SimpleITK 图像配准

SimpleITK 图像配准

在网上找的资源，效果不佳，等清楚了函数和原理再细改，调试效果。

  1 # -*- coding : UTF-8 -*-
  2 # @file   : regist.py
  3 # @Time   : 2021-11-12 17:00
  4 # @Author : wmz
  5 
  6 import SimpleITK as sitk
  7 
  8 # Utility method that either downloads data from the MIDAS repository or
  9 # if already downloaded returns the file name for reading from disk (cached data).
 10 # %run update_path_to_download_script
 11 # from downloaddata import fetch_data as fdata
 12 
 13 # Always write output to a separate directory, we don't want to pollute the source directory.
 14 import os
 15 OUTPUT_DIR = 'Output'
 16 
 17 import matplotlib.pyplot as plt
 18 # % matplotlib
 19 # inline
 20 
 21 from ipywidgets import interact, fixed
 22 from IPython.display import clear_output
 23 
 24 
 25 # Callback invoked by the interact IPython method for scrolling through the image stacks of
 26 # the two images (moving and fixed).
 27 def display_images(fixed_image_z, moving_image_z, fixed_npa, moving_npa):
 28     # Create a figure with two subplots and the specified size.
 29     plt.subplots(1, 2, figsize=(10, 8))
 30 
 31     # Draw the fixed image in the first subplot.
 32     plt.subplot(1, 2, 1)
 33     plt.imshow(fixed_npa[fixed_image_z, :, :], cmap=plt.cm.Greys_r);
 34     plt.title('fixed image')
 35     plt.axis('off')
 36 
 37     # Draw the moving image in the second subplot.
 38     plt.subplot(1, 2, 2)
 39     plt.imshow(moving_npa[moving_image_z, :, :], cmap=plt.cm.Greys_r);
 40     plt.title('moving image')
 41     plt.axis('off')
 42 
 43     plt.show()
 44 
 45 
 46 # Callback invoked by the IPython interact method for scrolling and modifying the alpha blending
 47 # of an image stack of two images that occupy the same physical space.
 48 def display_images_with_alpha(image_z, alpha, fixed, moving):
 49     img = (1.0 - alpha) * fixed[:, :, image_z] + alpha * moving[:, :, image_z]
 50     plt.imshow(sitk.GetArrayViewFromImage(img), cmap=plt.cm.Greys_r);
 51     plt.axis('off')
 52     plt.show()
 53 
 54 
 55 # Callback invoked when the StartEvent happens, sets up our new data.
 56 def start_plot():
 57     global metric_values, multires_iterations
 58 
 59     metric_values = []
 60     multires_iterations = []
 61 
 62 
 63 # Callback invoked when the EndEvent happens, do cleanup of data and figure.
 64 def end_plot():
 65     global metric_values, multires_iterations
 66 
 67     del metric_values
 68     del multires_iterations
 69     # Close figure, we don't want to get a duplicate of the plot latter on.
 70     plt.close()
 71 
 72 
 73 # Callback invoked when the IterationEvent happens, update our data and display new figure.
 74 def plot_values(registration_method):
 75     global metric_values, multires_iterations
 76 
 77     metric_values.append(registration_method.GetMetricValue())
 78     # Clear the output area (wait=True, to reduce flickering), and plot current data
 79     clear_output(wait=True)
 80     # Plot the similarity metric values
 81     plt.plot(metric_values, 'r')
 82     plt.plot(multires_iterations, [metric_values[index] for index in multires_iterations], 'b*')
 83     plt.xlabel('Iteration Number', fontsize=12)
 84     plt.ylabel('Metric Value', fontsize=12)
 85     plt.show()
 86 
 87 
 88 # Callback invoked when the sitkMultiResolutionIterationEvent happens, update the index into the
 89 # metric_values list.
 90 def update_multires_iterations():
 91     global metric_values, multires_iterations
 92     multires_iterations.append(len(metric_values))
 93 
 94 
 95 fixed_image =  sitk.ReadImage(r"E:\Data\right_knee\01-YCQ-cl-L.nrrd", sitk.sitkFloat32)
 96 moving_image = sitk.ReadImage(r"E:\Data\right_knee\02-LAXI-r-cl lv.nrrd", sitk.sitkFloat32)
 97 
 98 interact(display_images, fixed_image_z=(0,fixed_image.GetSize()[2]-1), moving_image_z=(0,moving_image.GetSize()[2]-1), fixed_npa = fixed(sitk.GetArrayViewFromImage(fixed_image)), moving_npa=fixed(sitk.GetArrayViewFromImage(moving_image)));
 99 
100 
101 initial_transform = sitk.CenteredTransformInitializer(fixed_image,
102                                                       moving_image,
103                                                       sitk.Euler3DTransform(),
104                                                       sitk.CenteredTransformInitializerFilter.GEOMETRY)
105 
106 moving_resampled = sitk.Resample(moving_image, fixed_image, initial_transform, sitk.sitkLinear, 0.0, moving_image.GetPixelID())
107 
108 interact(display_images_with_alpha, image_z=(0,fixed_image.GetSize()[2]), alpha=(0.0,1.0,0.05), fixed = fixed(fixed_image), moving=fixed(moving_resampled));
109 
110 registration_method = sitk.ImageRegistrationMethod()
111 
112 # Similarity metric settings.
113 registration_method.SetMetricAsMattesMutualInformation(numberOfHistogramBins=50)
114 registration_method.SetMetricSamplingStrategy(registration_method.RANDOM)
115 registration_method.SetMetricSamplingPercentage(0.01)
116 
117 registration_method.SetInterpolator(sitk.sitkLinear)
118 
119 # Optimizer settings.
120 registration_method.SetOptimizerAsGradientDescent(learningRate=1.0, numberOfIterations=100, convergenceMinimumValue=1e-6, convergenceWindowSize=10)
121 registration_method.SetOptimizerScalesFromPhysicalShift()
122 
123 # Setup for the multi-resolution framework.
124 registration_method.SetShrinkFactorsPerLevel(shrinkFactors = [4,2,1])
125 registration_method.SetSmoothingSigmasPerLevel(smoothingSigmas=[2,1,0])
126 registration_method.SmoothingSigmasAreSpecifiedInPhysicalUnitsOn()
127 
128 # Don't optimize in-place, we would possibly like to run this cell multiple times.
129 registration_method.SetInitialTransform(initial_transform, inPlace=False)
130 
131 # Connect all of the observers so that we can perform plotting during registration.
132 registration_method.AddCommand(sitk.sitkStartEvent, start_plot)
133 registration_method.AddCommand(sitk.sitkEndEvent, end_plot)
134 registration_method.AddCommand(sitk.sitkMultiResolutionIterationEvent, update_multires_iterations)
135 registration_method.AddCommand(sitk.sitkIterationEvent, lambda: plot_values(registration_method))
136 
137 final_transform = registration_method.Execute(sitk.Cast(fixed_image, sitk.sitkFloat32),
138                                               sitk.Cast(moving_image, sitk.sitkFloat32))
139 
140 print('Final metric value: {0}'.format(registration_method.GetMetricValue()))
141 print('Optimizer\'s stopping condition, {0}'.format(registration_method.GetOptimizerStopConditionDescription()))
142 
143 moving_resampled = sitk.Resample(moving_image, fixed_image, final_transform, sitk.sitkLinear, 0.0, moving_image.GetPixelID())
144 
145 interact(display_images_with_alpha, image_z=(0,fixed_image.GetSize()[2]), alpha=(0.0,1.0,0.05), fixed = fixed(fixed_image), moving=fixed(moving_resampled));
146 
147 sitk.WriteImage(moving_resampled, os.path.join(OUTPUT_DIR, 'RIRE_training_001_mr_T1_resampled.mha'))
148 sitk.WriteTransform(final_transform, os.path.join(OUTPUT_DIR, 'RIRE_training_001_CT_2_mr_T1.tfm'))