ESTRO 2025 - Abstract Book

S3551

Physics - Optimisation, algorithms and applications for ion beam treatment planning

ESTRO 2025

Keywords: FLASH, 3DRM, Quality Assurance

References: [1] Simeonov Y et al. A Fast 3D Range-Modulator Delivery Approach: Validation of the FLUKA Model on a Varian ProBeam System Including a Robustness Analysis. Cancers (Basel). 2024 Oct 16;16(20):3498. doi: 10.3390/cancers16203498.

3944

Poster Discussion Optimizing on one metric, evaluating on another: Consistent improvements in proton FLASH metrics via spot sorting algorithm Viktor Wase, Rasmus Nilsson, Sophie Wuyckens, Albin Fredriksson R&D, RaySearch Labs, Stockholm, Sweden Purpose/Objective: The FLASH effect holds significant promise for radiotherapy by sparing healthy tissues while maintaining tumor control. Recently, there has been an interest in applying sorting algorithms to increase the volume that satisfies the FLASH effect conditions, without altering the dose distribution [1, 2]. However, these conditions are not well understood, and multiple FLASH metrics exist [3]. We evaluate whether optimizing the spot delivery order based on one metric can improve the FLASH effect across other metrics, thereby enabling trustworthy clinical improvements.

Material/Methods: Three cases were tested:

1. Cubic target phantom surrounded by an OAR. 2. Cubic target phantom intersected by an OAR. 3. 2-beam prostate.

A conformal FLASH plan was created using a patient-specific 3D-printed “hedgehog”, consisting of spikes, treating the whole target with a single energy layer together with a range shifter and an aperture [4]. The dose rate of such plans is inherently high but was optimized further by sorting the spots such that a metric was maximized, in a particular OAR. For the phantom cases it was the OARs mentioned above, and for the prostate case it was an OAR consisting of the combined rectum and bladder. The sorting targeted three dose rate definitions: percentile dose rate at 90% and 95%, and average dose rate. These definitions were combined with two FLASH dose thresholds (4 Gy and 6 Gy) and two dose rate thresholds (40 Gy/s and 60 Gy/s), giving 12 metrics. Each case was re-sorted using each metric and then evaluated across all metrics. The sorting algorithm aimed to increase the dose rate for each voxel according to the relevant metric, even if the voxel doesn’t reach the FLASH conditions, but with a weighted priority on satisfying the conditions. This implies that the sorting can have an effect even if no voxels achieve the FLASH thresholds.