Extract radiomics from one ROI

Build an intensity mask, retain CT intensities from -400 to 400 HU, discretize the retained values into 32 bins, and extract intensity statistics and GLCM features. These parameters are illustrative; choose and record settings for your study using Re-segmentation guidelines and Discretization guidelines.

from zrad.image import Image
from zrad.preprocessing import IntensityMaskBuilder, Resegmenter, RoiData, TextureDiscretizer
from zrad.radiomics import Radiomics

image = Image.from_nifti("path/to/phantom.nii.gz")
mask = Image.from_nifti_mask("path/to/mask.nii.gz", reference=image)

# Keep original CT values inside the ROI and NaN outside it.
roi = IntensityMaskBuilder().apply(RoiData(image=image, morphological_mask=mask))
roi = Resegmenter(intensity_range=(-400, 400)).apply(roi)
roi = TextureDiscretizer(number_of_bins=32).apply(roi)

# Request only the feature families needed for this analysis.
features = Radiomics(aggr_dim="3D", aggr_method="AVER").extract_features(
    roi_data=roi, families=["intensity_statistics", "glcm"], include_metadata=True
)
print(f"Mean intensity: {features['stat_mean']:.2f} HU")
print(f"GLCM contrast: {features['cm_contrast_3D_avg']:.3f}")
print(f"ROI voxels: {features['no_voxels']}; occupied bins: {features['no_bins']}")

With the bundled phantom, the output includes approximately 25.07 HU for mean intensity and 9.567 for GLCM contrast. no_voxels describes the morphological ROI; it can exceed the number of voxels retained after intensity resegmentation. See Understanding results for feature and metadata meanings.