Ning-Ning Lu, MD, PhD; Ran Wei, PhD; Yuan Tian, PhD; Shi-Rui Qin, MS | National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College
Competition Sponsor: Chinese Academy of Medical Sciences
Award year: 2021
The online adaptive workflow provided by MR-Linac (MRL) can solve all the problems when treating prostate cancer patients with conventional linacs, including bad soft tissue contrast, invasive procedure due to fiducial markers or electromagnetic transponders insertion, target area and organs motion and deformation, etc.. However, the time-consuming MRL workflow is the main obstacle to widely use this technique. The online adaptive deformable registration aligned on MRL is not good enough for prostate cancer, as the bladder filling changes and adjacent organs motion would affect the radiotherapy target, including prostate and seminal vesicles. Therefore, we collect the MR images at pre-scan and during each fractionation, extract the prostate and adjacent organs motion data, and build the prostate and pelvic organs motion prediction model (PPMPM) by Finite Element Analysis (FEA) and Biomechanical Model Constrains (BMC). The accuracy and performance of this model will be testified by comparison of the similarity to the position verification 3D-MRI(PV-MRI), cine-MRI and post-MRI. Based on the PPMPM, a new set of auto-contouring software will be developed to help the adaptive contouring which is rate-limiting step during the Adapt-To-Shape (commonly used for prostate cancer) workflow of MRL. With expanded enrollment, population-based, convolutional neural network (CNN)-based algorithm will be used to improve the model accuracy and robustness. Finally, the total on-couch time would be significantly reduced, patients’ satisfaction increased and more patients could be treated each day. The optimized workflow will reduce the doctors’ workload and promote the clinical use of this high precision technique.