Mapping the brain-web morphology to characterise ageing and cognitive impairment

SHARP maps cosmic-web–like morphology from structural MRI and improves participant-independent MRI age prediction, reducing mean absolute error from 6.04 to 4.71 years in CamCAN.

• Why it matters: Age-related brain changes are hard to quantify geometrically, and cognitive impairment needs features that capture spatial organization beyond standard anatomical measures.
• What they did: SHARP transfers hierarchical cosmic-web classification to MRI, extracting knots, filaments, sheets, and voids at four spectral levels, then combines these with higher-order spatial statistics and anatomy. It is benchmarked against a neural-network baseline using 36 regional measurements, total brain volume, and sex.
• The result: In OASIS-3, age prediction error drops from 4.57 to 4.07 years, and brain-web morphology supports cognitive-status classification by identifying 85% of CDR-positive observations at 86% specificity.

AI summary of the abstract · may contain errors · not medical advice · read the paper.