将 3-D MRI 图像轴转置到不同的图像视图

Question

如何使用 SimpleITK 将 3D MRI 图像转置到新的图像视图，但保持初始视图中的纵横比？这是我的代码：

def show_n_slices(array, start_from=0, step=10, columns=6, figsize=(18,10)):
    """ Plot N slices of a 3D image.
    :param array: N-dimensional numpy array (i.e. 3D image)
    :param start_from: slice index to start from
    :param step: step to take when moving to the next slice
    :param columns: number of columns in the plot
    :param figsize: figure size in inches
    """
    array = np.swapaxes(array, 0, 2)
    fig, ax = plt.subplots(1, columns, figsize=figsize)
    slice_num = start_from
    for n in range(columns):
        ax[n].imshow(array[:, :, slice_num], 'gray')
        ax[n].set_xticks([])
        ax[n].set_yticks([])
        ax[n].set_title('Slice number: {}'.format(slice_num), color='r')
        slice_num += step

    fig.subplots_adjust(wspace=0, hspace=0)
    plt.show()

def print_sitk_info(itk_image):
    """ Prints SimpleITK image information
    :param itk_image: SimpleITK image object.
    """
    print(f"[INFO]: Shape - {itk_image.GetSize()}")
    print(f"[INFO]: Spacing - {itk_image.GetSpacing()}")
    print(f"[INFO]: Origin - {itk_image.GetOrigin()}")
    print(f"[INFO]: Direction - {itk_image.GetDirection()}\n")

print("[INFO]: Image Info before Resampling to Isotropic Resolution:")
print_sitk_info(itk_image)
show_n_slices(sitk.GetArrayFromImage(itk_image), start_from=20, step=5)

这是我尝试转置图像轴以获得不同图像视图的代码：

array2 = array.transpose(2, 0 ,1)
itk_image2 = sitk.GetImageFromArray(array2)
itk_image2.SetOrigin(itk_image.GetOrigin())
itk_image2.SetSpacing(itk_image.GetSpacing())
itk_image2.SetDirection(itk_image.GetDirection())
array2 = sitk.GetArrayFromImage(itk_image2)

print_sitk_info(itk_image2)
show_n_slices(array2, start_from=30, step=5)

我在这里做错了什么？接下来我需要重新采样图像大小吗？

编辑 [基于 Dave Chens 的回答]：

按正确顺序添加间距的结果：

array2 = array.transpose(2, 0 ,1)
spacing = itk_image.GetSpacing()
itk_image2 = sitk.GetImageFromArray(array2)
itk_image2.SetOrigin(itk_image.GetOrigin())
itk_image2.SetSpacing([spacing[2], spacing[0], spacing[1]])
itk_image2.SetDirection(itk_image.GetDirection())
array2 = sitk.GetArrayFromImage(itk_image2)

print_sitk_info(itk_image2)
show_n_slices(array2, start_from=30, step=5)

SimpleITK 的结果PermuteAxesImageFilter:

pa = sitk.PermuteAxesImageFilter()
pa.SetOrder([2, 0 , 1])
i = pa.Execute(itk_image)
print_sitk_info(i)
show_n_slices(sitk.GetArrayFromImage(i), start_from=30, step=5)

Answer 1

问题在于您使用的是原始图像的像素间距。您需要以与转置像素相同的方式转置像素间距。所以你会做这样的事情：

spacing = itk_image.GetSpacing()
spacing2 = [spacing[2], spacing[0], spacing[1]]
itk_image2.SetSpacing(spacing2)

另外，请注意 SimpleITK 有一个 PermuteAxes 函数，它可以做同样的事情，并且会保留图像元数据，所以你不必复制它。

更新：好的，下面是我使用的代码。我在 Jupyter Notebook 中运行它并使用 ITKWidgets 在笔记本中渲染 3d 图像。请注意，底部的每个视图调用都应位于其自己的笔记本单元格中。

import SimpleITK as sitk
import itkwidgets

fnames = []
for i in range(1,33):
    name = "Image-"+str(i)+".dcm"
    fnames.append(name)
reader = sitk.ImageSeriesReader()
reader.SetFileNames(fnames)
img = reader.Execute()
print(img)

pa = sitk.PermuteAxesImageFilter()
pa.SetOrder([2,0,1])
img2 = pa.Execute(img)
print(img2)

# Render cell #1
itkwidgets.view(img)

# Render cell #2
itkwidgets.view(img2)