基于薄板样条的医学图像弹性配准

Medical Image Elastic Registration Based on Thin-Plate Splines

非刚体配准是神经外科和放疗计划设计中的一个关键问题。使用薄板样条方法 ,利用两个对应标志点集对眼底、脑等多种医学图像进行弹性配准。其中 ,弹性插值法是将两个点集绝对对齐 ,常会出现严重的局部畸变 ;而弹性近似法充分考虑了整体平滑性的要求 ,对定位有误差的标志点约束的图像配准更为优越。实验结果表明 。

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.