anonymise_dicom_folder.py

import os
import cv2
import glob
import pydicom
from pydicom.dicomdir import DicomDir
from pydicom.errors import InvalidDicomError

## EDIT THESE TWO LINES IF NEEDED:
FOLDER_PATH = "./dicom_study/" # Path to your folder of DICOMS; default assumes they're in a folder called 'dicom_study' in the same folder as this file
BLANK_PROPORTION = 0.05  # Set this to 0.1 to blank out top 10% of video (and therefore anonymise)

## You shouldn't need to edit anything below this
CONVERT_TO_BGR = True # This reverses the red and blue channels, because CV2 (god knows why) uses BGR not RGB
IGNORED_EXTENSIONS = [".PDF", ".XML", ".ANT", ".AVI", ".avi", ".jpg", ".JPG", '.mp4', '.MP4'] # Don't convert these files

file_list = [file for file in glob.glob(os.path.join(FOLDER_PATH, "*")) if
             os.path.splitext(file)[1] not in IGNORED_EXTENSIONS and not os.path.isdir(file)]

print(f"Found {len(file_list)} files")

def convert_dicom_video(in_file_path, dicom, blank_proportion=None):
    fourcc = cv2.VideoWriter_fourcc(*'MP4V')
    out_file_path = f"{in_file_path}.mp4"
    fps = 1000 / dicom.FrameTime
    if len(dicom.pixel_array.shape) == 4:
        n_frames, width, height, channels = dicom.pixel_array.shape
        print(
            f"{os.path.basename(in_file_path)} -> {os.path.basename(out_file_path)} / FPS: {fps} / WIDTH: {width} / HEIGHT: {height} / CHANNELS: {channels}")
    else:
        (n_frames, width, height), channels = dicom.pixel_array.shape, 1
        print(
            f"{os.path.basename(in_file_path)} -> {os.path.basename(out_file_path)} / FPS: {fps} / WIDTH: {width} / HEIGHT: {height} / CHANNELS: {channels}")

    out = cv2.VideoWriter(out_file_path, fourcc, fps, (height, width), True)
    for i, frame in enumerate(dicom.pixel_array):
        if blank_proportion:
            row_to = int(height * blank_proportion)
            frame[0:row_to, :] = 0
        if channels == 1:
            frame = cv2.cvtColor(frame, cv2.COLOR_GRAY2RGB)
        else:
            if CONVERT_TO_BGR:
                frame = frame[..., ::-1]
        out.write(frame)
    out.release()
    cv2.destroyAllWindows()


def convert_dicom_frame(in_file_path, dicom, blank_proportion=None):
    out_file_path = f"{in_file_path}.jpg"
    if len(dicom.pixel_array.shape) == 2:
        width, height = dicom.pixel_array.shape
        frame = dicom.pixel_array
    else:
        width, height, channels = dicom.pixel_array.shape
        frame = dicom.pixel_array
        if CONVERT_TO_BGR:
            frame = frame[..., ::-1]
    if blank_proportion:
        row_to = int(height * blank_proportion)
        frame[0:row_to, :] = 0
    print(f"{os.path.basename(in_file_path)} -> {os.path.basename(out_file_path)} / WIDTH: {width} / HEIGHT: {height}")
    cv2.imwrite(out_file_path, frame)


if __name__ == "__main__":
    failed_files = []

    for file_path in file_list:
        try:
            dicom = pydicom.dcmread(file_path)
        except pydicom.errors.InvalidDicomError:
            failed_files.append(file_path)
            continue
        if type(dicom) == DicomDir:
            failed_files.append(file_path)
        else:
            try:
                if len(dicom.pixel_array.shape) == 4 or (
                        len(dicom.pixel_array.shape) == 3 and dicom.pixel_array.shape[2] != 3):
                    convert_dicom_video(file_path, dicom, blank_proportion=BLANK_PROPORTION)
                elif len(dicom.pixel_array.shape) == 2 or (
                        len(dicom.pixel_array.shape) == 3 and dicom.pixel_array.shape[2] == 3):
                    convert_dicom_frame(file_path, dicom, blank_proportion=BLANK_PROPORTION)
                else:
                    raise ValueError(f"Got shape: {dicom.pixel_array.shape}")
            except AttributeError:
                failed_files.append(file_path)

    newline = '\n'
    print(
        f"Completed converting {len(file_list)-len(failed_files)} of {len(file_list)} files; the following {len(failed_files)} files failed:\n{newline.join(failed_files)}")