Author: Ali Ajdari, Thomas R. Bortfeld, Zhongxing Liao, Mara Schubert, Katrin Teichert 👨🔬
Affiliation: The University of Texas MD Anderson Cancer Center, Department Of Radiation Oncology, Massachusetts General Hospital (MGH), Massachusetts General Hospital & Harvard Medical School, Fraunhofer ITWM 🌍
Purpose: Radiotherapy (RT) treatment planning often involves solving a multi-criteria optimization (MCO) problem. Conventionally, MCO considers a set of generic (population-wide) dosimetric criteria, ignoring patient-specific biological risk factors, which compromises clinical outcomes in high-risk groups. We propose a one-shot method to directly integrate biological risk factors within conventional MCO, enabling interactive plan navigation between dosimetric and biological endpoints.
Methods: A cohort (n=179) of non-small cell lung cancer patients receiving proton/photon RT was analyzed retrospectively. The clinical endpoint was the risk of symptomatic (grade 2+) radiation pneumonitis (RP), modeled using bootstrapped stepwise logistic regression, with interactions accounting for baseline lung function, smoking history, and dosimetric factors. For our risk-guided MCO (rg-MCO), we defined a special order relation to fuse the conventional MCO sandwiching algorithm with bi-level optimization, narrowing down the set of Pareto optimal plans to those with high gain in the secondary (biological) objective for any loss in primary (clinical) objectives. This allows for efficiently calculating the risk-guided counterparts of clinical plans in one shot (contrasted to reoptimization) with user-defined trade-offs. The algorithm’s performance was assessed in terms of clinical objectives and the RP predicted risk.
Results: The risk factors in the RP model were baseline breathing function, smoking history, V20Gy for total lung, and V5Gy for right lung, with interactions between smoking and V20Gy, and breathing and V5Gy. Comparing a systematically selected rg-MCO plan to a conventionally optimized plan per patient showed a 10.02±0.14% reduction in total lung V20Gy and a 22.26±5.18% reduction in right lung V5Gy across four patients. This translated into patient-specific RP risk reduction up to 5.3% and an average reduction of 2.2±0.05%, while maintaining target coverage and restricting trade-offs to spinal cord and heart dose.
Conclusion: Integrating personalized risk models in multi-criteria plan optimization can reduce the risk of radiation pneumonitis while preserving target coverage.