MRI-based hippocampal volume is an important biomarker of neurodegeneration in Alzheimer’s disease, but its reliability can be compromised by differences in scanners, acquisition protocols and image quality—particularly in multicentre studies and routine clinical practice.

Cette étude introduit une simple and interpretable quality-control approach to identify 3D T1-weighted MRI acquisition sequences associated with poor test-retest stability of hippocampal volumetry. The analysis included 446 MRI scans acquired with 96 scanners across 76 sites, with hippocampal volumes assessed using 18 different automatic volumetry tools.

Remarquablement, just two readily interpretable image characteristics—left-right field-of-view and contrast-to-noise ratio—were sufficient for the decision-tree model, which achieved 79.5% accuracy in distinguishing acquisition sequences with good versus poor volumetric stability.

Le projet approach increased the ratio of good-to-poor sequences from 3.5 to 7.4, and its benefit was largely independent of the volumetry software used.

En fournissant un easy-to-use, software-independent MRI quality-control strategy, this work could improve the reproducibility of hippocampal volumetry and strengthen quantitative MRI biomarkers in Alzheimer’s disease research, multicentre studies and longitudinal monitoring.