Optimized pipeline for voxel-wise detection of T1 relaxation time abnormalities

Motivation: Quantifying voxel-wise deviations in T1 relaxation times from a normative atlas enables the characterization of brain microstructural pathology with high spatial resolution. So far, registration errors and long runtimes have constrained its clinical adoption. Goal(s): Provide accurate T1 deviation maps for clinical use, with decreased runtimes, reduced registration errors, and enhanced interpretation using noise filtering. Approach: An existing algorithm was improved by extending the normative cohort with multi-centric data, replacing the registration method with a deep-learning model, and implementing a novel iterative noise filtering technique. Results: This approach substantially reduced processing time, improved spatial alignment, and reduced measurement noise in T1 deviation maps. Impact: Improved spatial alignment and a novel iterative noise filtering technique enhanced efficiency, accuracy and interpretability of a pipeline characterizing voxel-wise abnormalities of quantitative T1 relaxation times, paving the way for clinical adoption of pathology characterization with single-subject T1 deviations.