dice coeff

dice_coefficient.py

import nibabel as nib
import numpy as np
import csv
import sys
from tqdm import tqdm

def dice_coefficient(mask1, mask2):
    """
    Compute the Dice coefficient, a measure of set similarity.
    Parameters
    ----------
    mask1, mask2 : array-like, bool
        Any arrays of the same shape.
    Returns
    -------
    dice : float
        Dice coefficient as a float on range [0,1].
        Maximum similarity = 1
        No similarity = 0
        
    """
    intersection = np.logical_and(mask1, mask2)

    return 2. * intersection.sum() / (mask1.sum() + mask2.sum())

def process_pair(pair):
    mask1_path, mask2_path = pair

    # Read the masks into numpy arrays
    mask1_nii = nib.load(mask1_path)
    mask2_nii = nib.load(mask2_path)

    mask1_data = mask1_nii.get_fdata()
    mask2_data = mask2_nii.get_fdata()

    # Ensure the masks are boolean
    mask1_data = mask1_data.astype(bool)
    mask2_data = mask2_data.astype(bool)

    # Calculate the Dice coefficient
    dice = dice_coefficient(mask1_data, mask2_data)
    return mask1_path, mask2_path, dice

if __name__ == "__main__":
    if len(sys.argv) < 2:
        print("Usage: python dice_coeff.py <csv_file_with_paths>")
        sys.exit(1)
    
    csv_file_path = sys.argv[1]
    
    with open(csv_file_path, 'r') as file:
        reader = csv.reader(file)
        for row in tqdm(reader, desc="Processing pairs"):
            mask1_path, mask2_path = row
            _, _, dice_score = process_pair((mask1_path, mask2_path))
            print("Dice coefficient between\n{}\nand\n{}:\n{:.4f}\n".format(mask1_path, mask2_path, dice_score))