通用约定

image.pngimage.png
image.png

图像操作

  1. img1 = sitk.Image(24,24, sitk.sitkUInt8)
  2. img1[0,0] = 0
  4. img2 = sitk.Image(img1.GetSize(), sitk.sitkUInt8)
  5. img2.SetDirection([0,1,0.5,0.5])
  6. img2.SetSpacing([0.5,0.8])
  7. img2.SetOrigin([0.000001,0.000001])
  8. img2[0,0] = 255
  10. img3 = img1 + img2
  11. print(img3[0,0])

访问和索引切片

Slicing of SimpleITK images returns a copy of the image data.This is similar to slicing Python lists and differs from the “view” returned by slicing numpy arrays.
image.png

numpy的和SimpleITK之间的转换：

  1. SimpleITK和numpy的索引访问的顺序相反！
  2. SimpleITK：图像[X，Y，Z]
  3. numpy的：image_numpy_array [Z，Y，X]

1. 从SimplTK到numpy

image.png

2. 从numpy到SimpleITK

image.png

3. 区别于scikit-image和numpy的转换：

image.png

读写图像

  1. SimpleITK可以读写存储在单个文件或一组文件（例如DICOM系列）中的图像。
  3. 以DICOM格式存储的图像具有与之关联的元数据字典，其中包括了DICOM标签。
  4. 当将DICOM系列被读为单个图像时，元数据信息不可用，因为DICOM标签是针对每个单独文件的。
  5. 如果需要元数据，可以使用面向对象的接口的ImageSeriesReader类，或者ImageFileReader类，
  6. 设置特定切片的文件名，访问元数据图像使用GetMetaDataKeys，HasMetaDataKey和GetMetaData。

例如，读取DICOM格式的图像，并将其写为mhd。文件格式由文件扩展名推导出。还设置了适当的像素类型-您可以覆盖它并强制选择您要的像素类型。

  1. data_directory = os.path.dirname(fdata("CIRS057A_MR_CT_DICOM/readme.txt"))
  2. series_ID = '1.2.840.113619.2.290.3.3233817346.783.1399004564.515'
  4. # Get the list of files belonging to a specific series ID.
  5. reader = sitk.ImageSeriesReader()
  6. # Use the functional interface to read the image series.
  7. original_image = sitk.ReadImage(reader.GetGDCMSeriesFileNames(data_directory, series_ID))
  9. # Write the image.
  10. output_file_name_3D = os.path.join(OUTPUT_DIR, '3DImage.mha')
  11. sitk.WriteImage(original_image, output_file_name_3D)
  13. # Read it back again.
  14. written_image = sitk.ReadImage(output_file_name_3D)
  16. # Check that the original and written image are the same.
  17. statistics_image_filter = sitk.StatisticsImageFilter()
  18. statistics_image_filter.Execute(original_image - written_image)
  20. # Check that the original and written files are the same
  21. print('Max, Min differences are : {0}, {1}'.format(statistics_image_filter.GetMaximum(), 
  22.                                                    statistics_image_filter.GetMinimum()))

Fetching CIRS057A_MR_CT_DICOM/readme.txt
Max, Min differences are : 0.0, 0.0
将图像写为jpg格式，并且需要重新调整图像强度（默认为[0,255]），因为JPEG格式的显示需要转换为UInt8像素类型。

  1. sitk.WriteImage(sitk.Cast(sitk.RescaleIntensity(written_image), sitk.sitkUInt8), 
  2.                 [os.path.join(OUTPUT_DIR, 'slice{0:03d}.jpg'.format(i)) 
  3.                  for i in range(written_image.GetSize()[2])])

图像显示

  1. mr_image = sitk.ReadImage(fdata('training_001_mr_T1.mha'))
  2. npa = sitk.GetArrayViewFromImage(mr_image)
  4. # Display the image slice from the middle of the stack, z axis
  5. z = int(mr_image.GetDepth()/2)
  6. npa_zslice = sitk.GetArrayViewFromImage(mr_image)[z,:,:]
  8. # Three plots displaying the same data, how do we deal with the high dynamic range?
  9. fig = plt.figure()
  10. fig.set_size_inches(15,30)
  12. fig.add_subplot(1,3,1)
  13. plt.imshow(npa_zslice)
  14. plt.title('default colormap')
  15. plt.axis('off')
  17. fig.add_subplot(1,3,2)
  18. plt.imshow(npa_zslice,cmap=plt.cm.Greys_r);
  19. plt.title('grey colormap')
  20. plt.axis('off')
  22. fig.add_subplot(1,3,3)
  23. plt.title('grey colormap,\n scaling based on volumetric min and max values')
  24. plt.imshow(npa_zslice,cmap=plt.cm.Greys_r, vmin=npa.min(), vmax=npa.max())
  25. plt.axis('off');

image.png

参考：

【1】SimpleITK图像基础
【2】SimpleITK-Notebooks-03_Image_Details
【3】scikit-image中的numpy速成
【4】SITK Documentation docs