Multimodal AI predicts multiple dementia neuropathologies from brain MRI

A deep learning framework integrating 3D T1 brain MRI with demographics and genetic data shows promise for predicting six dementia-related neuropathologies, potentially enabling non-invasive profiling, according to an autopsy-confirmed study.

• The model predicts six pathologies including amyloid-β, tau, TDP-43, and alpha-synuclein using structural MRI and clinical/genetic covariates.
• Explainable AI techniques (occlusion sensitivity, Grad-CAM, Integrated Gradients) localized brain regions contributing to predictions, which were compared with voxel-based morphometry.
• Models were trained and validated on autopsy-confirmed neuropathology labels from individuals with antemortem MRI (sample size not reported).

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.