Skip to content

Feature Names

pictologics.features.FEATURE_NAMES lists the feature keys of each feature family, in the order of the results. The pipeline uses it to give each configuration all its feature names, also when a step fails (see Result Guarantees).

from pictologics.features import FEATURE_NAMES, GLCM_FEATURE_NAMES

print(list(FEATURE_NAMES))        # intensity, histogram, ivh, spatial_intensity, ...
print(len(GLCM_FEATURE_NAMES))    # 25
print(GLCM_FEATURE_NAMES[:2])     # ('joint_maximum_GYBY', 'joint_average_60VM')
Family Tuple Features
intensity INTENSITY_FEATURE_NAMES 18
histogram HISTOGRAM_FEATURE_NAMES 23
ivh IVH_FEATURE_NAMES 7
spatial_intensity SPATIAL_INTENSITY_FEATURE_NAMES 2
local_intensity LOCAL_INTENSITY_FEATURE_NAMES 2
morphology MORPHOLOGY_FEATURE_NAMES 27
glcm GLCM_FEATURE_NAMES 25
glrlm GLRLM_FEATURE_NAMES 16
glszm GLSZM_FEATURE_NAMES 16
gldzm GLDZM_FEATURE_NAMES 16
ngtdm NGTDM_FEATURE_NAMES 5
ngldm NGLDM_FEATURE_NAMES 17

A feature key is the feature name and its IBSI code, for example joint_maximum_GYBY. A number after the code tells the variant of the feature, for example _10 in volume_at_intensity_fraction_0.10_BC2M_10. RadiomicsPipeline.describe_features() gives the keys of a pipeline with their families and preprocessing.

Definitions

Pictologics computes each feature as the IBSI reference manual defines it. The manual gives the definition and the formula of each feature, so this documentation does not repeat them. Find a feature in the section of its family, or search the manual for its code, for example Q4LE.

Family Section of the IBSI reference manual
intensity Intensity-based statistical features
histogram Intensity histogram features
ivh Intensity-volume histogram features
spatial_intensity Morphological features (Moran's I N365, Geary's C NPT7)
local_intensity Local intensity features
morphology Morphological features
glcm Grey level co-occurrence based features
glrlm Grey level run length based features
glszm Grey level size zone based features
gldzm Grey level distance zone based features
ngtdm Neighbourhood grey tone difference based features
ngldm Neighbouring grey level dependence based features