Li‘s 影像组学radiomics视频学习笔记(35)-基于2D超声影像的影像组学特征提

Good

作者：北欧森林
链接：https://www.jianshu.com/p/f82d30289d68
来源：简书，已获转载授权
  RadiomicsWorld.com “影像组学世界”论坛：
影像组学世界/RadiomicsWorld
本笔记来源于B站Up主: 有Li 的影像组学系列教学视频
本节(35)主要介绍: 2D超声影像组学的特征提取
　
　
视频中李博士情境再现了小白研究者可能碰到的各种技术难题，并演示了解决这些问题的思路。
1、将dicom格式的2D 超声图像转为压缩的nifti格式，将其命名为test.nii.gz; 勾画ROI后同样保存为压缩的nifti格式，命名为mask.nii.gz

2、尝试提取影像学特征：

1. import radiomics
2. from radiomics import featureextractor
4. imageFile = "/Users/Mac/Documents/JianShuNotes/2D_Ultrasound/test.nii.gz"
5. maskFile = "/Users/Mac/Documents/JianShuNotes/2D_Ultrasound/mask.nii.gz"
6. extractor = featureextractor.RadiomicsFeatureExtractor()
7. featureVector = extractor.execute(imageFile, maskFile)
8. #print(featureVector.items())
9. for featureName in featureVector.keys():
10.     print("%s: %s" % (featureName, featureVector[featureName]))

复制代码

出现了如下报错信息：
　
sitk::ERROR: Pixel type: vector of 8-bit unsigned integer is not supported in 3D byN3itk6simple26LabelStatisticsImageFilterE
3、按照提示，有可能是数据格式的问题。回到源代码，修改文件相应的格式，将其命名为 test1.nii.gz

1. import SimpleITK as sitk
2. import numpy as np
4. folderPath = "/Users/Mac/Documents/JianShuNotes/2D_Ultrasound/"
6. reader = sitk.ImageSeriesReader()
7. dicom_names = reader.GetGDCMSeriesFileNames(folderPath)
8. reader.SetFileNames(dicom_names)
9. image = reader.Execute()
11. image_arr = sitk.GetArrayFromImage(image) # Note: order:z, y, x !!
12. size = image.GetSize()
13. origin = image.GetOrigin() #order: x, y, z
14. spacing = image.GetSpacing() #order:x, y, z
15. direction = image.GetDirection()
16. #print(spacing)
18. pixelType = sitk.sitkInt8 #注意这里是Int8
19. image_new = sitk.Image(size,pixelType)
21. #image_arr_new = image_arr[:,:,::-1] #镜像翻转操作
22. image_arr_new = image_arr
23. #print(image_arr_new.shape)
25. image_new = sitk.GetImageFromArray(image_arr_new)
26. image_new.SetDirection(direction)
27. image_new.SetSpacing(spacing)
28. image_new.SetOrigin(origin)
29. sitk.WriteImage(image_new,folderPath + "test1.nii.gz")

复制代码

４、将步骤[2]中的test.nii.gz替换为test1.nii.gz，再次执行该步骤，此时又出现了报错信息：

sitk::ERROR: Input "labelImage" for "LabelStatisticsImageFilter" has dimension of 3 which does not match the primary input's dimension of 2!

5、执行 print(image_arr_new.shape) 后显示为：

1. (1, 900, 1600, 3)

复制代码

Notes: 这里涉及python读入nifti文件的格式问题，其顺序为 z,y,x (对应这里的1,900,1600); 这里的3为不同的slice

6、探索这个“3”代表的意义：将每一个slice导出来，通过对比发现其中的端倪。执行代码：

1. image_arr_new = image_arr[:,:,:,0]
2. #image_arr_new = image_arr[:,:,:,1]
3. #image_arr_new = image_arr[:,:,:,2]
4. #print(image_arr_new.shape)
6. image_new = sitk.GetImageFromArray(image_arr_new)
7. image_new.SetDirection(direction)
8. image_new.SetSpacing(spacing)
9. image_new.SetOrigin(origin)
10. sitk.WriteImage(image_new,folderPath + "test-slice0.nii.gz")
11. #sitk.WriteImage(image_new,folderPath + "test-slice1.nii.gz")
12. #sitk.WriteImage(image_new,folderPath + "test-slice2.nii.gz")

复制代码

7、将test-slice0.nii.gz，test-slice1.nii.gz，test-slice2.nii.gz在软件中打开，查看它们之间的差异(肉眼似乎看不出啥区别)


8、使用写代码的方法来探索test-slice0.nii.gz，test-slice1.nii.gz，test-slice2.nii.gz之间的差异(视频里李博士演示了对比前二者)

1. import SimpleITK as sitk
2. import numpy as np
4. slicer0 = sitk.ReadImage("test-slice0.nii.gz")
5. slicer1 = sitk.ReadImage("test-slice1.nii.gz")
7. slicer0_arr = sitk.GetArrayFromImage(slicer0)
8. slicer1_arr = sitk.GetArrayFromImage(slicer1)
10. comp = slicer0_arr == slicer1_arr
11. print(comp)

复制代码

9、将差异的部分保存为影像导出

1. image_arr = slicer0_arr - slicer1_arr
2. image_arr[image_arr != 0] = 1
3. 　
4. size = slicer0.GetSize()
5. origin = slicer0.GetOrigin() #order: x, y, z
6. spacing = slicer0.GetSpacing() #order:x, y, z
7. direction = slicer0.GetDirection()
8. 　
9. image_new = sitk.GetImageFromArray(image_arr)
10. image_new.SetDirection(direction)
11. image_new.SetSpacing(spacing)
12. image_new.SetOrigin(origin)
13. sitk.WriteImage(image_new,"comp.nii.gz")

复制代码

10、此时在软件中打开comp.nii.gz文件，查看差异(在于右下角的水印部分)(视频中是左下角的水印)。

换句话说，这3张test-slice0.nii.gz，test-slice1.nii.gz，test-slice2.nii.gz是一样的，差别仅在于右下角的水印。所以，后续的工作只需选择其中的一个即可。

11、再次查看test-slice0.nii.gz和mask文件，发现mask文件和源文件不对应。

mask.nii.gz是基于test.nii.gz文件画出并保存的，但是在test.nii.gz保存为test1.nii.gz时存在一个翻转(我没有搞明白这一点)，所以应重新保存一下mask文件

1. import SimpleITK as sitk
2. import numpy as np
4. folderPath = "/Users/Mac/Documents/JianShuNotes/2D_Ultrasound/"
5. mask = sitk.ReadImage('mask.nii.gz')
6. mask_arr = sitk.GetArrayFromImage(mask)
8. reader = sitk.ImageSeriesReader()
9. dicom_names = reader.GetGDCMSeriesFileNames(folderPath)
10. reader.SetFileNames(dicom_names)
11. image = reader.Execute()
13. image_arr = sitk.GetArrayFromImage(image) # Note: order:z, y, x !!
14. size = image.GetSize()
15. origin = image.GetOrigin() #order: x, y, z
16. spacing = image.GetSpacing() #order:x, y, z
17. direction = image.GetDirection()
19. pixelType = sitk.sitkInt8 #注意这里是Int8
20. image_new = sitk.Image(size,pixelType)
21. mask_new = sitk.Image(size,pixelType)
23. #image_arr_new = image_arr[:,:,::-1] #镜像翻转操作
24. image_arr_new = image_arr[:,:,:,0]
25. print(image_arr.shape)
27. image_new = sitk.GetImageFromArray(image_arr_new)
28. image_new.SetDirection(direction)
29. image_new.SetSpacing(spacing)
30. image_new.SetOrigin(origin)
32. mask_new = sitk.GetImageFromArray(mask_arr) # 视频里勘误为mask_new,其实就应该是mask_arr, print二者可以看出区别来
33. mask_new.SetDirection(direction)
34. mask_new.SetSpacing(spacing)
35. mask_new.SetOrigin(origin)
37. sitk.WriteImage(image_new,"test0.nii.gz")
38. sitk.WriteImage(mask_new,"mask0.nii.gz")

复制代码

12、再次尝试提取影像组学特征(代码[2])，发现代码可以运行。

但是这里有一个隐藏的bug，就是mask文件虽然重新保存了，但是并不能和原来的位置匹配。
通过软件查看图像可以发现其中的区别。

13、再次回到原始代码里查找问题来源

1. import SimpleITK as sitk
2. import numpy as np
4. folderPath = "/Users/Mac/Documents/JianShuNotes/2D_Ultrasound/"
5. mask = sitk.ReadImage('mask.nii.gz')
6. mask_arr = sitk.GetArrayFromImage(mask)
8. #reader = sitk.ImageSeriesReader()
9. #dicom_names = reader.GetGDCMSeriesFileNames(folderPath)
10. #reader.SetFileNames(dicom_names)
11. #image = reader.Execute()
13. image = sitk.ReadImage('test.nii.gz')
15. image_arr = sitk.GetArrayFromImage(image) # Note: order:z, y, x !!
16. size = image.GetSize()
17. origin = image.GetOrigin() #order: x, y, z
18. spacing = image.GetSpacing() #order:x, y, z
19. direction = image.GetDirection()
21. pixelType = sitk.sitkInt8 #注意这里是Int8
22. image_new = sitk.Image(size,pixelType)
23. mask_new = sitk.Image(size,pixelType)
25. #image_arr_new = image_arr[:,:,::-1] #镜像翻转操作
26. image_arr_new = image_arr[:,:,:,0]
27. print(image_arr.shape)
29. image_new = sitk.GetImageFromArray(image_arr_new)
30. image_new.SetDirection(direction)
31. image_new.SetSpacing(spacing)
32. image_new.SetOrigin(origin)
34. mask_new = sitk.GetImageFromArray(mask_arr)
35. mask_new.SetDirection(direction)
36. mask_new.SetSpacing(spacing)
37. mask_new.SetOrigin(origin)
39. sitk.WriteImage(image_new,"test0.nii.gz")
40. sitk.WriteImage(mask_new,"mask0.nii.gz")

复制代码

14、历尽周折，现在test0.nii.gz文件和mask0.nii.gz文件相匹配了！可以愉快地提取影像组学特征了。

1. import radiomics
2. from radiomics import featureextractor
4. imageFile = "/Users/Mac/Documents/JianShuNotes/2D_Ultrasound/test0.nii.gz"
5. maskFile = "/Users/Mac/Documents/JianShuNotes/2D_Ultrasound/mask0.nii.gz"
6. extractor = featureextractor.RadiomicsFeatureExtractor()
7. featureVector = extractor.execute(imageFile, maskFile)
8. #print(featureVector.items())
9. for featureName in featureVector.keys():
10.     print("%s: %s" % (featureName, featureVector[featureName]))

复制代码

来源：https://blog.csdn.net/weixin_38594676/article/details/113144043
免责声明：如果侵犯了您的权益，请联系站长，我们会及时删除侵权内容，谢谢合作！