High-Frequency Spatial Feature Fusion with 3D CNN for Early Stage Schizophrenia Classification
Kamil Akhtar, Amirthavarshani Mahadevan
The paper develops a 3D CNN with high-pass spatial filtering and Laplacian initialization to classify early-stage schizophrenia using fMRI data, achieving 75.54% test accuracy.
bioRxiv · Jun 19, 2026 · PreprintNeuroscience