Optimization of pencil beam scanning pattern for FLASH proton therapy

TL;DR

This paper introduces a graph theory-based algorithm to optimize proton beam scanning patterns for FLASH therapy, significantly increasing dose rates while maintaining treatment quality, and offers a more efficient alternative to genetic algorithms.

Contribution

A novel graph-based optimization method for proton scan trajectories that enhances dose rate in FLASH therapy with improved efficiency over existing genetic algorithms.

Findings

Doubling of median dose rate with optimized pattern

Minor increase in DR95 compared to conventional patterns

Method requires fewer evaluations than genetic algorithms

Abstract

Purpose: The FLASH effect, which reduces the radiosensitivity of healthy tissue while maintaining tumor control at high dose rates, has shown potential for improving radiation therapy. Conformal FLASH proton therapy involves advanced beam-shaping technologies and specialized nozzle designs to confine the dose to the target volume. Optimizing the spot delivery pattern and range modulators can enhance the local dose rate, and genetic algorithms have been used to optimize scan patterns for stereotactic FLASH proton therapy of early-stage lung cancer and lung metastases. A fast and effective method based on graph theory is proposed to optimize the dose rate in specific regions of interest. Methods: We have created a graph-based algorithm to optimize the trajectory of proton spots to maximize the 100th percentile dose rate. Since this problem is NP-hard, we have employed an approximation algorithm that can solve this kind of Traveling Salesman Problem efficiently. Results: When compared to a conventional serpentine pattern, the optimized scanning trajectory led to a doubling of the median dose rate, but only a minor increase in DR95. Our approach is more efficient and requires fewer evaluations of the objective function and hyper-parameters compared to existing genetic algorithms. Conclusions: The optimized scanning trajectory led to a doubling of the median dose rate, but only a minor increase in DR95. The extent to which the dose rate can be increased depends on the size and shape of the region of interest. Future research could explore integrating FLASH objectives into treatment planning and incorporating the proposed method into plan optimization.