Towards a Virtual Biopsy: An improved multimodal imaging biomarker to guide treatment decisions in neuro-oncology by combining advanced tissue microstructure imaging with deep learning

Project Summary/Abstract Despite advances in surgery, radiotherapy, and chemotherapy, the prognosis for neuro- oncology patients remains poor, with a mean survival of 12-15 months for high-grade gliomas. One major reason for poor survival is that it remains difficult to accurately assess tumor progression and treatment-related changes on standard imaging. This limits necessary information for guiding biopsies or resecting malignant tissue. In addition, new therapies can cause radiological patterns that obfuscate the underlying course of the disease. As a result, there is a critical need for new quantitative imaging tools to evaluate brain tumors and their progression. The goal of this proposal is to develop novel quantitative imaging biomarkers using the combination of advanced tissue microstructure imaging and deep learning to accurately discriminate tumor from non-tumor tissue, measure tumor progression and treatment response, and predict clinical outcomes. This approach utilizes restricted spectrum imaging (RSI), an advanced diffusion-weighted imaging technique that models the restricted diffusion of water to improve tumor conspicuity. Phase I of this proposal will develop a hybrid multimodal, RSI-based biomarker for brain cancer, quantify the biomarker in abnormal sub-regions, and demonstrate the biomarker’s performance in predicting clinical outcomes. Phase II of this proposal will develop and deploy commercial-grade software to the CorTechs Labs cloud platform, demonstrate its clinical usability and utility, and generate the materials required for a 510K FDA submission. The AI technology developed through this proposal will ultimately serve as a clinical decision support tool to improve clinician performance in diagnosing and evaluating brain tumors and predicting tumor response to treatment.Project Narrative Quantitative biomarkers that can better delineate tumor from non-tumor tissue including radiation injury are essential for the diagnosis, prognosis, and monitoring of treatment response in neuro-oncology patients. The goal of this proposal is to develop an automated AI-based commercial software tool that is predictive of clinical outcomes including treatment response, progression and survival. This approach utilizes a novel combination of restricted spectrum imaging (RSI), an advanced diffusion-weighted MRI technique for probing tissue microstructure, and deep learning to integrate RSI microstructure data with multimodal imaging data.