The diagnosis of primary brain tumors requires a layered approach of histologic, anatomic and molecular features to generate an integrated diagnosis with clinical and prognostic significance. The diagnostic workup of diffuse gliomas in particular requires a panel of immunohistochemical stains with a subset of tumors requiring additional molecular testing to reach a diagnostic category recognized by the World Health Organization. In the United States and worldwide, scarce resources are available to perform these tests, so methods that improve pre-test probabilities and decrease false positive results have significant clinical and financial impact. Our long-term goal is to improve and standardize testing and diagnoses for brain tumor patients worldwide by validating new diagnostic workflows using digital imaging, immunohistochemical tests, open source computing platforms and machine learning algorithms to improve diagnostic capabilities. We will achieve these goals by completing three specific aims in this R03 Pilot/Feasibility project. First, we will determine the extent to which predictive diagnostic models developed from public domain data show generalizability to cases evaluated at a tertiary brain cancer care center. We have already generated a prototype statistical predictive model, which we will expand to all CNS tumor types and test with data from patients at James Cancer Hospital. Second, we will generate and validate models that predict the probability of false-positive 1p/19q FISH testing using histological features from OLIG2-immunostained brain tumor slides obtained from whole slide imaging. Lastly, we will consolidate data containing whole slide digital images, immunohistochemical features, clinical data and molecular features of diffuse gliomas. Consolidating these data will allow us to begin data analysis correlating histological images to immunohistochemical and molecular features. This dataset will represent the core dataset that upon which we will base our next R01-level proposal. To achieve these goals, we have assembled a multidisciplinary team composed of an image analysis expert and neuropathologist (JO), a molecular neuropathologist (DT), and a high dimensionality bioinformaticist (JZ). The Ohio State University is the first US cancer center to transition to complete whole slide imaging, and therefore we are in a unique position to generate a significant, vertical advance in improving diagnostic accuracy in neuropathology with modern Pathology Informatics approaches.
The proposed research is relevant to public health because it will develop predictive models of cancer diagnoses and consolidate various forms of patient data into one centralized source for research. This new knowledge will ultimately serve as the framework upon which interventions are designed that may improve morbidity and mortality of patients with primary brain tumors. Therefore, the proposed research is relevant to the part of NIH's mission that pertains to developing fundamental knowledge that will help reduce the burdens of human disease.
