Skip to content

HemiSpec v0.1.0

Pre-release
Pre-release

Choose a tag to compare

View all tags
@mqqq333 mqqq333 released this 28 Jun 23:43
v0.1.0
202ba32

HemiSpec v0.1.0 — first public beta

HemiSpec v0.1.0 is the first usable public release of HemiSpec, a research toolkit for reconstruction-derived hemispheric specificity analysis. It is intended as research software / public beta rather than a mature clinical or commercial product.

What is included

  • Unified HemiSpec repository, documentation website, Python package, CLI (hemispec), and GUI (hemispec-gui).
  • Public-safe synthetic quickstart for smoke-testing the command/file contract without private MRI data or model weights.
  • Standard workflow support for DGN inference, ANS/RNS map generation, optional ROI tables, hemisphere-classifier validation, and TRT reliability checks.
  • Windows CLI and compact GUI release artifacts.
  • Lightweight Python wheel and source distribution.
  • SHA256 checksums and a release-artifact manifest.

Download artifacts

  • HemiSpec-CLI-v0.1.0-win64.exe — Windows CLI executable; lightweight build, no bundled model/atlas payloads.
  • HemiSpec-GUI-v0.1.0-win64.zip — Windows compact GUI folder distribution; unzip and run hemispec_gui/hemispec_gui.exe.
  • hemispec_toolkit-0.1.0-py3-none-any.whl — Python wheel.
  • hemispec_toolkit-0.1.0.tar.gz — source distribution.
  • HemiSpec-v0.1.0-SHA256SUMS.txt — SHA256 checksums for the four upload candidates.
  • HemiSpec-v0.1.0-RELEASE_ARTIFACTS.txt — local build, validation, and known-limitation record.

Verification performed

  • py -3.12 -m pytest -q: 44 passed, with 2 known scipy precision warnings from synthetic near-identical arrays.
  • py -3.12 -m mkdocs build --strict: passed.
  • Python wheel/sdist build completed.
  • Windows CLI executable help and TRT smoke paths validated.
  • Windows GUI executable hidden smoke test passed.
  • Full Python CLI DGN workflow validated locally with external DGN/classifier bundles and real sample NIfTI inputs.
  • Forbidden payload scan over wheel/sdist/GUI zip/GUI folder found no blocked .nii, .nii.gz, .pth, .pt, .ckpt, .joblib, .pkl, or .xlsx payloads.
  • SHA256 checksums were recalculated and matched the staged artifacts.

External asset bundle policy

The default v0.1.0 Windows CLI/GUI artifacts are intentionally lightweight. They do not bundle torch, DGN weights, atlas payloads, classifier bundles, real MRI inputs, or generated outputs.

For model-enabled use, distribute approved assets separately, for example:

HemiSpec-Assets/
  models/dgn/
  models/hemisphere_classifier/
  atlases/glasser/
  ASSET_MANIFEST.yml
  SHA256SUMS.txt
  LICENSES/

Every external asset bundle should include version/date, source/provenance, license and citation requirements, compatible hemispec-toolkit version, expected local paths or environment variables, preprocessing assumptions, and checksums.

Useful environment variables:

HEMISPEC_ASSET_ROOT
HEMISPEC_DGN_MODEL_ROOT
HEMISPEC_CLASSIFIER_MODEL_DIR
HEMISPEC_GLASSER_ATLAS
HEMISPEC_GLASSER_LABEL_TABLE

Method boundary

The ANS/RNS metric framework and original cross-hemispheric DGN approach originate from Wang et al. 2024, Patterns, "Using a deep generation network reveals neuroanatomical specificity in hemispheres". HemiSpec packages and extends that workflow for the current software release.

Known limitation

The included compiled Windows artifacts are lightweight builds. Full DGN workflows require a local Python environment with torch and approved external DGN/model/atlas/classifier assets.

Assets 8
Loading