Skip to content

HemiSpec

HemiSpec is a research-software toolkit for converting preprocessed gray-matter maps into bilateral reconstruction-derived hemispheric measures: ANS (absolute neuroanatomical specificity) and RNS (relative neuroanatomical specificity).

Input boundary

HemiSpec does not accept raw T1-weighted MRI directly as DGN input. First convert each T1 image into an MNI152 1.5 mm masked gray-matter map (*_GM_masked.nii.gz) using the documented FSL preprocessing workflow.

Input and preprocessing Quick start Installation

End-to-end workflow

Cross-hemispheric reconstruction and ANS/RNS definitions
Conceptual reconstruction framework and ANS/RNS definitions from Wang et al. (2024). HemiSpec starts from FSL-preprocessed GM maps. ANS/RNS are non-negative GM-derived measures, not distances in millimetres. Hemisphere validation is optional; behavioral-phenotype analysis shown in the schematic is not a built-in workflow.
Stage Input Main operation Output
Preprocessing T1-weighted NIfTI FSL brain extraction, tissue segmentation, affine MNI registration, GM threshold/mask *_GM_masked.nii.gz
Reconstruction Preprocessed GM map Left-to-right and right-to-left DGN inference Reconstructed target hemispheres
Metric computation Actual and reconstructed GM ANS/RNS residual metrics ANS.L, ANS.R, RNS.L, RNS.R maps
Optional summaries Voxel maps + compatible atlas ROI aggregation Long and wide ROI tables
Optional validation Maps/ROI features Hemisphere classification and/or TRT analysis Validation tables and plots

What belongs to the original method, and what HemiSpec adds

The cross-hemispheric DGN framework and ANS/RNS definitions originate from Wang et al. (2024). HemiSpec packages that method into an installable API/CLI/GUI workflow, model-asset discovery, bilateral map export, ROI summaries, validation utilities, documentation, and release tooling.

  • Original method and metrics: cite Wang et al. (2024).
  • HemiSpec software or downstream studies: cite the relevant public software/manuscript record when available.

See Citation for the complete reference and citation boundaries.

Choose your path

  • Prepare real MRI input

    Start from T1-weighted NIfTI, run process_single_subject.sh, and verify the DGN input grid and quality-control checks.

    Input and preprocessing

  • Run HemiSpec

    Install the current source checkout with Git LFS, inspect model readiness, and run the GUI or bilateral CLI workflow.

    Quick start

  • Understand the method

    Review the reconstruction framework, original ANS/RNS definitions, and HemiSpec implementation details.

    Methods

  • Models and atlas assets

    Learn how DGN checkpoints, classifier bundles, and optional ROI atlases are located and distributed.

    Data and models

Current software scope

This site documents the current main source checkout. The archived v0.1.0 release does not include every command shown here, including quickstart and automatic model-cache installation; see Installation. The PyPI project is not public yet. Record the source commit when reporting software use.

The GUI and CLI generate voxel-wise ANS/RNS maps. ROI tables, hemisphere-classifier validation, and TRT validation are optional. ROI export requires an atlas on the input grid; the released classifier specifically requires its compatible Glasser atlas and labels. Each run should use a new output directory.

Source: github.com/mqqq333/HemiSpec. Documentation built with Material for MkDocs.