Medical Image Elastic Registration Based on Thin-Plate Splines
Haiping Ren · Beijing shengwu yixue gongcheng · 2003
Non rigid registration is a key issue for neurosurgery and radiotherapy planning. Two elastic registration methods based on two sets of corresponding feature points for human eyeground and brain images are presented. One method is thin plate spline interpolation, which forces the corresponding points to match each other exactly. Sometimes, this method would cause serious local distortions. Another method is thin plate spline elastic approximation, which takes the overall smoothness of the whole image into account, hence it is better when errors of feature points exist. Experiment results showed that good registration were obtained using the above mentioned two methods.