Program Information
Automated Plan Quality Survey Across Several Institutions Using Batch Knowledge-Based Planning
B Ziemer*, N Li , J Hoisak , R Rice , D Hoffman , I Dragojevic , M Cornell , K Moore , University of California, San Diego, La Jolla, CA
Presentations
WE-F-205-2 (Wednesday, August 2, 2017) 1:45 PM - 3:45 PM Room: 205
Purpose: Radiotherapy plan quality comparisons across multiple institutions are difficult for many reasons, but primarily because no shared set of patient cases exists to make equivalent comparisons. The purpose of this work was, using knowledge-based planning (KBP) to automate and standardize planning, to demonstrate a method to compare practices across multiple distinct institutions with minimal human effort.
Methods: One coordinating institution (MAIN) and three associated satellite clinics (SAT₁,₂,₃), each with separate clinical databases and human planners, exported and anonymized 20 prostate patients’ volumetric modulated arc therapy (VMAT) plans under an IRB-approved study. The previously-treated plans were imported into a central database running commercial KBP software (Varian RapidPlanᵀᴹ) and Eclipse Scripting Application Programming Interface (ESAPI) with read/write capabilities. A previously-validated RapidPlan prostate routine (separately trained with 105 patients from MAIN) was applied to all 80 cases via a batch ESAPI auto-planning script that used identical VMAT setup to the test clinical cases but optimized with KBP-driven parameters. Planning target volume (PTV) and organ-at-risk (OAR) dose-volume histogram (DVH) differencing ΔDVH=DVH_clinical-DVH_KBP quantified separation between clinical and KBP planning. Within each clinic’s 20-patient sample, averages and standard deviations of ΔDVH quantified aggregate performance against the KBP standard, and thereby each other. Unpaired t-test (significance threshold p<0.05) identified systematic differences between MAIN and SAT₁,₂,₃ samples on standard prostate radiotherapy DVH cutpoints for PTV and OARs.
Results: Most metrics were not appreciably different across the four clinics, showing equivalent plan quality for most prostate-specific parameters. However, statistically significant differences were observed in each satellite clinic for some key dosimetric variables – PTV ΔV100%(p≤0.01;SAT₂,SAT₃); rectum ΔDmax(p<0.01;all); bladder ΔDmax(p≤0.02;all) – showing some systematic differences between the satellite clinics and the main site.
Conclusion: A multi-patient, multi-institution quality survey was accomplished with no manual effort, save data transfer steps, using automated knowledge-based planning and an application programming interface.
Funding Support, Disclosures, and Conflict of Interest: KLM acknowledges research grants, travel, and honoraria from Varian Medical Systems.
Contact Email: