Purpose: To develop a stochastic time‐resolved model that accounts for spatiotemporal changes in tumor over the extended course of the treatment. Furthermore, we aim to mitigate the effect of uncertainties, yet generate clinically acceptable plans even in worst‐case scenarios. The performance of the method is demonstrated based on a clinical lung cancer case. Method: Distinct tumor regression scenarios (from slow shrinkage to fast shrinkage) and their associated probabilities are constructed based on data from past studies. We minimize deviations from prescribed dose by taking all regression scenarios into account. These deviations are taken with respect to entire volume of interest, which consists of (right lung, left lung, heart, esophagus, spinal cord, tumor and tissue), over all time‐periods (5 weeks). Results: Conventional models disregarding tumor regression lead to significant overdose on OAR as time progresses, producing V 2 0> > 30%, hence beyond the threshold. The nominal time‐resolved model allows for redistribution of the dose based on tumor regression, leading to a V 2 0 < 19%, whilst improving the tumor coverage. On the other hand, when regression assumptions deviate from the realized, tumor coverage is sub‐optimal and clinically not acceptable. The proposed stochastic time‐resolved model mitigates the effect of lack of specific knowledge about tumor regression. In particular, it meets dose criterion for tumor, whilst sparing the OAR (V2 0 < 23%) even in the worst‐case scenario. Conclusion: Traditional models degrade amidst tumor regression, especially at latter time‐periods. The nominal time‐resolved model allows redistribution of doses based on tumor regression, enhancing OAR sparing, but holds only when assumptions are met. The stochastic model accounts for uncertainties in tumor regression and produces clinically acceptable plans, even in the worst‐case scenario when assumptions are not met.
ASJC Scopus subject areas
- Radiology Nuclear Medicine and imaging