Towards patient-specific biology-driven heterogeneous radiation planning: using a computational model of tumor growth to identify novel radiation sensitivity signatures.

Purpose/Objectives: With the burgeoning evidence of a cancer stem cell architecture in a number of solid tumors, many of our assumptions about optimal dose and scheduling for radiation therapy are being called into question. Further, it is well established that spatial heterogeneity in oxygen concentration can cause regions of relative radiation resistance. How these two forms of heterogeneity synergize, however, is not known. We use a cellular automaton model of a cancer stem cell-driven tumor to identify characteristic cellular-oxygen distribution signatures of tumours and derive optimal therapeutic strategies, which we then utilize for planning of spatially heterogeneous dose maps to optimize tumor control for individual patients, in silico. Materials/Methods: We have extended a previously developed computational model of a stem cell-driven tumor to include host tissue interactions, oxygen uptake and heterogeneous vascularization. We simulate tumor growth on domains, which represent imaging voxels, under a variety of assumptions regarding vascular density/patterning and cell proliferation/differentiation. We correlate the distribution of cells (by type: stem vs. nonstem) versus the oxygen concentration that each cell experiences. Results: By simulating tumors with a range of biological parameters, we have identified a family of cellular-oxygen distribution signatures (cell number vs. oxygen concentration), each of which responds to radiation differently based on the spatial distribution of vessels and stem fraction. We then optimize dose and fractionation for each of these distributions under different assumptions governing repopulation and reoxygenation, allowing for optimization to be done at the voxel level. Conclusions: The ability to generate spatially resolved radiation sensitivity signatures for individual patients could usher in a new paradigm of imaging driven personalized radiotherapy. To achieve this, we have developed a novel method by which to optimize radiation dose and fractionation for solid tumors given a theoretical measure of spatially resolved cellular-oxygen distribution and stem distribution. While our result is preliminary, the cell scale resolution of the model offers the possibility of translation of this method using information gleaned from MRI and PET (CD-133 and F-MISO) imaging.