VoxelRadiomicsFeatures

Note

This page is a reference documentation. It only explains the class signature, and not how to use it. Please refer to the Habitat Guide and Python API guide (v2.0) for usage.

class VoxelRadiomicsFeatures(modalities: Sequence[str] = (), roi: str | None = None, params_file: str | None = None, params: Dict[str, Any] | None = None, kernel_radius: int = 3, voxel_batch: int | str = 1000, use_torch_radiomics: str | bool = 'auto', torch_device: str = 'auto', torch_dtype: str = 'float64', use_gpu_matrices: str | bool = 'auto', output_float32: bool = True, class_progress: bool = False, crop_to_roi: bool = True, cache_dir: str | None = None, modality: str | None = None, as_: str | None = None)[source]

Bases: object

Describe every ROI voxel by a PyRadiomics feature vector.

One extraction pass runs per modality and each column is suffixed with -{modality}, the v0.1 scheme for concat(voxel_radiomics(m1), voxel_radiomics(m2)).

Parameters:

  • modality – Single modality key – the explicit form used inside feature trees. Mutually exclusive with modalities.

  • modalities – Modality keys to extract from, in feature order; empty selects every image the subject carries.

  • as – Optional output-column alias. Valid only with exactly one resolved modality; the column suffix then uses the alias.

  • roi – Mask key defining the region of interest; None uses the subject’s single mask.

  • params_file – Path to a PyRadiomics parameter YAML; None selects the bundled voxel preset.

  • params – Inline PyRadiomics settings, for API callers holding settings in memory. Mutually exclusive with params_file.

  • kernel_radius – Neighbourhood radius in voxels; radius 1 is a 3x3x3 cube, radius 3 a 7x7x7 cube.

  • voxel_batch – ROI voxels per batch. Default 1000. Pass a larger integer on a 12–24 GB GPU, or "auto" to pick from VRAM.

  • use_torch_radiomics – "auto", True or False – whether to use the TorchRadiomics path when torch and CUDA are present.

  • torch_device – Torch device string, or "auto" to select one.

  • torch_dtype – "float64" (default) or "float32" for the torch path.

  • use_gpu_matrices – "auto", True or False – whether the TorchRadiomics texture matrices (GLCM, …) are built on GPU by habit.kernels.radiomics.gpumatrices instead of the single-threaded PyRadiomics C extension. "auto" follows the torch device. Bit-identical counts either way.

  • output_float32 – Downcast the feature columns to float32, the v0.1 default that keeps large voxel tables manageable.

  • class_progress – When True, print and tqdm each PyRadiomics class (firstorder, glcm, …). Default False: one execute() with no per-class lines; Cohort.map still shows subject progress.

  • crop_to_roi – When True (default), crop image and mask to the ROI bounding box plus kernel_radius padding before calling execute. PyRadiomics re-applies the identical crop internally, so feature values are bit-identical; the pre-crop just keeps the full-volume diagnostics (sitk.Hash, whole-image statistics) and mask checks off the big volume, saving several seconds per modality on whole-body scans.

  • cache_dir – Optional directory for extracted fields. A hit skips PyRadiomics. The cache key ignores voxel_batch and device knobs so a later run with a larger batch can reuse the file.

See also

habit.voxel_features.extract_voxel_texture

One image + mask call that constructs this extractor.

__init__(modalities: Sequence[str] = (), roi: str | None = None, params_file: str | None = None, params: Dict[str, Any] | None = None, kernel_radius: int = 3, voxel_batch: int | str = 1000, use_torch_radiomics: str | bool = 'auto', torch_device: str = 'auto', torch_dtype: str = 'float64', use_gpu_matrices: str | bool = 'auto', output_float32: bool = True, class_progress: bool = False, crop_to_roi: bool = True, cache_dir: str | None = None, modality: str | None = None, as_: str | None = None) → None[source]
property spec: Spec

Return the algorithm specification used for provenance.

__call__(subject: Subject) → VoxelFeatureField[source]

Compute per-voxel radiomics for one subject.

Parameters:

subject – Subject providing the requested modalities and mask.

Returns:

One row per ROI voxel, one column per feature and modality.

Raises:

HABITAPIError – If a requested modality is absent, or a modality’s extraction does not cover every ROI voxel.

Examples using habit.voxel_features.VoxelRadiomicsFeatures

Voxel texture and GPU

Voxel texture and GPU

Preprocessing voxel texture before clustering

Preprocessing voxel texture before clustering