compute_rad.py

"""
Compute and interpret radiomic distances between two datasets.
"""
import os
from argparse import ArgumentParser
from time import time

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

from src.radiomics_utils import convert_radiomic_dfs_to_vectors, compute_and_save_imagefolder_radiomics, compute_and_save_imagefolder_radiomics_parallel, interpret_radiomic_differences
from src.utils import frechet_distance

def main(
        image_folder1,
        image_folder2,
        force_compute_fresh = False,
        interpret = False,
        parallelize = True
):
    radiomics_fname = 'radiomics.csv'

    radiomics_path1 = os.path.join(image_folder1, radiomics_fname)
    radiomics_path2 = os.path.join(image_folder2, radiomics_fname)

    # if needed, compute radiomics for the images
    if force_compute_fresh or not os.path.exists(radiomics_path1):
        print("No radiomics found computed for image folder 1 at {}, computing now.".format(radiomics_path1))
        if parallelize:
            compute_and_save_imagefolder_radiomics_parallel(image_folder1, radiomics_fname=radiomics_fname)
        else:
            compute_and_save_imagefolder_radiomics(image_folder1, radiomics_fname=radiomics_fname)
        print("Computed radiomics for image folder 1.")
    else:
        print("Radiomics already computed for image folder 1 at {}.".format(radiomics_path1))

    if force_compute_fresh or not os.path.exists(radiomics_path2):
        print("No radiomics found computed for image folder 2 at {}, computing now.".format(radiomics_path2))
        if parallelize:
            compute_and_save_imagefolder_radiomics_parallel(image_folder2, radiomics_fname=radiomics_fname)
        else:
            compute_and_save_imagefolder_radiomics(image_folder2, radiomics_fname=radiomics_fname)
        print("Computed radiomics for image folder 2.")
    else:
        print("Radiomics already computed for image folder 2 at {}.".format(radiomics_path2))

    # load radiomics
    radiomics_df1 = pd.read_csv(radiomics_path1)
    radiomics_df2 = pd.read_csv(radiomics_path2)

    feats1, feats2 = convert_radiomic_dfs_to_vectors(radiomics_df1, 
                                                         radiomics_df2,
                                                         match_sample_count=True, # needed for distance measures
                                                         ) 
    # Frechet distance
    fd = frechet_distance(feats1, feats2)
    rad = np.log(fd)

    print("RaD = {}".format(rad))

    if interpret:
        run_tsne = True
        interpret_radiomic_differences(radiomics_path1, radiomics_path2, run_tsne=run_tsne)

if __name__ == "__main__":
    tstart = time()
    parser = ArgumentParser()

    parser.add_argument('--image_folder1', type=str, required=True)
    parser.add_argument('--image_folder2', type=str, required=True)
    parser.add_argument('--force_compute_fresh', action='store_true', help='re-compute all radiomics fresh')
    parser.add_argument('--interpret', action='store_true', help='interpret the features underlying radiomic distance')

    args = parser.parse_args()

    main(
        args.image_folder1,
        args.image_folder2,
        force_compute_fresh=args.force_compute_fresh,
        interpret=args.interpret
        )

    tend = time()
    print("compute time (sec): {}".format(tend - tstart))