Glioblastoma (GBM) is a common and aggressive form of brain cancer affecting up to 20,000 new patients in the US every year. Standard of care therapies include stereotactic surgical resection, radiation therapy, and adjuvant chemotherapy. After initial treatment, patient monitoring is guided by standard MR imaging performed at routine intervals. Despite these rigorous therapies, current median survival is only 15 months. Imaging is a central part of brain tumor management, but MRI findings in brain tumor patients can be challenging to interpret and is further confounded by interpretation variability. Accurate interpretation of imaging is particularly important during the post-treatment phase of patient care as it can help clinicians proactively manage disease through early, precise characterization of true tumor recurrence. Disease-specific structured reporting systems attempt to reduce variability in imaging results by implementing well-defined imaging criteria and standardized language. The Brain Tumor Reporting and Data System (BT-RADS), developed at Emory University, is one such framework streamlined for clinical workflows and includes quantitative criteria for more objective evaluation of follow-up imaging. While BT-RADS has had success with clinical adoptability, it still faces hurdles reducing interobserver variability and improving objective classification of disease state. This proposed study addresses an unmet need for unbiased, quantitative metrics for robust, objective interpretation of follow-up imaging for GBM patients. Where previous evaluative methods used two-dimensional, representative MRI slices for evaluating the extent of tumor, we propose to develop a deep learning segmentation tool, accurately calculating volumes of tumor after surgical resection, and trained from our expansive database of patient data with contours manually drawn by radiation oncologists. Further, we propose to develop computationally advanced software for predicting disease progression building upon the clinically developed BT- RADS criteria. Such tools would assist physicians in caring for brain tumor patients through post-treatment surveillance and guide future clinical decision making. In addition, we believe these quantitative metrics would have unbounded potential in clinical trial settings where it may be difficult to evaluate the efficacy of novel therapeutics for GBM. Therefore, these assistive tools will be tested on clinical trial data to determine if they are superior to conventional measurements alone. A fundamental goal of this proposal is ensuring the quantitative tools we develop are applicable for clinicians in their daily lives. Therefore, an effort will be made to collaborate extensively with clinicians to house the algorithms in sleek, intuitive software for physicians to utilize. Through the multidisciplinary environment, high performance computing, and clinical resources we have available, we believe this proposal will be successful in developing clinically assistive tools for unbiased, objective monitoring of GBM patients in clinical settings and evaluation of novel therapies in research settings.
Follow-up imaging for glioblastoma patients is challenging to interpret due to subjectivity in evaluating disease state and overlap between treatment effects and tumor progression. More objective follow-up criteria are vital for sensitive and accurate monitoring of glioblastoma patients after initial treatment to ensure proactive management of disease and accurate characterization of tumor recurrence. This research proposal will develop user-friendly quantitative software for lesion segmentation and disease-state assessment, providing unbiased metrics for post-treatment patient monitoring and guidance of management decisions in clinical and research settings.
