Quantitative Evaluation of Knowledge-Based Treatment Planning Models Integrated with Multi-Criteria Optimization
S Jayarathna*, K Reddy, H Li, K Guida, The University of Kansas Medical Center, Kansas City, KS
Presentations
MO-E115-IePD-F3-1 (Monday, 7/11/2022) 1:15 PM - 1:45 PM [Eastern Time (GMT-4)]
Exhibit Hall | Forum 3
Purpose: To investigate the clinical performance of knowledge-based planning (KBP) models derived with Multi-Criteria Optimization (MCO) for patients enrolled on a stage III A/B non-small cell lung cancer (NSCLC) clinical trial.
Methods: Two KBP models were generated for an in-house lung trial. The hypofractionated arm is designed to deliver a simultaneously-integrated boost of 62.5Gy and 45Gy. Prior to importing plans into the KBP models, an ESAPI script was utilized to ensure OAR constraints, target coverage, and plan quality parameters met trial standards. The first KBP model was generated by manual optimization (KBP_M) while the second model was generated by further optimizing the original plans with MCO (KBP_MCO). Outlier plans were eliminated iteratively by replanning or deleting. Model validation was performed using ten clinical cases. For each case, five plans were created: manual, KBP_M, KBP_M with secondary MCO optimization (KBP_M+MCO), KBP_MCO, and KBP_MCO with MCO optimization (KBP_MCO+MCO). During MCO optimization, trade-offs were made to ensure adequate tumor volume coverage while improving OAR sparing.
Results: All plans were able to achieve a V100% of 95% or higher for both PTVs for each patient in the study. For lungs, the difference of V20Gy was improved by about 8% for MCO-derived KBP models in comparison with manual planning. The D0.03cc for the heart and esophagus, which often abutted the PTV volumes, were improved by 2% and 7.5% on average, respectively, with KBP_MCO; utilizing the KBP_MCO model with additional MCO optimization further improved the heart D0.03cc by over 3.5%. MCO optimization improved upon spinal cord D0.03cc by 18% and 8% for KBP_M and KBP_MCO plans, respectively.
Conclusion: MCO-derived KBP models significantly improved the dose sparing of OARs while meeting the appropriate target coverage. KBP_MCO can achieve even greater OAR dose reduction with further MCO optimization, proving useful when for patients enrolled on clinical trials.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email