Room: AAPM ePoster Library
Purpose: A novel, automated intensity-modulated radiation therapy (IMRT) treatment planning workflow was developed to incorporate a personalized radiation therapy (RT) beamlet optimization step leveraging biomarkers and dose-based statistical models to estimate patient-specific toxicity and efficacy. To explore the potential benefit of this patient-specific optimization method, a preclinical investigation of the treatment planning workflow was performed to demonstrate feasibility and evaluate clinically-realistic lung RT plans generated with this process.
Methods: A prioritized optimization strategy was employed to maximize the utility metric of a plan subject to clinical dose constraints, followed by additional optimization steps to improve the clinical relevancy of the plan. The dose-dependent utility metric is defined by the probability of efficacy minus the sum of individually weighted toxicity probabilities. To evaluate this method, 5 non-small cell lung cancer (NSCLC) patients, previously treated on an IRB-approved clinical trial and representing a variety of patient geometries and patient-specific utility models, were re-planned using the new optimization method and compared to plans generated using traditional dose constraint optimization. For this NSCLC patient cohort, the patient specific utility metric incorporated models for local-regional progression-free survival, and grade 3+ cardiac events, pneumonitis, and esophagitis.
Results: In each of the five cases, the new IMRT treatment planning process successfully generated lung RT plans that directly incorporate patient-specific variability in radiation sensitivity while maintaining standard clinical constraints. Plans generated with this personalized optimization workflow resulted in an average 1.3% [range: 0.4% - 2.7%] improvement in plan utility when compared to standard dose optimization. Larger improvements were noted in cases with relatively large target volumes.
Conclusion: The successful implementation of this optimization method facilitates direct exploration and inclusion of patient-specific trade-offs within RT planning to maximize the probability of uncomplicated local control based on statistical models of control and toxicity with both dose and biomarkers as covariates.
Funding Support, Disclosures, and Conflict of Interest: This work is funded by NIH P01CA059827 and supported in part by Varian Medical Systems Inc.