Knowledge-based radiation therapy treatment planning decision support system for head and neck cancer utilizing...

Radiotherapy treats cancers through precise delivery of radiation to target volumes. Radiotherapy treatment plans, prescribing the delivery of therapeutic radiation, are presently created primarily from clinical experience and application of clinical protocols through trial-and-error rather than standardized quantitative methods. We developed an informatics infrastructure and decision support system to assist during treatment plan creation by providing access to applicable retrospective radiotherapy cases.

Radiotherapy treatment planning is based in part on tumor position and spatial relationships to surrounding structural anatomy. Our system data mines retrospective cases to identify cases and treatment plans with similar tumor position and surrounding anatomy (i.e., multi-organ-tumor constellation geometry) for clinicians to use as references during treatment plan creation. The system is based on a database of 390 DICOM RT dosimetry digital radiotherapy datasets with associated extracted quantitative features. Using data mining techniques, overall similarity between cases is calculated with features extracted from tumor volumes and organs at risk (OAR).

We implemented our radiotherapy treatment planning decision support system with a full-stack infrastructure, including a database of best practice retrospective cases, an algorithmic backend for feature extraction and similarity calculation, and a frontend web application for clinical use. Clinicians can upload current planning cases to the web application whereupon their similarity to knowledge base cases is calculated, and the most similar are presented to clinicians for selection as references during current treatment plan creation.

This radiotherapy treatment planning decision support system, by providing access to geometrically similar retrospective best practice reference cases, presents a novel tool to improve treatment planning. Development of a full system infrastructure increases standardization, facilitates creation of high quality plans, and assists clinicians, particularly during the critical initial beam configuration stage.