10 Pyradiomics - Extract to DF

Code

from pyhere import here
import pandas as pd

Code

from radiomics import featureextractor

# Initialize the RadiomicsFeatureExtractor
extractor = featureextractor.RadiomicsFeatureExtractor()

10.1 Extract One File

Code

import pandas as pd
import os

# Define a function to apply to each row
def extract_radiomics_features(imageFilepath, maskFilepath, label):
    # Check if image, mask file, and label exist
    if not os.path.exists(imageFilepath) or not os.path.exists(maskFilepath) or pd.isna(label):
        # If any of these are missing, return NA values
        return pd.Series([pd.NA] * len(extractor.enabledFeatures.keys()))
    
    # Try to extract features, return NA if an error occurs
    result = extractor.execute(imageFilepath=imageFilepath,
                               maskFilepath=maskFilepath,
                               label=int(label))
        
    return pd.Series(result)

Code

extract_radiomics_features("../../data/MRI-Brain-Eye/Case1/3 t1 axial Processed_CaPTk.nrrd",
                           "../../data/MRI-Brain-Eye/Case1/Segmentation.seg.nrrd",
                           1)

GLCM is symmetrical, therefore Sum Average = 2 * Joint Average, only 1 needs to be calculated

diagnostics_Versions_PyRadiomics                   v3.0.1
diagnostics_Versions_Numpy                         1.26.4
diagnostics_Versions_SimpleITK               2.3.1-g42ce2
diagnostics_Versions_PyWavelet                      1.5.0
diagnostics_Versions_Python                       3.10.10
                                            ...          
original_ngtdm_Busyness                3.5402573095824215
original_ngtdm_Coarseness           0.0009334119757378282
original_ngtdm_Complexity              1483.2263634997792
original_ngtdm_Contrast               0.10701153309807594
original_ngtdm_Strength                1.0848440176409921
Length: 129, dtype: object

10.2 Extract Features to DF

Code

import pandas as pd
import os
from radiomics import featureextractor


# Initialize the RadiomicsFeatureExtractor
extractor = featureextractor.RadiomicsFeatureExtractor()

def radiomics_extract_to_df(imageFilepaths, maskFilepaths, labels):
    """
    Extracts radiomics features for a list of image and mask file paths with specified labels.
    
    Parameters:
    - imageFilepaths: list or pandas series of paths to image files
    - maskFilepaths: list or pandas series of paths to mask files
    - labels: list or pandas series of labels to be used for feature extraction
    
    Returns:
    - DataFrame with radiomics features, one row per image, mask, and label combination.
    """
    # Create a DataFrame to store the file paths and labels
    df = pd.DataFrame({
        "imageFilepath": imageFilepaths,
        "maskFilepath": maskFilepaths,
        "label": labels
    })

    # Apply the feature extraction function to each row
    features_df = df.apply(lambda row: extract_radiomics_features(row["imageFilepath"], row["maskFilepath"], row["label"]), axis=1)

    # Combine original file information with extracted features
    result_df = pd.concat([df, features_df], axis=1)

    return result_df

Code

# Example lists of file paths and labels
imageFilepaths = ["../../data/MRI-Brain-Eye/Case1/3 t1 axial Processed_CaPTk.nrrd", "../../data/MRI-Brain-Eye/Case2/7 t2_Flair_axial Processed_CaPTk_2.nrrd"]
maskFilepaths = ["../../data/MRI-Brain-Eye/Case1/Segmentation.seg.nrrd", "../../data/MRI-Brain-Eye/Case2/Segmentation_Eye.seg.nrrd"]
labels = [1, 2]  # Example labels for each segment

# Call function
radiomics_df = radiomics_extract_to_df(imageFilepaths, maskFilepaths, labels)
radiomics_df

GLCM is symmetrical, therefore Sum Average = 2 * Joint Average, only 1 needs to be calculated
GLCM is symmetrical, therefore Sum Average = 2 * Joint Average, only 1 needs to be calculated

	imageFilepath	maskFilepath	label	diagnostics_Versions_PyRadiomics	diagnostics_Versions_Numpy	diagnostics_Versions_SimpleITK	diagnostics_Versions_PyWavelet	diagnostics_Versions_Python	diagnostics_Configuration_Settings	diagnostics_Configuration_EnabledImageTypes	...	original_glszm_SmallAreaHighGrayLevelEmphasis	original_glszm_SmallAreaLowGrayLevelEmphasis	original_glszm_ZoneEntropy	original_glszm_ZonePercentage	original_glszm_ZoneVariance	original_ngtdm_Busyness	original_ngtdm_Coarseness	original_ngtdm_Complexity	original_ngtdm_Contrast	original_ngtdm_Strength
0	../../data/MRI-Brain-Eye/Case1/3 t1 axial Proc...	../../data/MRI-Brain-Eye/Case1/Segmentation.se...	1	v3.0.1	1.26.4	2.3.1-g42ce2	1.5.0	3.10.10	{'minimumROIDimensions': 2, 'minimumROISize': ...	{'Original': {}}	...	143.74413264102546	0.0165588948802172	6.12011149362317	0.1779624893435635	1972.0849022912257	3.5402573095824215	0.0009334119757378282	1483.2263634997792	0.10701153309807594	1.0848440176409921
1	../../data/MRI-Brain-Eye/Case2/7 t2_Flair_axia...	../../data/MRI-Brain-Eye/Case2/Segmentation_Ey...	2	v3.0.1	1.26.4	2.3.1-g42ce2	1.5.0	3.10.10	{'minimumROIDimensions': 2, 'minimumROISize': ...	{'Original': {}}	...	89.63527189334039	0.011517574782977553	6.031264647874513	0.24972647702407003	171.04926886676446	1.4260178519990565	0.002340576074194839	289.673667008558	0.08092096635016202	0.3530084647821763

2 rows × 132 columns