ABSTRACT
Deformable image registration (DIR) is one of the major prob-lems in medical image processing, occurring in dose calculation [1], treatment planning [2], and scatter removal of cone-beam CT (CBCT) [3]. It is of prime importance to establish a pixel-to-pixel correspondence between two images in many clinical scenarios. For instance, the registration of a CT image to an MRI of a given patient taken at
CONTENTS 10.1 Introduction 149 10.2 Similarity Measures 151
10.2.1Mutual Information 151 10.2.2Bhattacharyya Distance 152
10.3Viscous Fluid Model 153 10.4 Algorithm 155 10.5GPU Implementation 156
10.5.1Histogram Computation 157 10.5.2 Convolution 158
10.6 Experiments 159 10.7 Conclusion 160 Acknowledgments 161 References 162
dierent times can provide complementary diagnostic information. For such applications, since the required deformation of the patient anatomy cannot be represented by a rigid transform, DIR is almost the sole means to establish the necessary mapping.