The brain tumor microenvironment is heterogeneous, with variable features of hypoxia, acidosis, nutrient depletion, and stromal and immune cell signaling. These have clear effects on the hallmarks of tumor growth, invasion, and migration, but have not been well characterized in the context of in vivo imaging and preclinical cell culture models.
The specific aims of our research over the next three years will test the overarching hypothesis that microenvironmental stressors along with stromal and immune components can:
- Be identified on in vivo clinical imaging and categorized by machine learning for brain tumor diagnosis, prognosis, and treatment design.
- Alter cell proliferation, invasion, migration, and treatment resistance of cultured cells used in preclinical models.
<a name="expander1">be identified on in vivo clinical imaging and categorized by machine learning for brain tumor diagnosis, prognosis, and treatment design.</a>
We are collaborating with Dr. Kristen Yeom at Stanford University on application of advanced imaging techniques to the study of pediatric brain tumors. Our work has described how patterns of microscopic water diffusion and blood perfusion can be used to distinguish histopathologic features that are otherwise ambiguous on conventional imaging. We have defined prognostically distinct tumor subtypes that correlate with clinical behavior. We also described a phenomenon of low blood perfusion in a specific type of high-grade glioma that is suggestive of a hypoxic state, and has formed the basis of our current in vitro experiments evaluating the effects of hypoxia on epigenetic modifications in patient-derived tumor cultures.
We have created a shared imaging server and data use agreement between Dayton Children’s Hospital and Stanford University, allowing us to share brain tumor data for validation of a novel deep learning model of approximately 250 separate MRI features, based on computer vision tasks that are completely independent of human input.
We have a similar collaboration with Dr. Jason Parker at Indiana University, also in the field of advanced imaging. Our work has also focused on machine learning for integrating and classifying MRI features of brain tumors, based on the rationale that tumor heterogeneity can contribute to treatment failure. This is a separate, second strategy of combining multiple imaging modalities into a single predictive framework, known as “nosologic imaging analysis,” and includes correlation of data collected from intraoperative stereotactic navigation images to biological specimens obtained during tumor resection.
<a name="expander2">alter cell proliferation, invasion, migration, and treatment resistance of cultured cells used in preclinical models.</a>
With financial support from the Gala of Hope, we established a Tissue Bank and culture facility at Dayton Children’s Hospital, now with a current inventory of over 20 cultured central nervous system tumors linked to highly detailed clinical and imaging data elements. We have recently become satellite members of the Children’s Brain Tumor Tissue Consortium. Work is in progress to expand our culture program to include orthotopic xenograft mouse models and lentoviral infection for immortalization of low-grade tumors prone to in vitro senescence.
In our laboratory at the Wright State University Neuroscience Engineering Collaboration Building, we are studying the effects of microenvironmental stress on autopsy-derived diffuse intrinsic pontine glioma (DIPG) cells obtained from Stanford University and VU Medical Center in Amsterdam.