Related Experiment Video
Updated: Aug 6, 2026

Proton Therapy Delivery and Its Clinical Application in Select Solid Tumor Malignancies
Published on: February 6, 2019
Development of a GPU-based dose calculation engine for MRI-guided proton therapy
Yue Gu1, Yuxiang Wang1,2, Meiqi Liu1
1Department of Engineering and Applied Physics, University of Science and Technology of China, Hefei, Anhui, China.
Background:
Magnetic resonance imaging (MRI)-guided proton therapy (MRPT) integrates the superior soft-tissue contrast of MRI with the high dose conformity of proton therapy. Clinical implementation of MRPT requires an efficient and accurate dose calculation engine capable of accounting for the heterogeneous magnetic field effects on proton beams.
Purpose:
This study developed a GPU-based proton and ion dose engine (gPRIDE) for fast and accurate dose calculations in MRPT under realistic clinical configurations.
Methods:
To account for magnetic focusing effects, a novel beam profile correction method was developed. Computational efficiency was enhanced through a region-of-interest-based dose calculation, a linear approximation in ray tracing, and hybrid pencil-beam- and voxel-level GPU parallelization. Three-dimensional magnetic field maps with varying magnetic field strengths were used for validation. gPRIDE was validated against Monte Carlo simulations in a water phantom and four patient cases (prostate, liver, lung, and brain) under two beam orientations: beams parallel and perpendicular to the main magnetic field.
Results:
gPRIDE achieved gamma passing rates exceeding 99% under the 2 %/2 mm criterion for all cases except the lung case, where pronounced anatomical heterogeneity posed challenges to analytical modeling. Despite this, gPRIDE still achieved a gamma passing rate 95% under the 3 %/3 mm criterion. Wall-clock time per treatment plan ranged from 2.7 s to 13.7 s on an NVIDIA GeForce RTX 3060 GPU.
Conclusions:
gPRIDE enables fast and accurate dose computations under realistic three-dimensional magnetic field maps and supports both beam orientations relative to the main magnetic field. The combination of computational efficiency, accuracy, and versatility facilitates its integration into clinical MRPT workflows.

