Browse Teach Curate

Explainable Brain Age Gap Estimation with Covariance Neural Networks in Neuroimaging

Uses coVariance Neural Networks (graph neural networks whose graph is the sample covariance matrix, provably equivalent to PCA) on structural-MRI cortical-thickness features to estimate the brain age gap and characterize distinct anatomic patterns across neurodegenerative conditions, with explainability derived from how the network weights eigenvectors of the anatomic covariance matrix.

Data Text

EXPLAINABLE BRAIN AGE GAP PREDICTION IN NEURODEGENERATIVE CONDITIONS USING COVARIANCE NEURAL

A coVariance Neural Network (VNN) is a graph neural network whose graph is the sample covariance matrix of the data; its convolution is a coVariance filter H(C)=sum_k h_k C^k with learned taps, which…