A fast DRR-based image regestrition algorithm for calculation of setup error in radiotherapy

In radiotherapy, patient setup error is one of the important factors that impact the final therapeutic effect. To help operators correct the patient setup accurately, this paper proposes a method to real-time calculate the offset of patient´s positions automatically. In this method, the mean of Graphics Processing Unit (GPU) is employed to help fast generate the digitally reconstructed radiographs (DRR), and iterative nearest neighbor based optimization algorithm is used to effectively find the setup error. The experimental results demonstrate that our method can realize the image registration fast with high precision. For the given 512 × 512 images, the average registration time is 20.025s with the average error of translation and rotation lower than 0.5mm and 0.5°.