import os import numpy as np import nibabel as nib from os import path # Read files from the specified directory at a given level (num). def readDir(path, num): """ Reads the directory and returns the filenames from the specified level (num). """ im = [] k = 0 for root in os.walk(path): if k == 0: im.append(root[num]) k = k+1 if k == 1: continue return im # Recursively get all filenames in a directory and its subdirectories. def get_filenames(dirpath, filenames): """ Recursively get all filenames in a directory and its subdirectories. """ if not path.isabs(dirpath): dirpath = path.abspath(dirpath) for pathname, dirs, files in os.walk(dirpath): if files: for f in files: filenames.append(path.join(pathname, f)) if dirs: for dir_ in dirs: get_filenames(path.join(pathname, dir_), filenames) # Save a numpy array as a NIfTI image def save_nii(arr, file_path, affine): """ Save a numpy array as a NIfTI image. """ nib.Nifti1Image(arr, affine).to_filename(file_path) # Load a NIfTI file and return its data and affine transformation matrix def load_nii_affine(filename): """ Loads a NIfTI file and returns its data and affine transformation matrix. """ if not os.path.exists(filename): return np.array([1]) nii = nib.load(filename) data = nii.get_fdata() # get_fdata() is preferred over get_data() in newer nibabel versions affine = nii.affine nii.uncache() return data, affine # Calculate the bounding box of a mask (region of interest) def get_bbox_from_mask(mask, outside_value=0): """ Calculate the bounding box of a mask (region of interest). """ mask_voxel_coords = np.where(mask != outside_value) minzidx = int(np.min(mask_voxel_coords[0])) maxzidx = int(np.max(mask_voxel_coords[0])) + 1 minxidx = int(np.min(mask_voxel_coords[1])) maxxidx = int(np.max(mask_voxel_coords[1])) + 1 minyidx = int(np.min(mask_voxel_coords[2])) maxyidx = int(np.max(mask_voxel_coords[2])) + 1 return [minzidx, maxzidx], [minxidx, maxxidx], [minyidx, maxyidx] # Crop the image to the given bounding box def crop_to_bbox(image, bbox): """ Crop the image to the given bounding box. """ assert len(image.shape) == 3, "Only supports 3D images." resizer = (slice(bbox[0][0], bbox[0][1]), slice(bbox[1][0], bbox[1][1]), slice(bbox[2][0], bbox[2][1])) return image[resizer] # Extract and crop the data from images def Crop_data(): """ Crop the images in the specified directory and save the cropped images to the target path. """ import warnings warnings.filterwarnings("ignore") all_file_names = readDir(original_img_path, 1)[0] img_size_list = [] for img_index in range(len(all_file_names)): print(f'== Processing {img_index + 1}/{len(all_file_names)}: {all_file_names[img_index]} ==') # File paths img_path_T1 = original_img_path + all_file_names[img_index] + '/' + all_file_names[img_index] + '_t1.nii.gz' img_path_T1C = original_img_path + all_file_names[img_index] + '/' + all_file_names[img_index] + '_t1ce.nii.gz' img_path_T2 = original_img_path + all_file_names[img_index] + '/' + all_file_names[img_index] + '_t2.nii.gz' img_path_FLAIR = original_img_path + all_file_names[img_index] + '/' + all_file_names[img_index] + '_flair.nii.gz' img_path_seg = original_img_path + all_file_names[img_index] + '/' + all_file_names[img_index] + '_seg.nii.gz' # Load data nii_data_1, affine = load_nii_affine(img_path_T1) nii_data_2, affine_2 = load_nii_affine(img_path_T1C) nii_data_3, affine_3 = load_nii_affine(img_path_T2) nii_data_4, affine_4 = load_nii_affine(img_path_FLAIR) nii_seg, affine_seg = load_nii_affine(img_path_seg) # Calculate bounding box combined_image = nii_data_1 + nii_data_2 + nii_data_3 + nii_data_4 bbox = get_bbox_from_mask(combined_image) # Crop images cropped_1 = crop_to_bbox(nii_data_1, bbox) cropped_2 = crop_to_bbox(nii_data_2, bbox) cropped_3 = crop_to_bbox(nii_data_3, bbox) cropped_4 = crop_to_bbox(nii_data_4, bbox) cropped_seg = crop_to_bbox(nii_seg, bbox) cropped_seg = (cropped_seg > 0).astype(np.uint8) # Get cropped image size img_size = cropped_1.shape # Save paths save_path_T1 = save_img_path + all_file_names[img_index] + '/T1.nii' save_path_T1C = save_img_path + all_file_names[img_index] + '/T1C.nii' save_path_T2 = save_img_path + all_file_names[img_index] + '/T2.nii' save_path_FLAIR = save_img_path + all_file_names[img_index] + '/FLAIR.nii' save_path_seg = save_img_path + all_file_names[img_index] + '/mask.nii' if not os.path.exists(save_img_path + all_file_names[img_index] + '/'): os.makedirs(save_img_path + all_file_names[img_index] + '/') # Save cropped images save_nii(cropped_1.astype(np.uint16), save_path_T1, affine) save_nii(cropped_2.astype(np.uint16), save_path_T1C, affine) save_nii(cropped_3.astype(np.uint16), save_path_T2, affine) save_nii(cropped_4.astype(np.uint16), save_path_FLAIR, affine) save_nii(cropped_seg.astype(np.uint8), save_path_seg, affine_seg) img_size_list.append(img_size) # Save the crop sizes of all images np.save(save_crop_size_path + 'crop_size.npy', img_size_list) return all_file_names def get_max_size(): img_size_list = np.load(save_crop_size_path + 'crop_size.npy') size_1 = np.zeros((len(img_size_list))) size_2 = np.zeros((len(img_size_list))) size_3 = np.zeros((len(img_size_list))) for size_temp in range(len(img_size_list)): size_1[size_temp] = int(img_size_list[size_temp][0]) size_2[size_temp] = int(img_size_list[size_temp][1]) size_3[size_temp] = int(img_size_list[size_temp][2]) print(np.max(size_1), np.max(size_2), np.max(size_3)) # 3D padding def Padding3d(data, padding_shape): """ Add padding to a 3D image. """ padding_h, padding_w, padding_d = (np.array(padding_shape) - np.array(np.shape(data))) // 2 padding_T = np.zeros([padding_h, np.shape(data)[1], np.shape(data)[2]], dtype=np.uint8) padding_B = np.zeros([padding_shape[0] - padding_h - np.shape(data)[0], np.shape(data)[1], np.shape(data)[2]], dtype=np.uint8) H_data = np.concatenate([padding_T, data, padding_B], 0) padding_L = np.zeros([np.shape(H_data)[0], padding_w, np.shape(H_data)[2]], dtype=np.uint8) padding_R = np.zeros([np.shape(H_data)[0], padding_shape[1] - padding_w - np.shape(H_data)[1], np.shape(H_data)[2]], dtype=np.uint8) W_data = np.concatenate([padding_L, H_data, padding_R], 1) padding_F = np.zeros([np.shape(W_data)[0], np.shape(W_data)[1], padding_d], dtype=np.uint8) padding_B = np.zeros([np.shape(W_data)[0], np.shape(W_data)[1], padding_shape[2] - padding_d - np.shape(W_data)[2]], dtype=np.uint8) data = np.concatenate([padding_F, W_data, padding_B], 2) return data # Pad the images with zeros and save them to the target directory. def Padding_data(all_file_names, padding_shape=(160, 192, 160)): """ Pad the images with zeros and save them to the target directory. """ img_size_list = np.load(save_crop_size_path + 'crop_size.npy', allow_pickle=True) img_index = 0 for img_size in img_size_list: print(f'== Padding {img_index + 1}/{len(img_size_list)} ==') # Get file paths img_path_T1 = save_img_path + all_file_names[img_index] + '/T1.nii' img_path_T1C = save_img_path + all_file_names[img_index] + '/T1C.nii' img_path_T2 = save_img_path + all_file_names[img_index] + '/T2.nii' img_path_FLAIR = save_img_path + all_file_names[img_index] + '/FLAIR.nii' img_path_seg = save_img_path + all_file_names[img_index] + '/mask.nii' # Load data nii_data_1, affine = load_nii_affine(img_path_T1) nii_data_2, affine_2 = load_nii_affine(img_path_T1C) nii_data_3, affine_3 = load_nii_affine(img_path_T2) nii_data_4, affine_4 = load_nii_affine(img_path_FLAIR) nii_seg, affine_seg = load_nii_affine(img_path_seg) # Perform padding padded_1 = Padding3d(nii_data_1, padding_shape) padded_2 = Padding3d(nii_data_2, padding_shape) padded_3 = Padding3d(nii_data_3, padding_shape) padded_4 = Padding3d(nii_data_4, padding_shape) padded_seg = Padding3d(nii_seg, padding_shape) if not os.path.exists(save_img_path_final + all_file_names[img_index] + '/'): os.makedirs(save_img_path_final + all_file_names[img_index] + '/') # Save padded images save_nii(padded_1.astype(np.uint16), save_img_path_final + all_file_names[img_index] + '/T1.nii', affine) save_nii(padded_2.astype(np.uint16), save_img_path_final + all_file_names[img_index] + '/T1C.nii', affine_2) save_nii(padded_3.astype(np.uint16), save_img_path_final + all_file_names[img_index] + '/T2.nii', affine_3) save_nii(padded_4.astype(np.uint16), save_img_path_final + all_file_names[img_index] + '/FLAIR.nii', affine_4) save_nii(padded_seg.astype(np.uint8), save_img_path_final + all_file_names[img_index] + '/mask.nii', affine_seg) img_index += 1 if __name__ == '__main__': # File path definitions original_img_path = './cropdata/beforedata/' # Path to the original images save_img_path = './cropdata/aftercrop/' # Path to save cropped images save_crop_size_path = './cropdata/' # Path to save crop sizes save_img_path_final = './cropdata/afterpadding/' # Path to save padded images # Step 1: Crop out the black borders from the data all_file_names = Crop_data() # Calls the Crop_data function to crop images # Step 2: Verify if the maximum size in the data is smaller than (160, 192, 160) get_max_size() # Calls the get_max_size function to check the maximum size in the data # Step 3: Add padding with zero values and zero slices to ensure each axis is a multiple of 3 or 2 # The default padding size is (160, 192, 160) Padding_data(all_file_names) # Calls the Padding_data function to add padding to the images