Standardisering, bearbetning och hierarkisk kvantifiering av [11C]PBR28-positronemissionstomografidata

This Master’s thesis addresses challenges in Positron Emission Tomography (PET) quantification of neuroinflammation using the second-generation radiotracer [11C]PBR28 through a dual approach: (1) standardising historical datasets into the Brain Imaging Data Structure (BIDS) to ensure reproducibility, and (2) evaluating the Simultaneous Multifactor Bayesian Analysis (SiMBA) framework with two-tissue compartmental models (2TCM and 2TCM-1K), marking the first application of SiMBA to [11C]PBR28 PET data, and the first implementation of the 2TCM-1K model for low-affinity binder (LAB) data using a second-generation radiotracer. Data from 56 healthy individuals (29 HABs, 23 MABs, 4 LABs) were converted to BIDS using pipelines (dcm2niix, PET2BIDS, sMRIprep), enabling standardised analysis. While conventional non-linear least squares (NLS) modeling confirmed genotype-driven gradients (HAB > MAB > LAB) and revealed a 38-45% reduction in total distribution volume VT with 2TCM-1K, the hierarchical Bayesian implementation (SiMBA) for 2TCM showed underfitting, and 2TCM-1K variants failed to converge. The work establishes reusable BIDS-formatted datasets and demonstrates the necessity of vascular trapping correction for accurate TSPO quantification, revealing persistent genotype-specific challenges in Bayesian kinetic modeling for the second-generation radiotracer [11C]PBR28.