Construction of binding potential maps of Pittsburgh Compound B by bolus-plus-infusion transformation (BPIT): Comparison to popular reference tissue approaches

Background and aims: Pittsburgh Compound B ([11C]PIB), an amyloid imaging agent, is being investigated in dementia and aging research. Construction of accurate binding potential (BP) maps is needed as voxel-wise statistical approaches are increasingly employed for detecting regional increases in amyloid deposition. We tested an approach to transfer bolus-only dynamic PET experiments to bolus-plus-infusion experiments in theory such that BP was estimated as voxel-to-reference region ratio minus 1 in a hypothetical steady-state (BPIT). Methods: Thirty-four Baltimore Longitudinal Study of Aging participants (age: 79.9 ± 7.6) underwent one 90-min dynamic PET scan following bolus injection of PIB (15mCi). Participants had normal cognition (n=28) or mild cognitive impairment (informant-based clinical dementia rating of 0.5). Volumes of interest (VOIs) were defined manually on a spoiled-gradient MRI for putamen, caudate nucleus, thalamus, and cerebellum. A standard VOI template, including 28 VOIs for cortical and subcortical structures, was adjusted to each individual's MRI using SPM2 spatial normalization. VOIs were transferred to PET space according to MRI-PET coregistration parameters from SPM2. BP maps of PIB were generated by BPIT, reference tissue graphical analysis (RTGA; Logan et al., 1996), and multi-linear reference tissue method with 2 parameters (MRTM2; Ichise et al., 2002) using the cerebellum as the reference region. Regional BP values were obtained by applying the VOIs on BP maps (Map analysis). Separately, regional time-activity curves (TACs) were obtained by applying the VOIs on dynamic PET frames to estimate regional BP values for each of the three approaches (TAC analysis). Results: Regional BP values of the three Methods: agreed for the TAC analysis (BPIT (=y) vs. RTGA (=x): y = 0.991?x + 0.00288, r2=0.992; BPIT vs. MRTM2: y = 0.995?x + 0.000147 r2=0.988). In the comparison of map (=y) vs. TAC (=x) analyses, BPIT showed no deviations of regional BP values (y = 1?x + 0.0000, r2=1.000) while RTGA showed bias and greater variability (y = 0.83?x + 0.0629, r2=0.928). MRTM2 displayed minimal bias and variability in this comparison (y = 0.989?x + 0.00353, r2=0.993). Computational times for constructing BP maps were 19.2 ± 1.4 seconds for BPIT, 26.5 ± 2.0 seconds for RTGA, and 586.3 ± 15.5 seconds for MRTM2 using a 3.4 MHz Pentium 4 PC. No outliers were observed (voxel values 10 times greater than mean of 6 surrounding voxels) for BPIT and RTGA but 34.5 ± 1.9 % of voxels (range: 30.6 ? 39.6%) were outliers with MRTM2 although the majority was outside the brain. Conclusions: This study validated the BPIT approach for obtaining BP of PIB scans against widely used RTGA and MRTM2 approaches on the basis of the TAC analysis. BPIT suffered no bias and variability in the map vs. TAC comparisons, which was a substantial improvement over RTGA. BPIT and RTGA displayed no outliers but MRTM2 suffered noticeable numbers of outliers. BPIT substantially shortened image construction times. Thus, BPIT appears advantageous over RTGA and MRTM2 in constructing BP maps for PIB scans in aging and dementia research.