nii convert to 2D image【python】

可以自己精简，我的label是二分类

import SimpleITK as sitk
import cv2
from PIL import Image
import numpy as np
import nibabel as nib  # nii格式一般都会用到这个包
import imageio  # 转换成图像
import os

import numpy as np
from scipy.ndimage import rotate
from scipy.ndimage import median_filter
import matplotlib.pyplot as plt



xy = 128
vol1 = int(xy/2)
vol2 = int(xy/4)
vol3 = int(vol2+16)

def preprocess(image):
    result = median_filter(image, size=3)

    """# 添加高斯噪声
    noise = np.random.normal(0, 25, size=image.shape)
    noise_img = image + noise.astype('uint8')

    # 双边滤波
    result = cv2.bilateralFilter(noise_img, 9, 75, 75)

    # 显示图像
    cv2.imshow('src', image)
    cv2.imshow('noise', noise_img)
    cv2.imshow('result', result)
    cv2.waitKey()
    cv2.destroyAllWindows()"""

    return result


def create_dirs(out_path, num):
    for i in range(1, num):  # 这里需要注意，i取不到6，因为range()是前闭后开的，即i的取值范围为1-5。

        dir = os.path.join(r'crop/test4/ct/')
        # 前者为路径，后者为待创建文件夹的名称。注意，批量创建文件夹时不能有重复名称的，因此可以对文件夹加上序号信息。

        isExists = os.path.exists(dir)
        if not isExists:
            os.mkdir(dir)

def mask2d(input_path, output_folder, idx):
    # 加载NIfTI文件
    img = nib.load(input_path)
    data = img.get_fdata()

    # 获取数据的形状信息
    num_slices = data.shape[2]  # 切片数量
    print(data.shape, num_slices)

    # 遍历每个切片并保存为PNG图像
    for i in range(num_slices):
        slice_data = data[:, :, i]  # 提取当前切片数据
        # Image.fromarray(255*img_array_list[foo].astype('int')).convert('L')
        image = Image.fromarray(255 * slice_data.astype('int')).convert('L')
        image = image.rotate(270)
        # image = Image.fromarray(slice_data)  # 创建PIL图像对象

        output_name = f"{output_folder}slice_{idx}_{i}.png"  # 设置输出文件名
        image.save(output_name)  # 保存为PNG图像



def nii2d(img_addr, target_folder, idx):

    img_addr_n = nib.load(img_addr)
    # Convert them to numpy format,
    data = img_addr_n.get_fdata()

    # clip the images within [-125, 275],
    data_clipped = np.clip(data, -125, 275)

    # normalize each 3D image to [0, 1], and
    data_normalised = (data_clipped - (-125)) / (275 - (-125))

    split_root = img_addr.split('\\')  # 通过\\来进行截断
    print(split_root) # ['crop/test4/ct/volume-0.nii']

    # extract 2D slices from 3D volume for training cases while
    # e.g. slice 000
    for i in range(data.shape[2]):
        formattedi = "{:03d}".format(i)
        slice000 = data_normalised[:, :, i] * 255
        image = Image.fromarray(slice000)
        image = image.convert("L")
        image = image.rotate(270)
        image = image.transpose(Image.Transpose.FLIP_LEFT_RIGHT)

        image.save(target_folder +str(idx)+"-"+str(i)+ ".png")





for i in range(20):
    k = i + 1
    image_path = "ct/volume-{}.nii".format(str(i))
    label_path = "label/segmentation-{}.nii.gz".format(str(i))

    label = sitk.ReadImage(label_path, sitk.sitkInt16)
    label_array = sitk.GetArrayFromImage(label)

    image = sitk.ReadImage(image_path, sitk.sitkInt32)
    image_array = sitk.GetArrayFromImage(image)  # 分别读图像和标签数据
    print("\nimage_array=",image_array.shape, " label_array=",label_array.shape)

    center_x = (image_array.shape[1]) // 2
    center_y = (image_array.shape[2]) // 2
    center_z = (image_array.shape[0]) / 2  # 分别计算出xyz方向上的中心
    print("center_x=", center_x, "center_y=", center_y, "center_z=",center_z)

    center_x = center_x - vol2

    image_array = image_array[:, center_x - vol3:center_x + vol3, center_y - xy:center_y + xy]
    label_array = label_array[:, center_x - vol3:center_x + vol3, center_y - xy:center_y + xy]  # 在XY裁剪出一个256 * 256的区域

    #####只需要保存有标签的序列就行了
    z = np.any(label_array, axis=(1, 2))
    start_slice, end_slice = np.where(z)[0][[0, -1]]

    # 截取保留区域
    image_array = image_array[start_slice:end_slice + 1, :, :]
    label_array = label_array[start_slice:end_slice + 1, :, :]
    # print("Preprocessed shape:",ct_array.shape,seg_array.shape)

    new_image = sitk.GetImageFromArray(image_array)
    new_image.SetDirection(image.GetDirection())
    new_image.SetOrigin(image.GetOrigin())
    new_image.SetSpacing(image.GetSpacing())

    new_seg = sitk.GetImageFromArray(label_array)
    new_seg.SetDirection(label.GetDirection())
    new_seg.SetOrigin(label.GetOrigin())
    new_seg.SetSpacing(label.GetSpacing())

    sitk.WriteImage(new_image, "crop/test4/ct/volume-{}.nii".format(str(i)))
    sitk.WriteImage(new_seg, "crop/test4/label/segmentation-{}.nii.gz".format(str(i)))

    nii2d("crop/test4/ct/volume-" + str(i) + ".nii", "crop/png_ct/", i)

    mask2d("crop/test4/label/segmentation-" + str(i) + ".nii.gz", "crop/png_label/", i)

    """对于label来说是ok的，但是对于那个来说不行"""