Yi Rong

Clinical feasibility of MR-assisted CT-based cervical brachytherapy using MR-to-CT deformable image registration

The purpose of this study is to evaluate the feasibility of using deformable image registration algorithms to improve high-dose-rate high-risk clinical target volume (HR-CTV) delineation between preapplicator implantation MRI (pre-MRI) and postapplicator implantation CT (post-CT) in the treatment of locally advanced cervical cancer (LACC). Twenty-six patients were identified for the study. Regions of interest were segmented on MRI and CT. A HR-CTV was delineated on pre-MRI and compared with the previously contoured HR-CTV on the post-CT. Two commercially available algorithms, ANACONDA (anatomically constrained) and MORFEUS (biomechanical model based) with various controlling structure settings, including the cervix, uterus, etc., were used to deform pre-MRI to post-CT. MRI-to-CT deformed targets are denoted as HR-CTV'. Quantitative deformation metrics include Dice index, distance to agreement, and center of mass displacement. Qualitative clinical usefulness of deformations was scored based on HR-CTV identification on CT images. For ANACONDA and MORFEUS deformations, using a cervix controlling region of interest resulted in the highest Dice, lowest distance to agreement, and lowest center of mass displacement for HR-CTV'. With MORFEUS deformations, the deformed HR-CTV' proved clinically useful in 23 patients. Prebrachytherapy implantation MRI can aid target contours for CT-based brachytherapy through ANACONDA or MORFEUS algorithms with appropriate parameter selection for LACC patients.