Synthetic Image-Derived Phenotypes Using Physics-based Simulations: Predicting Variability in Subcortical Volumes

Motivation: MR image-derived phenotypes (IDPs), such as subcortical volumes, vary widely across acquisition protocols, complicating comparisons in large cohort studies. Goal(s): To develop a physics-based simulation pipeline to synthesize IDPs, helping predict how different protocols affect IDP variability. Approach: Given quantitative relaxometry maps and sequence parameters, this abstract proposes a pipeline for synthesizing IDPs using MR physics-based simulations, which outputs both the synthetic T1w images and FSL-segmented deep grey matter (DGM) structure volumes. Results: Simulation experiments preliminary validate the synthesization pipeline. Experiments show synthesised DGM volume variations my be induced by variations in acquisition protocols and varying protocol sensitivities to partial volume effects. Impact: This pipeline is a first step in using image synthesis to generate matched IDPs precisely. It offers a foundation for exploring protocol sensitivity in neuroimaging biomarkers and can serve as a basis for addressing cross-protocol data harmonization challenges.