Knowledge‐based automated planning for oropharyngeal cancer

• DOI: 10.1002/mp.12930
• Publication Date: 2018-04-21T14:15:47Z
• Authors: Aaron Babier; Justin J. Boutilier; Andrea L. McNiven; Timothy C.Y. Chan
• Keywords: Benchmarking

The purpose of this study was to automatically generate radiation therapy plans for oropharynx patients by combining knowledge-based planning (KBP) predictions with an inverse optimization (IO) pipeline.We developed two KBP approaches, the bagging query (BQ) method and generalized principal component analysis-based (gPCA) method, predict achievable dose-volume histograms (DVHs). These approaches generalize existing methods predicting physically feasible organ-at-risk (OAR) target DVHs in sites multiple targets. Using leave-one-out cross validation, we applied both models a large dataset 217 patients. predicted were input into IO pipeline that generated treatment (BQ gPCA plans) via intermediate step estimated objective function weights model. compared clinical benchmarking. To assess complete pipeline, BQ plans. isolate effect predictions, put through produce optimized (CIO) This approach also allowed us estimate complexity benchmarked against CIO using DVH differences criteria. Iso-complexity (relative CIO) evaluated.The tended less dose is delivered than what observed while more similar DVHs. Both populations reproduced within median difference 3 Gy. Clinical criteria OARs satisfied most frequently (74.4%), 6.3% points Meanwhile, (90.2%), 21.2% However, once constrained plans, performance degraded significantly. In contrast, still notable improvement being criteria.Our automated can successfully use high-quality without human intervention. Between methods, tend achieve comparable as even when controlling plan complexity, whereas underperform.