Upstream Dependencies and Documentation Links

This page summarizes official third-party library documentation and concept references for HABIT modules. Implementation details, full hyperparameters, and mathematical definitions are authoritative upstream; the HABIT user manual focuses on config keys, CLI, and data contracts (see Configuration Reference).

Image preprocessing (`habit preprocess`)

Purpose	Library / tool	Documentation
Registration (`backend: ants`)	ANTsPy (`ants.registration`)	Registration · ants.registration API
Registration (`backend: elastix`)	Official elastix / transformix executables (CLI `-f` / `-m` / `-out` / `-p` / `-tp`, etc.; see elastix command-line reference)	Website / download · Model Zoo (LKEB, by data/task) · Model Zoo (elastix.dev)
Registration (`backend: simpleitk`)	SimpleITK (`ImageRegistrationMethod`)	ImageRegistrationMethod · Fundamental Concepts
ANTs core	ANTs	ANTsX/ANTs
Resampling / N4 / majority filter	SimpleITK	User Guide · Fundamental Concepts · ResampleImageFilter
DICOM to NIfTI	dcm2niix	rordenlab/dcm2niix

Habitat segmentation (`habit get-habitat`)

Clustering, dimensionality reduction, and related steps rely heavily on scikit-learn; superpixel SLIC etc. match the implementing class.

Purpose	Library	Documentation
K-Means / GMM / pipeline and metrics	scikit-learn	Clustering · Metrics
Feature scaling, unsupervised preprocessing (sklearn pipeline concepts)	scikit-learn	Preprocessing

Habitat features and traditional radiomics (`habit extract` / `habit radiomics`)

Purpose	Library	Documentation
Radiomics features	PyRadiomics	PyRadiomics
Image I/O / some geometry	SimpleITK	same as preprocessing section

Machine learning (`habit model` / `habit cv`)

Purpose	Library	Documentation
Preprocessing, classifiers, pipelines, metrics	scikit-learn	User Guide · API Reference
Class imbalance resampling (if enabled)	imbalanced-learn (optional)	imbalanced-learn
Gradient boosting etc. (if configured)	XGBoost / LightGBM, etc.	XGBoost Python (as needed)
AutoML (if configured)	AutoGluon	AutoGluon Tabular
SHAP plots (if enabled)	shap	SHAP

Model comparison (`habit compare`)

Purpose	Library / method	Documentation
ROC / PR / calibration plots	matplotlib / in-house plotting (see source)	Medical papers often cite TRIPOD / CONSORT-AI reporting guidelines (not library docs)
DeLong test (if enabled)	see implementation and literature	refer to HABIT output docs and cited statistical literature

ICC / Test–retest / other

Purpose	Library	Documentation
Statistics and matrix operations	NumPy / SciPy / pandas	NumPy · SciPy · pandas

ROI delineation (external tools)

HABIT does not include delineation; common tools: