Skip to content

TRT and classifier tasks

This note documents runtime tasks without publishing manuscript-specific result values.

TRT reliability

hemispec trt estimates test-retest reliability from paired ANS/RNS maps. The command writes similarity matrices, within-subject summaries, summary plots, and validation_summary.csv.

hemispec trt   --maps-dir "/path/to/workflow/intermediate/combined_maps"   --out-dir "/path/to/workflow/validation/trt"   --kinds ANS,RNS   --hemis L,R

Before public release, use only approved/synthetic examples in this document. Do not publish private cohort sizes, exact p-values, or manuscript-only summary statistics here.

Hemisphere-classifier validation

hemispec hemi-classify loads saved sklearn/joblib classifier artifacts and applies them to ROI-level ANS/RNS features. Training is not a product feature.

hemispec hemi-classify   --maps-dir "/path/to/workflow/voxel_maps"   --roi-csv "/path/to/workflow/tables/roi_features_bilateral.csv"   --classifier-model-dir "/path/to/classifier_bundle"   --out-dir "/path/to/workflow/validation/hemi_classify"

Supported runtime modes:

single
paired_residual

Publish model performance only through an approved model card or manuscript-safe release note. The model card should include training/validation provenance, feature transforms, software versions, checksums, and citation boundaries.