TH‐AB‐304‐09: A Patient‐Specific Heterogeneous Radiobiological Model for Glioblastoma Multiforme

Abstract

Purpose:Gliomas tumors proliferate and invade healthy brain tissue rapidly, yielding short life expectancies. Radiation therapy is commonly used, but gliomas are known to be radio‐resistant, and almost always recur following treatment. Current radiotherapy protocols are based on the classic linear‐quadratic radiobiological model, assuming a homogeneous (one‐component) tumor. Gliomas are very heterogeneous, consisting of normoxic, hypoxic, and necrotic tissues, each responding differently to radiation. An enhanced linear‐quadratic model which takes into account the different responses of the heterogeneous tumor regions to radiation can guide treatment planning to optimize dose distributions to maximize the therapeutic effect.Methods:We used a set of differential equations to model the growth of glioma tumors. Our model expands on the one‐component model developed by (Rockne,2010) by including normoxic, hypoxic, and necrotic components and radiosensitivity values for each component. Proliferation and diffusion parameters are extracted by contouring the tumor on two sets of pre‐treatment MRIs and modeling it as a volume equivalent sphere.Results:We compared two examples of glioma tumors presented in (Rockne,2010) with our three component model and find better predictive capabilities for post‐RT tumor volumes using the three component model. In both cases, the one‐component model over predicts the effects of radiation on the tumor core, while the three‐component model more accurately fits the data by accounting for radioresistance of the hypoxic core. Additionally, we applied the model to a UCSF patient with a recurring tumor and use the model to predict time to fatal tumor burden.Conclusion:The three‐component model accurately matches tumor growth dynamics derived from MRIs for two example cases presented in literature, and one case from UCSF. Spatial information about hypoxic, necrotic, and normoxic cell densities are derived from the model providing information needed to more intelligently prescribe dose distributions tailored to a specific patient's tumor.[Rockne,et.al.,PhysMedBio,2010]