photons are attenuated by the tissues between the detectors and the source.
PET, SPECT and neuroreceptor studies in schizophrenia
9
M icrosphere Beads
Chemical Trapp ing
Fig. 7. Reduction of the reconstruction activIty at the center of the head can be expected if attenuation correction is not applied in SPECT
The attenuation correction is easy to apply in PET and much more difficult for SPECT. The problem has not been so much in the mathematics, but in the commercial implementation of fast, yet accurate methods. Both computer memory and computational times have been limiting in the past. The problem as illustrated in Fig. 7 is the proper use of attenuation correction mathematics. For the usual radionuclides the head attenuation results in a lowering of the apparent concentration by as much as 300%. Quantitative interinstitutional and logitudinal studies
Finally, the use of practical models which will allow the quantitative comparison between patients, and from time to time in the same patient, demands some measure of the arterial input function which is usually determined by drawing arterial blood every 5-10 seconds. The time rate of change of arterial concentration can also be determined from tomographic data obtained over the heart using dynamic PET or SPECT. The basic model for quantitating the uptake of a tracer which does not wash out of the brain is: FE = Q(T) /
~~ A(t)dt
(2)
T . F. Budinger
10
#=0
\\
~
\',
. II
,n \" 20 -
ORIGINAL
1/
:~
1\
II '
.
"
~il \ \ .....',
IJ. = 0 .05
'11
I
i/· :' ,f
\1 /'
10
.f
._,#=005
.:
1.1
1/
• Y' \.... / '\":,t" "", ::-"'/ ','/.
/1
•..•
\L \ ~iJ. = i.., '\ / \ '-. '" '_J'" 0 11
\ /\ \..
\.
/
!
/"~/ ! '-oJ
., I .-._./ ......:
r-J
\.{ = 0 15
IJ. = 0.11
IJ.
o
= 0.15
23 em
Fig, 8. The correspondence between microspheres method of measuring flow and the use of radiopharmaceuticals which are trapped in the tissues (see Fig. 9 for tracers which are not trapped)
where F is flow, E is extraction, Q is the amount in the tissue at some time , T, and A is the arterial concentration (Budinger and Kuesman, 1985) . We assume that the arterial concentration is measured from an arterial line or determined from dynamic tomographic data taken over the heart during the initial few minutes after the injection. Thus, in this equation Q(T) is spatial distribution of concentration made by moving the head into the tomograph at time T. Note that for micro spheres or ligands that behave like microspheres, E in the above equation is 1 (Fig. 8). More complex models involving interchange between extravascular compartments and receptor systems as well as interchange between extravascular compartments and nonspecific accumulation sites generally require data collection strategies which place severe demands on patients, control subjects and statistical limits of the data, particularly if information from small regions is being sought. Using a model as simple as equation 2 or a slightly more complex model for the partly diffusible tracer (Fig. 9) wherein a wash-out (k2) is significant, reasonably good quantitative values can be determined. In this case flow in (kd and wash-out (k2) are derived from fitting the data from each subregion of the brain to the equation. Q(T) = (k l )
Jo dtA(t)eT
k2(-t-t)
(3)
PET, SPECT and neuroreceptor studies in schizophrenia
11
Diffusible
-
I
'~
0\. . . . .::.
..........
Fig. 9. The two-compartment model for a diffusible tracer such as ISO-H20
In the case of equation 3, Q must be measured simultaneously with A(t) thus requiring arterial sampling by catheter or yet to be developed instrumentation. Activation studies
Finally, much of the work currently under way using activation analysis with PET and H 215 0 relies on the phenomenon modeled in equation 3. Most of the information being derived on the spatial localization of cognitive processes evaluate brain image information collected by integrating Q(t) with the assumption that any changes of the arterial activity behavior A( t) will be normalized by the process of rationing or subtracting the integrated images for the conditions of control vs. activation. An alternate approach using a non-diffusible flow tracer 22 I HIPDM; T1I2 = 3.6min) and the model of equation 2 can give accurate results without the need to measure the input function if relative differences are being sought.
e
Conclusion
Major advances in the use of noninvasive techniques in the study of schizophrenia can be made if the restrictions of resolution and statistics are kept in mind when analyzing data which has been properly corrected for attenuation and scatter. The present resolution of commercial instruments prohibits accurate measurements of the relative concentrations of neuroreceptors between the cortex and mid-brain gray matter. The expectations for very high resolution PET instruments can be realized to the extent that structures as small as the locus ceruleus and inner vs. outer cerebral cortex can be evaluated. This will require 2 mm resolution and image acquisition times in excess of 15 min. In the interim, studies which report flow, metabolism
12
T. F. Budinger: PET, SPECT and neuroreceptor studies in schizophrenia
and neuroreceptor concentrations in mental disorders vs. normal controls will be improved in their contribution if the limitations of resolution are addressed frontally. Some cautions regarding inferences from radiopharmaceuticals
The importance of specificity of the ligand is made clear when examining the cross specificity of compounds such as methylspiperone to both the 5HT2 and the dopamine D2 system. On the other hand, a highly specific ligand for the D2 system such as raclopride might not be an ideal tracer of neuroreceptors because the low affinity ligand might be replaced in vivo by endogenous compounds such as dopamine itself. Thus, the evaluation of the D2 system requires the careful attention to the affinity of the ligand as well as its specificity. A parallel argument can be made for selection of flow tracers. Whereas, some cerebral blood flow tracers are readily available, they might have a metabolic component which will lead to inaccurate flow results. References Budinger TF (1990) Advances in emission tomography instrumentation: Quo Vadis? J Nucl Med 31:628-631 Budinger TF, Huesman RH (1985) Ten percepts for quantitative data acquisition and analysis. Circulation 72:IV-53- IV-62 Budinger TF, Derenzo SE, Huesman RH (1984) Instrumentation for positron emission tomography. Ann Neurol 15:§35-§43 Moses WW, Derenzo SE, Budinger TF (1992) Design for a high-rate, high resolution PET module using room-temperature silicon photodiodes for crystal identification. J Nucl Med 33 Author's address: Dr. T. F. Budinger, The Lawrence Berkeley Laboratory, University of California, MS 55-127, Berkeley, CA 94720, U.S.A.
© Springer-Verlag 1992
Critical review of PET, SPECT and neuroreceptor studies in schizophrenia T. F. Budinger Lawrence Berkeley Laboratory, University of California, Berkeley, CA, U.S.A.
Hypotheses having to do with the neurochemical basis for mental disease can be tested by noninvasive measurement of flow, metabolism and neuroreceptor concentrations in various parts of the brain. The capabilities of these technologies for accurate absolute quantitation are limited by the instrument resolution and the limited statistics associated with allowable radiation doses. Nevertheless, very important new information regarding the brain function in normal and diseased states can be gleaned from the techniques of positron emission tomography (PET) and single photon emission tomography (SPECT). This chapter presents the critical problems which face the clinical researcher who endeavors to test hypotheses regarding mental illnesses using PET or SPECT. First, the quantitative capabilities of emission tomography will be discussed; then the special requirements of SPECT for attenuation correction will be reviewed. Next, the physiological principles required for inter-institutional comparisons are presented.
Summary.
Resolution and quantitation
Commercial PET and SPECT instruments have the ability to give valid numbers for flow, metabolism, and receptor densities only in 10 mm sized regions of the brain. Whereas the stated resolution of a PET instrument might be 6 mm, the quantitative recovery of data in regions smaller than the thickness of the cortex (e.g. 4mm) will be less than 50% due to the fact that the limited resolution will result in the activity being spread over a region larger than the cerebral cortex, for example (Budinger et al., 1984). This spreading or blurring of activity results in a lowering of the concentration giving rise to erroneous values for flow, metabolism, and receptor density (Fig. 1). An accurate estimate of the concentration of fluorodeoxyglucose in the cortex cannot be expected unless the instrument resolution is better than the thickness of the cerebral cortex. For example. the error in glucose metabolism can be greater that 30% if the measurements are made with an instrument with a resolution of 8 mm as shown in Fig. 2. A region of
A
10 5 Events
10 7 Events
Low Resolut ion
M
Hi gh
1llt
Reso l ution
Fig. 1. Theoretical expectation of the spread of activity from an system which have a resolution comparable to the size of the actual activity distribution
RELATION BETWEEN RESOLUTION AND GMRglu OR FLOW ACCURACY
GMRgl u fx
10.0 x 0.7 3.0 x 0.3
High Resolution
True Observed
7.0
2....2 7.9
= 1.O..Jl 7. 9
= 1.3
Low Resolution
Fig. 2. Comparison between the imaging capabilities and the quantitative information for two PET instruments. Upper shows the results on the same patient as the lower image
T. F. Budinger: PET, SPECT and neuroreceptor studies in schizophrenia
5
interest over the low resolution image is shown to include both white and gray matter. Though these facts severely limit the absolute quantitative ability of PET or SPECT, the limitations do not invalidate the techniques for measurements of relative concentrations - metabolism, receptor concentrations, and flow. When comparing one area of the cortex to another, the relative concentration is generally a reliable estimate. Indeed, if the activity adjacent to the gray matter cortical ribbon is low and the tracer is distributed uniformly in a cerebral cortical area, it is possible to correct the PET or SPECT data based on a priori anatomical information provided by NMR. In general, when examining large structures such as the striatum, the absolute quantitation will be reliable. If the resolution allows, the concentration within the putamen, for example, can be quantitated but it is not possible to achieve accurate ratios of concentration between portions of the putamen or hippocampus with the resolution of available commercial instruments. Two areas of contemporary interest provide examples of the problem. It is of interest to evaluate the metabolism of the hippocampus in schizophrenia. Due to the limited size of the hippocampus relative to the resolution of available instruments, one must view with extreme caution even the relative concentrations of activity left to right and anterior to posterior because we can expect errors not only due to the blurring of activity but also due to partial volume contributions which are related to the slice thickness and position of the tomographic slices relative to the hippocampus. A second aspect of research in schizophrenia which cannot be pursued easily with the present instruments is the evaluation of the balance of neuroreceptors between different parts of the cerebral cortex and between the cerebral cortex and striatum. Both the dopamine and serotonin systems exist in the cerebral cortex and hippocampal gyri. These neuroreceptors are found within thin layers of the cortex less than 1 mm thick. Considerinb the specific to nonspecific accumulation of ligands, the limitations of 6 mm resolution instruments in detecting concentrations in 1 mm regions and the statistical noise associated with emission tomography, it is unlikely that we will be able to explore the question of neuroreceptor relative concentration with instruments having the present resolution. Let us examine the determinants of resolution for a PET device. The general concept of PET detection shown in Fig. 3 shows that the first determinant of resolution is the detector size. Because of the geometry of the detection of annihilation photons at 180 degrees, the geometric resolution is 1/2 the detector thickness. Thus, for a detector of 3 mm width, we expect 1.5 mm resolution. But there are two other physical factors which limit resolution (Fig. 3). The angulation error of 0.25 degrees at the time of photon production through the annihilation of a positron with an electron can cause a error of a few mm in the reconstructed image. The positron leaves the nucleus with kinetic energy which means it travels a short distance in tissue and this range effect can seriously affect resolution if the
6 POSItron annihilation photons 1800
_
0.25 0
v
Fig. 3. Positron emission tomography concept
20 r - - - - - - - - -
16
E
12
.a
8
.§. .~
~ a:
4
The evolution of resolution.
Year
Fig. 4. Factors affecting resolution
kinetic energy is high as is the case for some positron emitters (e.g. iodine122, iodine-124, rubidium-82) ; however, this effect is not as serious as some have speculated as it results in a resolution deterioration of less than 1 mm (Fig. 4). Important image resolution over the last 15 years is shown in Fig. 5. (Budinger, 1990). The Donner 600-crystal tomograph was built to demonstrate the possibilities of achieving 2.6 mm resolution using 3 mm crystals. Results from this instrument in the measurement of deoxygiucose concentration in Alzheimer's are shown in Fig. 6.
PET, SPECT and neuroreceptor studies in schizophrenia FACTOR
a.
7
CONTRIBUTION
~ ------=--------""-I:=Jt 3.0 mm
~:,
D e tector G ometry
A
1.5 mm
I O·±.2S·
b.
1.3 mm
D
0.5 mm Range Theoretical R
(a 2 +b2
solution C2)~
2.0 mm
Resolution after Recon s truction
Calculated as 2.35 x rms deviations for
2.6 mm
18F
Fig. 5. Improvement in resolution over the past 15 years. Arrow indicates point sources 4 mm apart PET 600 Control
AD
Fig. 6. PET images of a control and Alzheimer's patient
It is now believed to be possible to achieve 2 mm resolution using a new detector design (Moses et aI., 1992). However, even with an instrument with a resolution of 2 mm, we would not expect an accurate concentration value for neuroreceptors because the neuroreceptors might reside in layers
8
T. F. Budinger
thinner than 1 mm. The requirement for increased sensitivity in order to achieve sufficient statistics to give reliable data with 2 mm resolution can be satisfied by a new design approach which involves new detectors just now under development and time-of-flight implementation to improve the signal-to-noise. Sensitivity
Sensitivity is defined as the number of detected good events of the tomographic system (resolution) per a given source activity in an attenuating medium. The sensitivity of a PET system is dependent on the detector size and efficiency. A typical sensitivity of a commercial PET system is 15,000 events per second per J..I.Ci in a 20 cm diameter attenuator for 1 cm axial resolution. This rather lengthy characterization of sensitivity is needed in order to be able to compare one system to another and to be able to predict feasibility of new types of quantitative studies. The major determinant of sensitivity of SPECT systems is the amount of detector material surrounding (usually how many cameras) the subject and the resolution which is provided by the collimator. The better the resolution, the lower the sensitivity. SPECT systems have less sensitivity than PET by the factor given in the equation below: PET _ 150 (mm) d(mm) SPECT
(1)
where d is the resolution in millimeters. This equation assumes an axial section 1 cm thick. The major reason for this difference in sensitivity lies in the need to collimate the photons for the SPECT single photon radionuclides. The collimation for PET radionuclides in given by the electronic acceptance of only those photons which arrive in time coincidence - the assumption is that they arose from a source located on a line between the detectors which registered the detection (Fig. 4). Not only is SPECT convenient for blood flow studies wherein the resolution is adequate to give numbers reflecting blood flow in one part of the cortex relative to another, but SPECT does provide an opportunity to evaluate the receptor activities in the striatum because of the availability of 123I-Iabeled ligands specific for the D2 and 5-HT2 systems. Attenuation compensation
A critical aspect of quantitative SPECT which distinguishes this technique as potentially less quantitative than PET is the need for special manipulations of the data during reconstruction. Both in PET and SPECT, the photons are attenuated by the tissues between the detectors and the source.
PET, SPECT and neuroreceptor studies in schizophrenia
9
M icrosphere Beads
Chemical Trapp ing
Fig. 7. Reduction of the reconstruction activIty at the center of the head can be expected if attenuation correction is not applied in SPECT
The attenuation correction is easy to apply in PET and much more difficult for SPECT. The problem has not been so much in the mathematics, but in the commercial implementation of fast, yet accurate methods. Both computer memory and computational times have been limiting in the past. The problem as illustrated in Fig. 7 is the proper use of attenuation correction mathematics. For the usual radionuclides the head attenuation results in a lowering of the apparent concentration by as much as 300%. Quantitative interinstitutional and logitudinal studies
Finally, the use of practical models which will allow the quantitative comparison between patients, and from time to time in the same patient, demands some measure of the arterial input function which is usually determined by drawing arterial blood every 5-10 seconds. The time rate of change of arterial concentration can also be determined from tomographic data obtained over the heart using dynamic PET or SPECT. The basic model for quantitating the uptake of a tracer which does not wash out of the brain is: FE = Q(T) /
~~ A(t)dt
(2)
T . F. Budinger
10
#=0
\\
~
\',
. II
,n \" 20 -
ORIGINAL
1/
:~
1\
II '
.
"
~il \ \ .....',
IJ. = 0 .05
'11
I
i/· :' ,f
\1 /'
10
.f
._,#=005
.:
1.1
1/
• Y' \.... / '\":,t" "", ::-"'/ ','/.
/1
•..•
\L \ ~iJ. = i.., '\ / \ '-. '" '_J'" 0 11
\ /\ \..
\.
/
!
/"~/ ! '-oJ
., I .-._./ ......:
r-J
\.{ = 0 15
IJ. = 0.11
IJ.
o
= 0.15
23 em
Fig, 8. The correspondence between microspheres method of measuring flow and the use of radiopharmaceuticals which are trapped in the tissues (see Fig. 9 for tracers which are not trapped)
where F is flow, E is extraction, Q is the amount in the tissue at some time , T, and A is the arterial concentration (Budinger and Kuesman, 1985) . We assume that the arterial concentration is measured from an arterial line or determined from dynamic tomographic data taken over the heart during the initial few minutes after the injection. Thus, in this equation Q(T) is spatial distribution of concentration made by moving the head into the tomograph at time T. Note that for micro spheres or ligands that behave like microspheres, E in the above equation is 1 (Fig. 8). More complex models involving interchange between extravascular compartments and receptor systems as well as interchange between extravascular compartments and nonspecific accumulation sites generally require data collection strategies which place severe demands on patients, control subjects and statistical limits of the data, particularly if information from small regions is being sought. Using a model as simple as equation 2 or a slightly more complex model for the partly diffusible tracer (Fig. 9) wherein a wash-out (k2) is significant, reasonably good quantitative values can be determined. In this case flow in (kd and wash-out (k2) are derived from fitting the data from each subregion of the brain to the equation. Q(T) = (k l )
Jo dtA(t)eT
k2(-t-t)
(3)
PET, SPECT and neuroreceptor studies in schizophrenia
11
Diffusible
-
I
'~
0\. . . . .::.
..........
Fig. 9. The two-compartment model for a diffusible tracer such as ISO-H20
In the case of equation 3, Q must be measured simultaneously with A(t) thus requiring arterial sampling by catheter or yet to be developed instrumentation. Activation studies
Finally, much of the work currently under way using activation analysis with PET and H 215 0 relies on the phenomenon modeled in equation 3. Most of the information being derived on the spatial localization of cognitive processes evaluate brain image information collected by integrating Q(t) with the assumption that any changes of the arterial activity behavior A( t) will be normalized by the process of rationing or subtracting the integrated images for the conditions of control vs. activation. An alternate approach using a non-diffusible flow tracer 22 I HIPDM; T1I2 = 3.6min) and the model of equation 2 can give accurate results without the need to measure the input function if relative differences are being sought.
e
Conclusion
Major advances in the use of noninvasive techniques in the study of schizophrenia can be made if the restrictions of resolution and statistics are kept in mind when analyzing data which has been properly corrected for attenuation and scatter. The present resolution of commercial instruments prohibits accurate measurements of the relative concentrations of neuroreceptors between the cortex and mid-brain gray matter. The expectations for very high resolution PET instruments can be realized to the extent that structures as small as the locus ceruleus and inner vs. outer cerebral cortex can be evaluated. This will require 2 mm resolution and image acquisition times in excess of 15 min. In the interim, studies which report flow, metabolism
12
T. F. Budinger: PET, SPECT and neuroreceptor studies in schizophrenia
and neuroreceptor concentrations in mental disorders vs. normal controls will be improved in their contribution if the limitations of resolution are addressed frontally. Some cautions regarding inferences from radiopharmaceuticals
The importance of specificity of the ligand is made clear when examining the cross specificity of compounds such as methylspiperone to both the 5HT2 and the dopamine D2 system. On the other hand, a highly specific ligand for the D2 system such as raclopride might not be an ideal tracer of neuroreceptors because the low affinity ligand might be replaced in vivo by endogenous compounds such as dopamine itself. Thus, the evaluation of the D2 system requires the careful attention to the affinity of the ligand as well as its specificity. A parallel argument can be made for selection of flow tracers. Whereas, some cerebral blood flow tracers are readily available, they might have a metabolic component which will lead to inaccurate flow results. References Budinger TF (1990) Advances in emission tomography instrumentation: Quo Vadis? J Nucl Med 31:628-631 Budinger TF, Huesman RH (1985) Ten percepts for quantitative data acquisition and analysis. Circulation 72:IV-53- IV-62 Budinger TF, Derenzo SE, Huesman RH (1984) Instrumentation for positron emission tomography. Ann Neurol 15:§35-§43 Moses WW, Derenzo SE, Budinger TF (1992) Design for a high-rate, high resolution PET module using room-temperature silicon photodiodes for crystal identification. J Nucl Med 33 Author's address: Dr. T. F. Budinger, The Lawrence Berkeley Laboratory, University of California, MS 55-127, Berkeley, CA 94720, U.S.A.
