Request code for BraTS dataset

[wyxogo]

thx for your research. Could you please provide the preprocessing code for the BraTS 2021 dataset to make it compatible with this project? I am particularly interested in working with medical image data.

[xcyao00]

You may can try the following code to prepare dataset from the original BraTS data.

import os
import PIL
import cv2
import tarfile
import numpy as np
import pydicom
import pandas as pd
from glob import glob
import nibabel as nib
from PIL import Image
import matplotlib.pyplot as plt
from scipy import ndimage as ndi
files = os.listdir('C:/Users/i-Dov/Downloads/BraTS2021_Training_Data')
img_ids = [f.split('_')[-1] for f in files]

SCAN_NUM = 83  # the original dataset has many scans, we only randomly use one scan to reduce the number of extracted images
for img_id in img_ids:
    img_nii = f'C:/Users/i-Dov/Downloads/BraTS2021_Training_Data/BraTS2021_{img_id}/BraTS2021_{img_id}_flair.nii.gz'
    image = nib.load(img_nii).get_fdata()  # (240, 240, 155)
    image = image[:, :, SCAN_NUM]
    min_val, max_val = np.min(image), np.max(image)
    image = (image - min_val) / (max_val - min_val)
    image = image * 255
    pil_image = Image.fromarray(image)
    
    seg_nii = f'C:/Users/i-Dov/Downloads/BraTS2021_Training_Data/BraTS2021_{img_id}/BraTS2021_{img_id}_seg.nii.gz'
    seg_mask = nib.load(seg_nii).get_fdata()  # (240, 240, 155)
    seg_mask = seg_mask[:, :, SCAN_NUM]
    if np.unique(seg_mask).shape[0] == 1:
        seg_image = Image.fromarray(seg_mask)
        seg_mask = seg_mask.astype(np.uint8)
        seg_image = Image.fromarray(seg_mask)
        seg_image.save('./brats/gt/normal/' + img_id + '_mask.png', format='png')
        pil_image.convert('RGB').save('./brats/normal/' + img_id + '.jpg', format='jpeg')
    else:
        seg_mask[seg_mask != 0] = 255
        seg_mask = seg_mask.astype(np.uint8)
        seg_image = Image.fromarray(seg_mask)
        seg_image.save('./brats/gt/abnormal/' + img_id + '_mask.png', format='png')
        pil_image.convert('RGB').save('./brats/abnormal/' + img_id + '.jpg', format='jpeg')

Or, you can download the organized dataset by us from https://huggingface.co/datasets/xcyao00/resad-data/tree/main, please download brats.zip. The structure of the dataset is the same as the MVTec dataset, so its usage is the same as the MVTec dataset. You can use the Dataset class (BRATS) in datasets/brats.py to load data.

[wyxogo]

thx