Dose Linear Energy Transfer Volume Histogram and Dosimetric Seed Spot Analysis in Spot Scanning Proton Therapy - PROJECT SUMMARY/ABSTRACT Radiation therapy may induce severe adverse events (AEs), and result in significant detriment to patients’ quality of life, especially in head and neck, such as brain necrosis (BN), osteoradionecrosis (ORN), etc. It is important to develop safer radiation therapy delivery models to minimize the risks of AEs. Spot scanning proton therapy (SSPT) provides such an opportunity due to its unique physics characteristics, i.e., Bragg peak. But it also introduces additional complexities since proton deposits most of the energy within a short distance, resulting in high linear energy transfer (LET) compared to traditional photon therapy. Therefore, its relative biologically effective (RBE) dose may vary significantly even for the same delivered physical dose. However, clinical use of LET is limited because of the uncertainties and discrepancies in RBE modelling and calculation. Thus, current SSPT plan evaluation exclusively relies on physical dose and ignores critical LET information. In this project, we propose a novel tool of dose-LET volume histogram (DLVH) to visualize dose, LET, and structure volume in one 3D surface plot. While avoiding parameters with large uncertainties in RBE modelling, we will correlate AEs with well-defined quantities of dose/LET to establish dose-LET volume constraints (DLVCs) at the organ level. The follow-up medical images (CT, PET, and MRI etc.) showed that only a small portion of regions (seed spots) at the AE sites originated from the synergistic effects of high dose and LET. These seed spots then expand to nearby voxels with low dose/LET due to biological processes to form AE sites identified later in the follow-up images. The inclusion of these voxels with low dose/LET will introduce lots of “noise” in the data analysis. Thus, it is important to filter these “noise” from the true “signals” if we want to understand the dose-LET effects in AE initialization. We propose a novel voxel-based method, dosimetric seed spot analysis, for studying the dose-LET effects upon AE initialization at the voxel level. Seed spot analysis is akin to the patch-based methods in medical image analysis and identifies seed spots. Empirical AE models will be developed via dosimetric seed spot analysis with critical voxels selected by the population-based DLVC. The inter-patient variation will be considered by the two-level mixed effect logistic regression model. The normal tissue complication probability models will be established based on DLVCs and be applied in clinical practice for plan evaluation, and we will then develop DLVH-based treatment planning using DLVC in SSPT to prospectively minimize the possible incidence of AEs. The proposed project will be carried out by experts from Mayo Clinic, MD Anderson, New York Proton Center/Memorial Sloan Kettering, and the University of Georgia. Our project will fill a critical technology void in SSPT and is one of the largest in-vivo outcomes-driven studies to investigate the impact of LET. The success of this study will become a major advancement to the current standard of care of SSPT plan evaluation and optimization using physical dose alone, leading to precision radiation therapy and improved clinical outcomes.
