Fast optimization methods for image registration in adaptive radiation therapy
Yuchuan Qiao · Data Archiving and Networked Services (DANS) · 2017
In this thesis we developed several stochastic optimization methods for fast image registration, leading to a 5-10 fold speedup over previous approaches. All proposed methods are implemented using C++ and integrated in the open source registration package elastix. We also exploited the usage of high performance computation resources -- the life science grid (lsgrid) to perform over $10^6$ registrations, which significantly reduced computation time for large scale computational tasks. As we have evaluated the proposed method in the application of online adaptive IMPT for prostate cancer, we expect that these methods can achieve the desirable performance for use in clinical practice.