Exhibit Hall | Forum 3
Purpose: A recent study indicates that dose-averaged linear energy transfer (LET) may only be suitable for characterizing the relative biological effectiveness (RBE) of protons along a pristine Bragg curve – a condition which never occurs in clinical practice. We believe that more versatile models of RBE require knowledge of the full lineal energy spectrum. Currently, voxel-level determination of lineal energy spectra requires repeated Monte Carlo simulations which are extremely computationally expensive. We are developing software which rapidly generates microdosimetric spectra indistinguishable from those developed using Monte Carlo methods. Our aim is to make possible the clinical and research use of voxel-level microdosimetric information.
Methods: Proton tracks of energies from 0.1 to 100 MeV in water were generated using the Geant4-DNA Option 2 physics model and stored in binary-encoded files. Using a GPU-accelerated algorithm, proton tracks were superimposed on spherical targets within a cubic voxel. The energy imparted to each spherical target was used to determine the lineal energy spectrum. Tracks were oversampled (re-used) to reduce statistical uncertainties.
Results: On typical workstation hardware, the computational efficiency of our software can reach 3500x greater than a Geant4-DNA based application to determine lineal energy of 50.0 MeV protons. Dose-mean lineal energy of protons in 1 μm diameter spheres calculated using our approach are in excellent agreement with published results. A dependence of lineal energy spectrum on voxel size was observed which has not previously been reported. Lineal energy spectra in spheres of diameters spanning 10 nm – 10 μm were calculated.
Conclusion: We have developed SuperTrack, a GPU-accelerated application for computational microdosimetry. A database of proton lineal energy spectra in targets from the nanometer through to the micrometer scale has been generated. Unlike existing Monte Carlo engines our software allows for determination of microdosimetric spectra in voxels filled with greater than 10²⁷ targets.
Funding Support, Disclosures, and Conflict of Interest: This work was supported by the UTHealth Innovation for Cancer Prevention Research Training Program Pre-doctoral Fellowship (Cancer Prevention and Research Institute of Texas grant #RP210042). The authors acknowledge the support of the High Performance Computing for Research facility at the University of Texas MD Anderson Cancer Center.