SU-D-BRB-03: Full Pre-Clinical Validation of Comprehensive Knowledge-Based Planning for Stereotactic Radiosurgery.

PURPOSE: As knowledge-based planning (KBP) would augment and potentially supplant manual human-driven treatment planning, it is imperative to ensure that the KBP system maintains or improves overall plan quality in the aggregate. The purpose of this study was to: (I) demonstrate non-inferiority of KBP-driven automated stereotactic radiosurgery (SRS) planning with blinded physician review and (II) determine what factors, if any, contributed to the preference of manual plans over KBP in the blinded review.

METHODS: Automated KBP routines in non-coplanar VMAT were developed for three clinical SRS scenarios: isolated brain lesions (ISOLATED), lesions closely abutting (<3cm) non-brain OARs (INVOLVED), and single-isocenter multiple metastases (MULTIMET). Two SRS-specializing physicians were given the patient's history/diagnosis/prescription and then presented with the candidate plans with the identities (manual or KBP) obscured. The preferred plan was then selected by the physician reviewer. If two plans were judged to be clinically indistinguishable, this was scored as a KBP selection as the automated plan had thus achieved the same quality as the manual plan. Multiple parameters were explored as possible correlates to manual preference (i.e. KBP failure): planning date, planner, attending physician, reviewing physicist, PTV volume, nearest OAR-to-PTV distance and prescription dose. Correlation coefficients were calculated using the Kendall-Tau statistical test with significance threshold p<0.05.

RESULTS: For the ISOLATED, INVOLVED and MULTIMET scenarios, the KBP plans were chosen 86% (90/105), 77% (20/26) and 76% (13/17) of the time, respectively. Across the groups, no factors proved to be even moderately correlated or predictive. The only correlation coefficient approaching significance was 0.126 (p=0.052) for PTV volume, implying that KBP plans were selected at a slightly lower rate in the case of smaller PTVs.

CONCLUSION: The clinically-deliverable KBP automated SRS plans were equivalent or superior to manual SRS plans for a clear majority of the cases in all three clinical scenarios. J.A. Hattangadi: Research Grant; Varian Medical Systems; K.L. Moore: Research Grant; Varian Medical Systems.