Nii转png，Nii转dcm，Dcm转png
Nii转png，Nii转dcm，Dcm转png
- 前言
- Nii转PNG
- Nii转DCM
- DCM转PNG
- 完整代码
- 结束语
前言
这篇文章主要是对之前使用的nii文件进行一些格式的转换，nii文件是之前用的LITS数据集，因为需要将其转为png图片和DICOM格式的二维图片。原理基本一致，先把需要转换的文件读取成numpy矩阵，然后再进行转化，主要是保存的时候更改后缀名而已。
Nii转PNG
先将nii中的每个切片读取进来，然后再一张张切片分开保存，代码如下：
单个nii转png
def nii2png_single(nii_path, IsData = True):
ori_data = sitk.ReadImage(nii_path) # 读取一个数据
data1 = sitk.GetArrayFromImage(ori_data) # 获取数据的array
if IsData: #过滤掉其他无关的组织，标签不需要这步骤
data1[data1 > 250] = 250
data1[data1 < -250] = -250
img_name = os.path.split(nii_path) #分离文件名
img_name = img_name[-1]
img_name = img_name.split('.')
img_name = img_name[0]
i = data1.shape[0]
png_path = './png/single_png' #图片保存位置
if not os.path.exists(png_path):
os.makedirs(png_path)
for j in range(0, i): #将每一张切片都转为png
if IsData: # 数据
#归一化
slice_i = (data1[j, :, :] - data1[j, :, :].min()) / (data1[j, :, :].max() - data1[j, :, :].min()) * 255
cv2.imwrite("%s/%s-%d.png" % (png_path, img_name, j), slice_i) #保存
else: # 标签
slice_i = data1[j, :, :] * 122
cv2.imwrite("%s/%s-%d.png" % (png_path, img_name, j), slice_i) # 保存
批量转换
def nii2png(data_path, label_path):
dataname_list2 = os.listdir(label_path)
for nii_label in dataname_list2:
print("process:", nii_label)
data1 = sitk.ReadImage(os.path.join(data_path, nii_label.replace('segmentation', 'volume'))) # 读取一个数据
liver_tumor_data = sitk.GetArrayFromImage(data1) # 获取数据的array
data2 = sitk.ReadImage(os.path.join(label_path, nii_label)) # 读取一个数据
liver_tumor_label = sitk.GetArrayFromImage(data2) # 获取数据的array
img_name = os.path.split(nii_label.replace('segmentation', 'volume')) # 分离文件名，用以保存时的文件名前缀
img_name = img_name[-1]
img_name = img_name.split('.')
img_name = img_name[0]
i = liver_tumor_label.shape[0]
liver_tumor_data[liver_tumor_data > 250] = 250
liver_tumor_data[liver_tumor_data < -250] = -250
outpath = r'./png/data_png' # 数据保存文件夹
outpath2 = r'./png/label_png' # 标签保存文件夹
if not os.path.exists(outpath):
os.makedirs(outpath)
if not os.path.exists(outpath2):
os.makedirs(outpath2)
for j in range(0, i):
if liver_tumor_label[j, :, :].max() > 0: # 只保存有肝脏的切片
img1 = (liver_tumor_data[j, :, :] - liver_tumor_data[j, :, :].min()) / (liver_tumor_data[j, :, :].max() - liver_tumor_data[j, :, :].min()) * 255
img2 = liver_tumor_label[j, :, :] * 122
cv2.imwrite("%s/%s-%d.png" % (outpath, img_name, j), img1)
cv2.imwrite("%s/%s-%d.png" % (outpath2, img_name, j), img2)
Nii转DCM
注意，因为原来的nii文件中就没有DICOM文件中特有的头文件信息，所以转换出来的DICOM文件实际上是缺少了头文件的信息的，算是伪DICOM文件，在使用的时候个人感觉并没有什么不对劲的地方，DICOM文件转为PNG图片显示出来的效果基本和Nii转PNG的效果一致，但是要特别注意的是读取DICOM文件的时候要强制读取。
单个Nii转DCM
def nii2dcm_single(nii_path, IsData = True):
save_folder = './dcm/dcm_single'
if not os.path.exists(save_folder):
os.makedirs(save_folder)
ori_data = sitk.ReadImage(nii_path) # 读取一个数据
data1 = sitk.GetArrayFromImage(ori_data) # 获取数据的array
if IsData: # 过滤掉其他无关的组织，标签不需要这步骤
data1[data1 > 250] = 250
data1[data1 < -250] = -250
img_name = os.path.split(nii_path) #分离文件名
img_name = img_name[-1]
img_name = img_name.split('.')
img_name = img_name[0]
i = data1.shape[0]
# 关键部分
castFilter = sitk.CastImageFilter()
castFilter.SetOutputPixelType(sitk.sitkInt16)
for j in range(0, i): #将每一张切片都转为png
if IsData: # 数据
slice_i = data1[j, :, :]
data_img = sitk.GetImageFromArray(slice_i)
# Convert floating type image (imgSmooth) to int type (imgFiltered)
data_img = castFilter.Execute(data_img)
sitk.WriteImage(data_img, "%s/%s-%d.dcm" % (save_folder, img_name, j))
else: # 标签
slice_i = data1[j, :, :] * 122
label_img = sitk.GetImageFromArray(slice_i)
# Convert floating type image (imgSmooth) to int type (imgFiltered)
label_img = castFilter.Execute(label_img)
sitk.WriteImage(label_img, "%s/%s-%d.dcm" % (save_folder, img_name, j))
批量转换
def nii2dcm(data_path, lable_path, IsTrain = False):
save_folder = './dcm'
nii_list = os.listdir(lable_path)
for nii_label in nii_list:
print("process:", nii_label)
data1 = sitk.ReadImage(os.path.join(lable_path, nii_label)) # 读取一个数据
liver_tumor_label = sitk.GetArrayFromImage(data1) # 获取数据的array
print(np.unique(liver_tumor_label))
data2 = sitk.ReadImage(os.path.join(data_path, nii_label.replace('segmentation', 'volume'))) # 读取一个数据
liver_tumor_data = sitk.GetArrayFromImage(data2) # 获取数据的array
slice_num = liver_tumor_label.shape[0] # 获取切片数量
if IsTrain:
save_img = save_folder+'/train'
else:
save_img = save_folder + '/val'
label_path = save_folder+'/label' # 设置保存路径
if not os.path.exists(save_img): # 判断文件目录是否存在，不存在则创建
os.makedirs(save_img)
if not os.path.exists(label_path):
os.makedirs(label_path)
img_name = os.path.split(nii_label.replace('segmentation', 'volume')) # 分离文件名，用以保存时的文件名前缀
img_name = img_name[-1]
img_name = img_name.split('.')
img_name = img_name[0]
for i in range(0, slice_num):
# 关键部分
if liver_tumor_label[i, :, :].max() > 0: # 肝脏标签图片
label_img = sitk.GetImageFromArray(liver_tumor_label[i, :, :] * 122)
data_img = sitk.GetImageFromArray(liver_tumor_data[i, :, :])
castFilter = sitk.CastImageFilter()
castFilter.SetOutputPixelType(sitk.sitkInt16)
# Convert floating type image (imgSmooth) to int type (imgFiltered)
label_img = castFilter.Execute(label_img)
data_img = castFilter.Execute(data_img)
sitk.WriteImage(label_img, "%s/%s-%d.dcm" % (label_path, img_name, i))
sitk.WriteImage(data_img, "%s/%s-%d.dcm" % (save_img, img_name, i))
DCM转PNG
单个DCM转PNG
def dcm2png_single(dcm_path):
save_pa = r'./png/single'
if not os.path.exists(save_pa):
os.makedirs(save_pa)
img_name = os.path.split(dcm_path.replace('.dcm', '.png')) # 替换后缀名并分离路径
img_name = img_name[-1]
ds = pydicom.read_file(dcm_path, force=True) # 注意这里，强制读取
img = ds.pixel_array # 提取图像信息
cv2.imwrite(os.path.join(save_pa, img_name), img)
批量转换
def dcm2png(dcm_path, label_path):
labellist = os.listdir(label_path)
save_data = r'./png/dcm_png_data'
save_lab = r'./png/dcm_png_lab'
if not os.path.exists(save_data):
os.makedirs(save_data)
if not os.path.exists(save_lab):
os.makedirs(save_lab)
for lab in labellist:
print('process', lab)
ds = pydicom.read_file(os.path.join(label_path, lab), force=True) # 注意这里，强制读取
label = ds.pixel_array # 提取图像信息
ds1 = pydicom.read_file(os.path.join(dcm_path, lab), force=True)
img = ds1.pixel_array # 提取图像信息
cv2.imwrite(os.path.join(save_data, lab.replace('.dcm', '.png')), img)
cv2.imwrite(os.path.join(save_lab, lab.replace('.dcm', '.png')), label)
完整代码
import os
import SimpleITK as sitk
import cv2
import pydicom
import numpy as np
def nii2png_single(nii_path, IsData = True):
ori_data = sitk.ReadImage(nii_path) # 读取一个数据
data1 = sitk.GetArrayFromImage(ori_data) # 获取数据的array
if IsData: #过滤掉其他无关的组织，标签不需要这步骤
data1[data1 > 250] = 250
data1[data1 < -250] = -250
img_name = os.path.split(nii_path) #分离文件名
img_name = img_name[-1]
img_name = img_name.split('.')
img_name = img_name[0]
i = data1.shape[0]
png_path = './png/single_png' #图片保存位置
if not os.path.exists(png_path):
os.makedirs(png_path)
for j in range(0, i): #将每一张切片都转为png
if IsData: # 数据
#归一化
slice_i = (data1[j, :, :] - data1[j, :, :].min()) / (data1[j, :, :].max() - data1[j, :, :].min()) * 255
cv2.imwrite("%s/%s-%d.png" % (png_path, img_name, j), slice_i) #保存
else: # 标签
slice_i = data1[j, :, :] * 122
cv2.imwrite("%s/%s-%d.png" % (png_path, img_name, j), slice_i) # 保存
def nii2png(data_path, label_path):
dataname_list2 = os.listdir(label_path)
for nii_label in dataname_list2:
print("process:", nii_label)
data1 = sitk.ReadImage(os.path.join(data_path, nii_label.replace('segmentation', 'volume'))) # 读取一个数据
liver_tumor_data = sitk.GetArrayFromImage(data1) # 获取数据的array
data2 = sitk.ReadImage(os.path.join(label_path, nii_label)) # 读取一个数据
liver_tumor_label = sitk.GetArrayFromImage(data2) # 获取数据的array
img_name = os.path.split(nii_label.replace('segmentation', 'volume')) # 分离文件名，用以保存时的文件名前缀
img_name = img_name[-1]
img_name = img_name.split('.')
img_name = img_name[0]
i = liver_tumor_label.shape[0]
liver_tumor_data[liver_tumor_data > 250] = 250
liver_tumor_data[liver_tumor_data < -250] = -250
outpath = r'./png/data_png' # 数据保存文件夹
outpath2 = r'./png/label_png' # 标签保存文件夹
if not os.path.exists(outpath):
os.makedirs(outpath)
if not os.path.exists(outpath2):
os.makedirs(outpath2)
for j in range(0, i):
if liver_tumor_label[j, :, :].max() > 0: # 只保存有肝脏的切片
img1 = (liver_tumor_data[j, :, :] - liver_tumor_data[j, :, :].min()) / (liver_tumor_data[j, :, :].max() - liver_tumor_data[j, :, :].min()) * 255
img2 = liver_tumor_label[j, :, :] * 122
cv2.imwrite("%s/%s-%d.png" % (outpath, img_name, j), img1)
cv2.imwrite("%s/%s-%d.png" % (outpath2, img_name, j), img2)
def nii2dcm_single(nii_path, IsData = True):
save_folder = './dcm/dcm_single'
if not os.path.exists(save_folder):
os.makedirs(save_folder)
ori_data = sitk.ReadImage(nii_path) # 读取一个数据
data1 = sitk.GetArrayFromImage(ori_data) # 获取数据的array
if IsData: # 过滤掉其他无关的组织，标签不需要这步骤
data1[data1 > 250] = 250
data1[data1 < -250] = -250
img_name = os.path.split(nii_path) #分离文件名
img_name = img_name[-1]
img_name = img_name.split('.')
img_name = img_name[0]
i = data1.shape[0]
# 关键部分
castFilter = sitk.CastImageFilter()
castFilter.SetOutputPixelType(sitk.sitkInt16)
for j in range(0, i): #将每一张切片都转为png
if IsData: # 数据
slice_i = data1[j, :, :]
data_img = sitk.GetImageFromArray(slice_i)
# Convert floating type image (imgSmooth) to int type (imgFiltered)
data_img = castFilter.Execute(data_img)
sitk.WriteImage(data_img, "%s/%s-%d.dcm" % (save_folder, img_name, j))
else: # 标签
slice_i = data1[j, :, :] * 122
label_img = sitk.GetImageFromArray(slice_i)
# Convert floating type image (imgSmooth) to int type (imgFiltered)
label_img = castFilter.Execute(label_img)
sitk.WriteImage(label_img, "%s/%s-%d.dcm" % (save_folder, img_name, j))
def nii2dcm(data_path, lable_path, IsTrain = False):
save_folder = './dcm'
nii_list = os.listdir(lable_path)
for nii_label in nii_list:
print("process:", nii_label)
data1 = sitk.ReadImage(os.path.join(lable_path, nii_label)) # 读取一个数据
liver_tumor_label = sitk.GetArrayFromImage(data1) # 获取数据的array
print(np.unique(liver_tumor_label))
data2 = sitk.ReadImage(os.path.join(data_path, nii_label.replace('segmentation', 'volume'))) # 读取一个数据
liver_tumor_data = sitk.GetArrayFromImage(data2) # 获取数据的array
slice_num = liver_tumor_label.shape[0] # 获取切片数量
if IsTrain:
save_img = save_folder+'/train'
else:
save_img = save_folder + '/val'
label_path = save_folder+'/label' # 设置保存路径
if not os.path.exists(save_img): # 判断文件目录是否存在，不存在则创建
os.makedirs(save_img)
if not os.path.exists(label_path):
os.makedirs(label_path)
img_name = os.path.split(nii_label.replace('segmentation', 'volume')) # 分离文件名，用以保存时的文件名前缀
img_name = img_name[-1]
img_name = img_name.split('.')
img_name = img_name[0]
for i in range(0, slice_num):
# 关键部分
if liver_tumor_label[i, :, :].max() > 0: # 肝脏标签图片
label_img = sitk.GetImageFromArray(liver_tumor_label[i, :, :] * 122)
data_img = sitk.GetImageFromArray(liver_tumor_data[i, :, :])
castFilter = sitk.CastImageFilter()
castFilter.SetOutputPixelType(sitk.sitkInt16)
# Convert floating type image (imgSmooth) to int type (imgFiltered)
label_img = castFilter.Execute(label_img)
data_img = castFilter.Execute(data_img)
sitk.WriteImage(label_img, "%s/%s-%d.dcm" % (label_path, img_name, i))
sitk.WriteImage(data_img, "%s/%s-%d.dcm" % (save_img, img_name, i))
def dcm2png_single(dcm_path):
save_pa = r'./png/single'
if not os.path.exists(save_pa):
os.makedirs(save_pa)
img_name = os.path.split(dcm_path.replace('.dcm', '.png')) # 替换后缀名并分离路径
img_name = img_name[-1]
ds = pydicom.read_file(dcm_path, force=True) # 注意这里，强制读取
img = ds.pixel_array # 提取图像信息
cv2.imwrite(os.path.join(save_pa, img_name), img)
def dcm2png(dcm_path, label_path):
labellist = os.listdir(label_path)
save_data = r'./png/dcm_png_data'
save_lab = r'./png/dcm_png_lab'
if not os.path.exists(save_data):
os.makedirs(save_data)
if not os.path.exists(save_lab):
os.makedirs(save_lab)
for lab in labellist:
print('process', lab)
ds = pydicom.read_file(os.path.join(label_path, lab), force=True) # 注意这里，强制读取
label = ds.pixel_array # 提取图像信息
ds1 = pydicom.read_file(os.path.join(dcm_path, lab), force=True)
img = ds1.pixel_array # 提取图像信息
cv2.imwrite(os.path.join(save_data, lab.replace('.dcm', '.png')), img)
cv2.imwrite(os.path.join(save_lab, lab.replace('.dcm', '.png')), label)
if __name__ == "__main__":
# 单个nii文件转png
nii_single = r'G:\Python\liversegment\ImageResource\ImageTraning\data\volume-0.nii'
nii2png_single(nii_single, IsData=False)
# 按照标签提取切片，只保存有肝脏的切片
niipath = r'G:\Python\liversegment\ImageResource\ImageTraning\data'
labpath = r'G:\Python\liversegment\ImageResource\ImageTraning\label'
nii2png(niipath, labpath)
# 单个nii转dcm
nii2dcm_single(nii_single, True)
# 批量nii转dcm
nii2dcm(niipath, labpath, IsTrain=True)
# 单个dcm转png
dcm_single = r'F:\Liver\Unet_4_23\dcm\train\volume-0-55.dcm'
dcm2png_single(dcm_single)
# 批量dcm转png
dcm_pa = r'F:\Liver\Unet_4_23\dcm\train'
lab_dcm = r'F:\Liver\Unet_4_23\dcm\label'
dcm2png(dcm_pa, lab_dcm)
结束语
代码基本上都是自己从网上找到，然后根据自己的实际来修改使用的，代码全部都测试过，自己用的时候没有问题。本人的水平有限，如有不当之处，敬请指正。