TY - JOUR T1 - Template based PiB PET-image analysis: Template driven variability JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 551 LP - 551 VL - 51 IS - supplement 2 AU - Charles Laymon AU - Bedda Rosario AU - Andrew Redfield AU - Michael Berginc AU - William Klunk AU - Chester Mathis AU - Rebecca Mcnamee AU - Julie Price Y1 - 2010/05/01 UR - http://jnm.snmjournals.org/content/51/supplement_2/551.abstract N2 - 551 Objectives Based on large PiB (a PET amyloid tracer) study volumes, there is a growing use of automated methods of analysis, e.g., template based procedures for defining regions of interest (ROIs). We examine the variability introduced into such an analysis based on template choice. Methods Using a defined methodology, ROIs for 14 regions of the brain were drawn on each of 12 MR scans (T1-T12) that served as templates. Six (T1-T6) healthy control (HC) and 6 (T7-T12) Alzheimer dementia (AD) scans were so utilized. The 12 templates and associated ROIs were separately warped to the native MR space of each of 58 HC and 16 AD subjects using SPM8. Regional PiB SUVs were extracted from subjects’ coregistered PET scans using the warped template ROIs resulting in 12 SUVs for each ROI per subject. For each subject and region, the SUV coefficient of variation (COV) was calculated using 3 separate MR template ensembles: T1-T6, T7-T12, and T1-12. SUV and COV averages were computed over all AD and all HC subjects for each ensemble over 4 primary cortical (i.e., SUVcort, COVcort) areas [anterior cingulate, frontal cortex, parietal, precuneus]. Results We find for each of 3 analyses for AD subjects: 1) T1-T6 templates: SUVcort=1.69, COVcort= 3.6% ; 2) T7-T12 templates: SUVcort=1.65, COVcort= 4.2%; 3) T1-T12 templates: SUVcort=1.67, COVcort=4.4%. For each of 3 analyses for HC subjects: 1) T1-T6 templates: SUVcort=0.92, COVcort= 4.4%; 2)T7-T12: SUVcort=0.89, COVcort= 4.0%; 3) T1-T12: SUVcort=0.90, COVcort= 4.5%. Similar results were observed for the cerebellum (a reference region for nondisplaceable uptake). Conclusions For AD and HC subjects, the SUVcort measure of PiB retention was relatively insensitive to choice of template ensemble (T1-T6, T7-T12, or T1-12). Interclass variability (percent difference in SUVcort between ensembles) was generally less than the intraclass variability (COVcort). This work validates the use of template-based procedures and suggests that template averaging techniques could be used to increase reliability. Research Support NIH AG024904-0 ER -