Using Class-Solution Optimization and Knowledge-Based Planning, a Physician-Driven Treatment Planning Workflow Demonstrates Improved OAR-Sparing for EBRT of the Prostate

Abstract

<h3>Purpose/Objective(s)</h3> To simplify the planning process for a novel cone-beam CT-based online adaptive radiotherapy (ART) platform, its integrated treatment planning system (OA-TPS) limits user inputs to target prescription, target coverage goals, and organs-at-risk (OAR) sparing goals. As an option, the system allows for knowledge-based plan (KBP) models to aid in dosimetric optimization. By utilizing KBP models within the framework of the system's ART-focused TPS, we propose a streamlined alternative to the traditional planning paradigm of iterative fine-tuning of patient-specific optimization objectives and time-consuming discussions of multiple clinical shareholders. To this end, we have developed a physician-driven treatment planning workflow using prostate cancer OA-TPS planning templates. <h3>Materials/Methods</h3> Six external-beam prostate cancer patients' treatment plans were generated in both a traditional TPS to serve as the benchmark and the OA-TPS using the approach mentioned above. All patients were planned to be treated in two phases: 4500cGy in 25 fractions (prostate+seminal vesicles+nodes) and 3600 cGy in 20 fractions (prostate) with identical target coverage and OAR-sparing dosimetric goals. In-house KBP models were trained using 39 (phase 1) and 36 (phase 2) plans from our institutional database. OA-TPS plans were generated using (1) prescription goals (PG) and (2) PG + KBP model. Plans generated by the OA-TPS were not modified in any way. <h3>Results</h3> The OA-TPS produced clinically-acceptable plans, per physician review, for 6/6 patients (PG) and 5/6 patients (PG+KBP model) with one plan not achieving target coverage by <1.5%. Dosimetric goals were considered achieved if they either met rotely or were better or within 2% of the dosimetry achieved by the benchmark TPS. Plans generated via the PG+KBP model demonstrated improved OAR sparing for most prescription dose metrics for all patients. PG+KBP model plans also demonstrated considerable OAR sparing at dose levels not specified in prescription goals (<i>e.g.,</i> Rectum V4050cGy in the tabulated data below.) <h3>Conclusion</h3> The OA-TPS generated prostate treatment plans largely meet clinical objectives both with and without using a KBP model. However, by including KBP models, there is a potential to develop streamlined "universal" class-solution approaches to site treatment planning that increase OAR sparing and maintain target coverage without iterative adjustments of patient-specific optimization parameters. Physician-driven treatment planning workflows, facilitated by universal patient plan templates, show promise to generate clinically-acceptable plans for a variety of patient anatomies.