In the current year we have refined our data analysis methods for estimation of rates of cerebral protein synthesis (rCPS) with L-1-C-11leucine and PET. Because rCPS tends to change only modestly under different physiological conditions, it is essential that the analyses achieve the highest possible accuracy. We also examined the impact of simplifying patient scanning protocols on estimation of rCPS. (1) Refinement of kinetic model analysis methods. Quantification with PET tracers utilizes compartmental models that generally assume homogeneity for each tissue region-of-interest (ROI) with respect to relevant physiological variables. The kinetic model of the C-11leucine PET method was originally applied to ROI data with the assumption that each ROI is homogeneous with respect to concentrations of amino acids, blood flow, transport and metabolism of amino acids, and rCPS. Due to the limited spatial resolution of PET, however, most regions contain a kinetically heterogeneous mixture of tissue components; this may bias estimation results. We have been refining our analyses to minimize effects of tissue heterogeneity on estimates of kinetic parameters and rCPS. We reasoned that a substantial reduction in the volume of a tissue region should reduce the impact of heterogeneity. We developed a method to apply the homogeneous tissue model to analysis of PET data at the voxel level, the Basis Function Method (BFM). We found that voxel-level BFM estimates of rCPS averaged over a ROI were substantially less biased than estimates based on direct fitting of the ROI time-activity curve (TAC) with a homogeneous tissue model. Model fits of the TACs showed that effects of tissue heterogeneity had been reduced, but not eliminated. We developed a second approach that explicitly takes heterogeneity within a ROI into account, spectral analysis with an iterative filter (SAIF). When applied to ROI-level data, SAIF-ROI produced low bias, low variance estimates of rCPS. It performed comparably to the voxel-level BFM method when count rates are normal, but at low count rates it performed better. Although SAIF allows for heterogeneity in a ROI, it does require an assumed constraint on the relationship among the kinetic parameters within the heterogeneous tissues. This has the most impact when kinetics of the various tissues within the ROI are most dissimilar. Under the premise that the dissimilarity among the tissues would be less at the voxel level, we extended the SAIF method for analysis of voxel-level data. In normal count rate studies, rCPS estimated with SAIF-voxel was approximately 5-15% higher than with SAIF-ROI analysis; inter-subject variability was comparable. Based on simulation studies we conclude that the difference is predominantly due to underestimation of rCPS with SAIF-ROI, i.e., the performance SAIF-voxel is better. As a final step, we investigated which data analysis method is optimal under various PET scanning conditions. When scanning with the high resolution research tomograph (HRRT), we found that voxel-level BFM and voxel-level SAIF analyses yielded comparable estimates of rCPS at both high and low count rates. However, when scanner resolution is degraded, SAIF-voxel performs better. We conclude that future C-11leucine studies on the high resolution PET scanner can be analyzed with the computationally simpler voxel-level BFM. (2) Improvement of PET reconstruction algorithms. Reconstruction algorithms are necessary to create three dimensional images of radioactivity distribution from positron emission data collected during the scanning process. PET scanners measure radioactivity in an object by detecting the two annihilation gamma rays emitted when a positron decays. The annihilation gammas are emitted simultaneously at 180 degrees from each other, and the scanner records the event in two of its detectors. The scanner cannot identify exactly where the decay event occurred, only that it occurred somewhere on a line between the two detectors. By combining data from millions of lines of response between detector pairs, an image of the original 3D distribution of radioactivity can be calculated with a reconstruction algorithm. If the object moves during the scanning process the image will be degraded and attenuation corrections applied during the reconstruction procedure will be in error. At the NIH, motion during PET scans of the brain on the HRRT has been addressed by continuously recording head position throughout the scanning procedure with the Polaris optical tracking system. This system has an accuracy of 0.5-0.6 mm; its recorded position data are used for motion correction of the emission data during image reconstruction. This motion correction has been shown to be sufficient for subjects who can lie fairly still in the scanner and therefore do not make large or abrupt motions during scanning. We found, however, that the motion correction is not sufficient for some of our subjects, particularly those who are asleep when the scan is initiated and who startle awake causing large and abrupt motion later in the scanning interval. Inadequate motion correction in the data leads to corrupted time courses of tissue activity and potentially large errors in estimated rates of cerebral protein synthesis. We collaborated with physicists in the NIH PET Department to test a new reconstruction procedure that provides additional motion and attenuation corrections. We analyzed C-11leucine PET data acquired in subjects with large abrupt motion during scanning and in subjects without such motion. Motion was evaluated by examining Polaris recordings of three dimensional translations and rotations of the head during scan acquisition. PET emission data were reconstructed by means of the standard Polaris-based motion correction only or the new procedure that introduces additional motion and attenuation corrections. Effectiveness of the new reconstruction procedure was assessed visually by comparing images of radioactivity distribution in brain during different intervals after administration of C-11leucine; substantial failures of motion correction lead to systematic shifting of the images between early and late intervals of the scan. Smaller failures of motion correction can be seen by examining TACs in selected ROIs. We found that the new reconstruction procedure had negligible impact on the scans in which motion was small (no image degradation was detected), but made substantial improvement in the scans with large abrupt motions. (3) Simplification of scanning protocols. One of our goals is to be able to study subjects with intellectual disabilities without the need for sedation, so any changes in the procedure that make the study less onerous for subjects may help us to achieve that goal. The C-11leucine method requires arterial blood sampling and has a 90 min scan time. We investigated the effects of shortening the scan duration to 50-60 min and found that rCPS values were unaffected. We also examined the effects of substituting venous for arterial blood sampling. We use a population-derived arterial input function scaled by the subjects' own measured venous data. Preliminary analyses of 28 scans supported the feasibility of utilizing venous in place of arterial blood sampling. We have now expanded the number of scans and conditions in which we have tested this hypothesis. We compared rCPS values determined with measured arterial blood data vs rCPS based on venous-rescaled population input functions in an additional 22 scans; we found no statistically significant differences in rCPS in either whole brain or any of the 16 regions examined in any group. Based on these findings, we conclude that the simplified scanning protocols for the C-11leucine PET method can be used without compromising accuracy of rCPS measurements.
