SimpleITK保存.nii.gz的文件

            print('img_path', img_path[0])
            img_img = load_itk(img_path[0])   # 正常读取，sitk.ReadImage(path)

            # img_img = nibabel.load(img_path[0]).get_data()
            img_array = sitk.GetArrayFromImage(img_img)  # [channel, height, width,]
            img_array_copy = sitk.GetArrayFromImage(img_img)
            channel, height, width = img_array.shape  # [channel, h, w]
            for j in range(channel):
                image = img_array[j]
                image = cv2.resize(image, (480, 480))  # resize成网络训练的h*w
       
                image_x = image
                image_x = np.expand_dims(image_x, axis=-1)
                image_x = np.expand_dims(image_x, axis=0)   # 这两个是添加成[b, h ,w ,c]的维度
                feed_dict = {
                    image_placeholder:image_x,
                    training_flag:False,
                }
                result, idx_map = sess.run([logits, seg_image], feed_dict=feed_dict)
				# idx_map是输出的预测图[1, h, w] --上面为分类好的类