Spatially-adaptive dose calculation grid and planning method for proton minibeam radiation therapy

Proton minibeam radiation therapy (pMBRT) poses computational challenges in achieving accurate dose calculations for sub-millimeter minibeams, necessitating fine spatial resolution (e.g., 0.2 mm) in the dose calculation grid.

This work proposes a spatially adaptive dose calculation method that considers pMBRT beam characteristics, including high-dose gradients and spatial fractionation. The method dynamically adjusts the dose calculation grid resolution for computational efficiency while preserving accuracy, employing strategies such as adaptive mesh refinement and hierarchical dose calculation.

The proposed spatially-adaptive method addresses the intrinsic characteristics of pMBRT, focusing on sharp dose gradients and spatial fractionation at the beam entrance. Within this adaptive dose calculation framework, the conventional uniform dose grid is replaced by a multiscale adaptive grid, incorporating fine spatial sampling in areas of high peak-to-valley dose ratio (PVDR) to ensure precise dose calculation. The sampling rate gradually decreases as depth increases from the beam entrance, optimizing computational efficiency. A dual-grid inverse optimization method is developed for spatially-adaptive grid-based treatment planning. Validation of the adaptive method involves comparing dose calculation and treatment planning results on an adaptive dose grid ("ADAPTIVE") with those on a fine dose grid ("FINE") with a resolution of 0.2 mm and a coarse dose grid ("COARSE") with a resolution of 3 mm. Both ADAPTIVE and COARSE are interpolated to align with the FINE, facilitating accurate dose comparisons.

The efficacy of ADAPTIVE for pMBRT is demonstrated by comparing dose calculation and treatment planning with FINE and COARSE. The dose calculation from ADAPTIVE was highly consistent with that from FINE (e.g., Gamma index passing rate in reference to FINE was 100% for the prostate case), while COARSE (Gamma index passing rate 73%) deviated substantially from FINE. On the other hand, the number of dose calculation grids for ADAPTIVE was only 1/15 of that for FINE, and the computational time for ADAPTIVE was less than one-tenth of that for FINE. In terms of the treatment planning difference, ADAPTIVE provided the same PVDR and almost the same dose distribution (e.g., Gamma index passing rate 100% for the prostate case) as FINE, while COARSE (Gamma index passing rate 64%) again was drastically different from FINE.

This work introduces a novel spatially adaptive high-resolution dose calculation method for pMBRT. The proposed spatially adaptive method efficiently and accurately calculates voxel-by-voxel dose distribution and PVDR for pMBRT, significantly reducing the number of required dose calculation grids, compared to the spatially-uniform method.