[ACCESS RESTRICTED TO THE UNIVERSITY OF MISSOURI AT AUTHOR'S REQUEST.] Knowledge-based planning (KBP) has become a prominent area of research in radiation oncology in the last five years. The development of KBP aims to address the lack of systematic quality control and plan quality variability in radiotherapy treatment planning by providing achievable, patient-specific optimization objectives derived from a model trained with a cohort of previously treated, site-specific plans. This dissertation intended to develop, evaluate, and implement a knowledge-based planning system to reduce variability and improve radiotherapy treatment plan quality. The project aimed to 1) develop and validate an algorithm to train mathematical models that predict dose-volume histograms for organs at risk in radiotherapy planning, 2) implement the algorithm into a software application in order to transfer the technology into clinical practice, and 3) evaluate the impact of the software system (algorithm + application) on reducing variability and improving radiotherapy treatment plan quality through knowledge transfer. The presented work demonstrates that a KBP model is beneficial to radiotherapy planning. The developed models adequately describe what is dosimetrically achievable for patient specific anatomy and have proven useful in outlier detection for quality control of radiotherapy planning. The KBP paradigm has also demonstrated ability to improve treatment plan quality through benchmarking and transfer of knowledge between institutions.
PV-0175: Knowledge-based DVH predictions for automated individualised treatment plan quality assurance
