IcoConv : Explainable brain cortical surface analysis for ASD classification.
May 7, 2024
This study presents a unique method for analyzing 3D shapes through 2D CNNs with custom pooling, aimed at identifying high-risk ASD subjects by analyzing brain measurements mapped onto a sphere, featuring a novel icosahedron convolution operator and generating explainability maps that corroborate and potentially expand on existing ASD research.
