Brain morphometry and normative modeling with ScanOMetrics

Affil. ist Inselspital

Morphometric analysis of brain MRI has significantly advanced our understanding of both healthy and pathological aging. However, translating research-grade morphometry tools into clinical practice demands time-efficient robust structural MRI reconstruction pipelines. In this context, deep learning (DL)-based methods have emerged as the leading approach for whole-brain segmentation and parcellation. In this work, we introduce ScanOmetrics, an open-source, research-grade tool designed to detect statistical anomalies in individual MRI scans using both surface- and voxel-based metrics. The tool leverages DeepSCAN, a deep learning-based brain segmentation and cortical parcellation method, as input to an 11-minute surface reconstruction pipeline adapted from FreeSurfer. We present the first validation results of ScanOmetrics, benchmarking its performance against standard FreeSurfer outputs. The analysis was focuses on OASIS3, a large and freely available dataset containing clinical grade high-resolution isotropic T1-weighted MRI scans of patients with Alzheimer’s disease (AD) and healthy controls (HC). We demostrate that cortical thickness anomalies in patient scans were mainly detected in regions that are known as predilection areas of cortical atrophy in AD. In contrast, anomaly detections in HCs were up to twenty-fold reduced and spatially unspecific. Progression of the atrophy pattern with clinical dementia rating (CDR) was clearly observable. Notably, ScanOmetrics delivered results in under 25 minutes, over 15 times faster than FreeSurfer.