| import cv2 | |
| import matplotlib.pyplot as plt | |
| import nibabel as nib | |
| import os | |
| import glob | |
| # from scipy.ndimage import zoom | |
| import numpy as np | |
| import skimage.transform | |
| import torch.optim | |
| from skimage import transform | |
| from scipy.ndimage import binary_fill_holes,zoom | |
| from scipy.ndimage import map_coordinates | |
| #todo:先裁剪出bounding box，在resize成统一大小 | |
| # from imblearn.over_sampling import SMOTE | |
| # image_paths=glob.glob('./train/label/*.png') | |
| # cnt_mp={0:0,1:0,2:0,3:0,4:0,5:0,6:0,7:0} | |
| # for path in image_paths: | |
| # image=cv2.imread(path)[:,:,0] | |
| # for i in cnt_mp: | |
| # cnt_mp[i]+=np.sum(image==i) | |
| # | |
| # # break | |
| # print(cnt_mp) | |
| # cnt_mp.pop(0) | |
| # cnt=0 | |
| # for i in cnt_mp: | |
| # cnt+=cnt_mp[i] | |
| # for i in cnt_mp: | |
| # print(i,(cnt/len(cnt_mp)/cnt_mp[i])) | |
| # image_paths=glob.glob('./mr_train/*_image.nii.gz') | |
| # for path in image_paths: | |
| # # print(path) | |
| # | |
| # filename=path.split('\\')[-1].split('.')[0] | |
| # | |
| # print(filename) | |
| # image=nib.load(path).dataobj | |
| # image=np.floor((image-np.min(image))/(np.max(image)-np.min(image))*255) | |
| # # image=zoom(image,[256/image.shape[0],256/image.shape[1],128/image.shape[2]],order=0) | |
| # # print(image.dtype) | |
| # | |
| # for i in range(image.shape[-1]): | |
| # | |
| # cv2.imwrite(os.path.join('train/image',f'{filename}_{i}.png'),image[:,:,i]) | |
| # | |
| # # break | |
| # # | |
| # os.path.join('') | |
| # image_paths=glob.glob('./ct_train/*_image.nii.gz') | |
| # for path in image_paths: | |
| # data_info=nib.load(path) | |
| # h_,w_,d_=data_info.header['pixdim'][1:4] | |
| # h,w,d=data_info.shape | |
| # data=data_info.get_fdata() | |
| # | |
| # print(data.shape) | |
| # print('实际大小',int(h*h_),int(w*w_),int(d*d_)) | |
| # # part_image=data[data!=-1024] | |
| # # data=(data-np.mean(data))/np.std(data) | |
| # # data=cv2.imread('../xm12/train_image/01001.png') | |
| # # data=(data-np.mean(part_image))/np.std(part_image) | |
| # print('取值范围', data.min(), data.max()) | |
| # # plt.imshow(data[:, :, 0],cmap='gray') | |
| # # plt.show() | |
| # | |
| # # nonzero_mask = np.zeros(data.shape[1:], dtype=bool) | |
| # # for c in range(data.shape[0]): | |
| # # this_mask = data[c] != 0 | |
| # # nonzero_mask = nonzero_mask | this_mask | |
| # # nonzero_mask = binary_fill_holes(nonzero_mask) | |
| # break | |
| def handle_image_and_label(): | |
| cnt=0 | |
| image_paths=glob.glob(r'C:\Users\zhang\PycharmProjects\mmsegmentation\data\Task02_Heart\labelsTr\*.nii.gz') | |
| for path in image_paths: | |
| # print(path) | |
| # if cnt<16: | |
| # folder='mri_train_2d' | |
| # else: | |
| # folder='mri_test_2d' | |
| folder='mri_train_2d' | |
| filename=path.split('\\')[-1].split('.')[0].replace('label','image') | |
| print(filename) | |
| image=nib.load(path).dataobj | |
| image=np.array(image,dtype=np.int8) | |
| # print(image.shape) | |
| # print(np.unique(image)) | |
| ''' | |
| label_map = [0, 1] | |
| for i, v in enumerate(label_map): | |
| image = np.where(image == v, i, image) | |
| image = np.where(image == 421, 2, image) | |
| # 能用的label-resize | |
| rows, cols, dim = 256,256,image.shape[-1] | |
| orig_rows, orig_cols, orig_dim = image.shape | |
| row_scale = float(orig_rows) / rows | |
| col_scale = float(orig_cols) / cols | |
| dim_scale = float(orig_dim) / dim | |
| map_rows, map_cols, map_dims = np.mgrid[:rows, :cols, :dim] | |
| map_rows = row_scale * (map_rows + 0.5) - 0.5 | |
| map_cols = col_scale * (map_cols + 0.5) - 0.5 | |
| map_dims = dim_scale * (map_dims + 0.5) - 0.5 | |
| coord_map = np.array([map_rows, map_cols, map_dims]) | |
| image=map_coordinates(image, coord_map, order=1) | |
| ''' | |
| # 自己写的label-resize | |
| # print(image.shape,type(image),image.dtype,np.unique(image)) | |
| # break | |
| # print(np.unique(image)) | |
| # final_index=[] | |
| # # print(np.unique(image,axis=0).shape) | |
| # # print(np.unique(image, axis=1).shape) | |
| # # print(np.unique(image, axis=2).shape) | |
| # temp_index=[] | |
| # for i in range(0,image.shape[0]): | |
| # if len(np.unique(image[i,:,:]))!=1: | |
| # temp_index.append(i) | |
| # break | |
| # for i in range(image.shape[0]-1,0,-1): | |
| # if len(np.unique(image[i,:,:])) != 1: | |
| # temp_index.append(i) | |
| # break | |
| # final_index.append(temp_index) | |
| # | |
| # temp_index = [] | |
| # for i in range(0,image.shape[1]): | |
| # if len(np.unique(image[:, i, :])) != 1: | |
| # temp_index.append(i) | |
| # break | |
| # for i in range(image.shape[1] - 1, 0, -1): | |
| # if len(np.unique(image[:, i, :])) != 1: | |
| # temp_index.append(i) | |
| # break | |
| # final_index.append(temp_index) | |
| # | |
| # temp_index = [] | |
| # for i in range(0,image.shape[2]): | |
| # if len(np.unique(image[:, :, i])) != 1: | |
| # temp_index.append(i) | |
| # break | |
| # for i in range(image.shape[2] - 1, 0, -1): | |
| # if len(np.unique(image[:, :, i])) != 1: | |
| # temp_index.append(i) | |
| # break | |
| # final_index.append(temp_index) | |
| # | |
| # print(final_index) | |
| # image=image[final_index[0][0]:final_index[0][1], | |
| # final_index[1][0]:final_index[1][1], | |
| # final_index[2][0]:final_index[2][1]] | |
| # 1 2 3 | |
| # 注意用保存为图片时，数值类型不要unsigned | |
| image=image.astype(np.int8) | |
| print(np.unique(image),type(image),image.shape,image.dtype) | |
| # print(np.unique(image[:,:,60])) | |
| for i in range(image.shape[-1]): | |
| cv2.imwrite(os.path.join(f'{folder}/label',f'{filename}_{i}.png'),image[:,:,i]) | |
| # np.save(os.path.join('./train-3d/label',filename),image) | |
| # if len(np.unique(image>=3)): | |
| # print(image.shape) | |
| # for i in range(image.shape[-1]): | |
| # cv2.imwrite(os.path.join('train/label',f'{filename}_{i}.png'),image[:,:,i]) | |
| # filename=filename.replace('label','image') | |
| image = np.array(nib.load(os.path.join('imagesTr', filename + '.nii.gz')).dataobj) | |
| print(np.unique(image)) | |
| # f,ax=plt.subplots(2,1) | |
| # ax[0].imshow(image[:,:,0]) | |
| # ct处理方式 | |
| # image=((image+1024)/4095)*255 | |
| # mri处理方式 | |
| image=((image-np.min(image))/(np.max(image)-np.min(image)))*255 | |
| # print(np.unique(image)) | |
| # ax[1].imshow(image[:,:,0]) | |
| # plt.show() | |
| # print(image[:,:,0][128]) | |
| # image = np.floor( | |
| # (image - np.min(image)) / (np.max(image) - np.min(image)) * 255) | |
| # image = image[final_index[0][0]:final_index[0][1], | |
| # final_index[1][0]:final_index[1][1], | |
| # final_index[2][0]:final_index[2][1]] | |
| # image=(image-np.min(image))/(np.max(image)-np.min(image))*255 | |
| # 能用的image-resize | |
| ''' | |
| image=skimage.transform.resize(image,(256,256,image.shape[-1]),order=3) | |
| ''' | |
| # image=image.astype(np.float32) | |
| # np.save(os.path.join('./train-3d/image',filename),image) | |
| print(image.dtype) | |
| # print(np.unique(image[:,:,60])) | |
| for i in range(image.shape[-1]): | |
| cv2.imwrite(os.path.join(f'{folder}/image',f'{filename}_{i}.png'),image[:,:,i]) | |
| cnt+=1 | |
| # break | |
| if __name__ == '__main__': | |
| # paths=glob.glob(r'C:\Users\zhang\PycharmProjects\mmsegmentation\data\Task02_Heart\imagesTr\*.nii.gz') | |
| # for p in paths: | |
| # | |
| # img=nib.load(p).dataobj | |
| # print(np.min(img),np.max(img)) | |
| handle_image_and_label() | |
| img=cv2.imread('./mri_train_2d/image/la_003_60.png',0) | |
| label=cv2.imread('./mri_train_2d/label/la_003_60.png',0) | |
| print(np.unique(label)) | |
| plt.subplot(1,2,1) | |
| plt.imshow(img) | |
| plt.subplot(1,2,2) | |
| plt.imshow(label,cmap='gray',interpolation='none') | |
| plt.show() | |
| # img_3d=nib.load('./mr_train/mr_train_1011_image.nii.gz').dataobj | |
| # label_3d=nib.load('./mr_train/mr_train_1011_label.nii.gz').dataobj | |
| # | |
| # # print(img_3d.shape) | |
| # img_slice=img_3d[:,:,90] | |
| # | |
| # label_slice=label_3d[:,:,90] | |
| # # label_slice=np.where(label_slice==420,2,label_slice) | |
| # # label_slice = np.where(label_slice == 850, 7, label_slice) | |
| # print(np.unique(label_slice)) | |
| # fig,ax=plt.subplots(1,2) | |
| # ax[0].imshow(img_slice,cmap='gray') | |
| # ax[1].imshow(label_slice,cmap='CMRmap') | |
| # plt.show() |