What problem does it solve? It provides model-level guidance for running classical, non-deep-learning disease classification on neuroimaging data, producing sparse and interpretable voxel-wise weight maps instead of opaque deep learning predictions. ## Core Features & Use Cases - Case-Control Classification: Train supervised SpaceNet models to discriminate patient versus control groups from aligned subject-level voxel maps. - Interpretable Outputs: Export predictions, decision scores, cross-validation metrics (accuracy, AUC, sensitivity, specificity), and coefficient maps highlighting discriminative brain regions. - Delegated Workflow: Coordinates with fmri-skill and smri-skill for preprocessing and nilearn-tool for concrete SpaceNet fitting. - Use Case: Given a moderate-sized cohort of aligned fMRI or sMRI images with diagnosis labels, fit a sparse SpaceNet classifier and visualize which brain regions drive the patient/control distinction. ## Quick Start Ask the agent to run a SpaceNet disease classification on your prepared voxel maps and labels, delegating preprocessing to fmri-skill or smri-skill and fitting to nilearn-tool.
