活动地点：

活动类型：学术报告

主讲人：陈韵梅 (University of Florida)

活动时间：

活动内容：

1.Deformable image registration in radiation therapy

 Deformable image registration is a critical technology for imaging guided radiation therapy, since it is able to account for internal organ deformation which is problematic for accurately evaluating the cumulative dose delivered throughout the course of radiation therapy. Existing deformable registration algorithms have problems of being sensitive to noise and the choices of parameters, and requiring a large number of iterations to achieve a reasonable result in accuracy. This limits the online applications of deformable image registration for clinical radiotherapy.

We propose a hybrid algorithm, which optimizes the deformation field and variance of the matching  by minimizing an energy functional.  The energy consists of the smoothness measure of the deformation field and the negative log-likelihood function of the target image and deformed source image. An semi-implicit finite difference scheme with the Additive Operator Splitting method is used to improve the computation efficiency.  A multi-resolution scheme is also applied for faster convergence and reducing the error from interpolation. The numerical results show the improvement of robustness and accuracy of the proposed algorithm . The existence of the model solution  is also proved.