Development of Lineal Energy Spectrum-Based Biological Effects Models for Protons

In this dissertation, methods are developed and described to allow for the rapid calculation of microdosimetric spectra (specifically, lineal energy) for protons. SuperTrack, a GPU-accelerated tool for calculation of microdosimetric spectra was developed and is capable of computing lineal energy spectra up to 5000x faster than using Geant4 directly. Proton lineal energy spectra generated by SuperTrack are indistinguishable from those generated by Geant4. With SuperTrack, large libraries of lineal energy spectra for monoenergetic protons spanning 0-300 MeV have been developed. The proton lineal energy spectra calculated by SuperTrack have been compared to experimental measurements made by a tissue equivalent proportional counter and demonstrate reasonable agreement. A method to sum monoenergetic lineal energy spectra to yield the lineal energy spectrum of a polyenergetic beam is described and validated. The summation approach for calculation of lineal energy spectra, along with the libraries generated by SuperTrack have been incorporated into a treatment planning system, RayStation IonPG-2023B. Having made the rapid calculation of proton lineal energy spectra possible, investigations to establish and determine the optimal mathematical formulation of a mathematical radiobiological model for the prediction of the biological effects of protons began. Using previously gathered clonogenic cell survival response data of H460, H1437, U87, and AGO cell lines following proton irradiation, mathematical models describing the relative biological effectiveness of proton therapy as a function of lineal energy and linear energy transfer were developed. It was determined that the potential benefits of lineal energy spectrum-based radiobiological models for protons may only be meaningful in conditions where cells are subject to multiple irradiation conditions with differing underlying proton energy spectra at the same linear energy transfer. Following this, mathematical models to predict in-vivo treatment outcomes following proton therapy were developed. Four distinct analysis approaches were applied to a cohort of pediatric ependyoma patients treated with proton therapy, first identified in a prior study by Peeler et al. 2016. The analysis approaches attempted to determine whether a correlation with increasing linear energy transfer and the appearance of hyperintense regions on T2-weighted magnetic resonance imaging post-treatment were correlated. I found that analysis approaches which grouped voxel-level response data from all patients together indicated that higher linear energy transfer was correlated with increasing risk of post-treatment image change. However, analysis methods which considered the risk of each individual patient’s risk of developing image changes found that most patients did not demonstrate increasing image change risk with increasing linear energy transfer. Additional work remains to be done to extend the lineal energy spectrum-based models developed to predict clonogenic cell survival to the prediction of in-vivo treatment response.