Radiomics features¶

pydcm.radiomics computes IBSI radiomic features over an ROI — 7 feature classes (firstorder, glcm, glrlm, glszm, gldm, ngtdm, shape), with standard radiomics feature names and validated against published baselines (104 features).

From a DICOM image + mask¶

from pydcm.radiomics import radiomics

feats = radiomics("ct.dcm", mask="mask.dcm")     # pixels → HU, spacing from geometry
feats["original_firstorder_Mean"]
feats["original_shape_VoxelVolume"]
feats["original_glcm_Contrast"]

Pixels are decoded to real-world values (HU) and the spacing is read from the image geometry (PixelSpacing / SliceThickness), so the feature values are in physical units without any manual setup.

From arrays¶

import numpy as np

feats = radiomics(image_array, mask=mask_array, spacing=(z, y, x))

Binning and preprocessing knobs¶

The same surface as the dcmradiomics CLI:

feats = radiomics(
    "ct.dcm", mask="mask.dcm",
    bins=32,                       # fixed bin count …
    bin_width=25.0,                # … or a fixed bin width (mutually exclusive)
    value_range=(-1024.0, 3071.0),
    resample=1.0,                  # isotropic resample (mm) before extraction
    normalize=False,
    wavelet=False, log_sigma=None, # optional image-filter families
)

Extractor object API¶

For code that expects the familiar extractor object, that API is available too:

from pydcm.radiomics import RadiomicsFeatureExtractor

extractor = RadiomicsFeatureExtractor()
extractor.enableFeatureClassByName("firstorder")
result = extractor.execute("ct.dcm", "mask.dcm")