Comparison of deformable image registration algorithms for planning CT and CBCT in prostate cancer patients treated with external beam radiation therapy.

Abstract

This study evaluated the performance of three categories of deformable image registration (DIR), namely intensity-based, spline-based, and hybrid methods, for planning CT (pCT)-to-cone-beam CT (CBCT) registration to evaluate the accuracy of different DIR approaches for pCT-to-CBCT registration and assess their potential applicability to CBCT-based adaptive radiotherapy. Fourteen default registration methods were compared with the Anatomically Constrained Deformation Algorithm (ANACONDA) using a CIRS phantom and five prostate cancer patients. Registration accuracy was assessed using Target Registration Error (TRE), Dice Similarity Coefficient (DSC), and image quality metrics. Deformed CT (dCT) consistently outperformed CBCT, demonstrating substantial improvements in image quality across Phantom and patient studies. ANACONDA achieved the highest registration accuracy, with TRE 0.90 for target volumes and organs-at-risk. ANACONDA outperformed intensity-based and spline-based registration methods, providing improved image quality and anatomical registration accuracy, thereby facilitating more accurate image guidance for adaptive radiation in pelvic radiotherapy.