Tangible advantages of multi-stage brain motion compensation for PET imaging demonstrated in multiple studies

Head motion is a persistent challenge in positron emission tomography (PET) brain imaging, reducing quantitative accuracy, degrading time-activity curves (TACs), and complicating kinetic modeling. We evaluated a fully automated, multi-stage motion compensation framework, COMBRA (Correction of Motion and BRain Alignment), across more than one hundred scans and four independent PET studies using different radiotracers (¹⁸F-MK-6240, ¹¹C-Raclopride, ¹¹C-PBR28, and ¹¹C-Martinostat). COMBRA implements a staged registration and reconstruction pipeline, enabling both inter-frame realignment and intra-frame motion-gated correction. It operates in a PET-driven mode but can incorporate magnetic resonance imaging (MRI)-based motion tracking when available. Across dynamic datasets, COMBRA improved model fitting precision, as reflected by reduced standard error in distribution volume ratio estimates using MRTM2 modeling, and yielded more consistent test-retest outcomes. In static cohort studies, the framework enhanced regional standardized uptake value ratios, improved image contrast, and mitigated motion-induced bias. Compared to frame-to-frame realignment and MRI-guided approaches, PET-only COMBRA demonstrated equivalent or superior performance, with misalignments reduced to sub-voxel levels. Importantly, its static reconstructions provided better signal-to-noise properties and more robust quantification in high-motion cases. At the population level, COMBRA increased statistical power, suggesting that smaller sample sizes may be sufficient for future studies. Collectively, these results demonstrate that COMBRA offers a scalable, tracer-independent solution for motion compensation in brain PET imaging. By improving accuracy, reliability, and sensitivity, this framework has the potential to strengthen both mechanistic research studies and clinical investigations where subtle group differences are critical.