# Quantitative analysis of robustness‐based versus LET‐based optimization in intensity‐modulated proton therapy for pediatric brain tumors

## TL;DR

This study compares two optimization methods in proton therapy for pediatric brain tumors, finding a trade-off between treatment robustness and radiation energy transfer effects.

## Contribution

The paper quantitatively evaluates the trade-off between robustness-based and LET-based optimization in proton therapy for pediatric brain tumors.

## Key findings

- Robust optimization reduced mean LET in the target by 8%-15% compared to nominal and 11%-20% compared to LET optimization.
- LET optimization reduced high LET in normal tissues like brainstem and spinal cord by 29%-41% and 13%-36%, respectively.
- Both robust and LET optimization resulted in less variation in LET for organs at risk compared to nominal optimization.

## Abstract

To perform a quantitative analysis of the trade‐off between robustness‐based and linear energy transfer (LET)‐based optimization of intensity‐modulated proton therapy for anatomically challenging pediatric brain tumor cases.

Three pediatric brain tumor patients were included in this study. Three plans were generated for each case based on: (1) nominal optimization, without considering uncertainties and LET; (2) robust optimization, including proton range and patient setup uncertainties; and (3) LET optimization, considering increasing LET in the target and reducing LET in normal tissues. All plans were optimized with individually fine‐tuned objective weighing to obtain highest achievable target coverage and meet dose limits of critical structures considering robustness or LET criteria. We then evaluated the impact of robust optimization on LET distribution and assessed the robustness of the LET optimization plan, intercompared with the nominal plan. We also compared the dose and LET distributions, as well as biological effect, for each individual beam across the three plans.

Robust optimization consistently achieved robust target coverage among all cases, but it reduced the mean LET in the target by 8%–15% compared to nominal optimization and by 11%–20% compared to LET optimization. LET optimization was effective in reducing high LET in normal tissues. In patient cases, it reduced the maximum LET in the brainstem and spinal cord by 29%–41% and 13%–36%, respectively, compared to nominal optimization. Robust optimization also reduced high LET in the brainstem and spinal cord, but to a slightly lesser extent than LET optimization. Moreover, both robust and LET optimization resulted in less variation in the mean and maximum LETd for OARs compared to nominal optimization.

An inherent conflict between robust target coverage and high LET in the target in IMPT planning was demonstrated in our study for pediatric brain tumor patients. Robust optimization resulted in lower LET not only in the target but also in nearby critical structures, such as brainstem and spinal cord. In contrast, LET optimization improved the LET distribution but worsened the plan robustness compared to nominal optimization. The trade‐off effect between LET enhancement and plan robustness can vary in respect to anatomical relationships and needs to be carefully evaluated in IMPT planning.