Advancing Neurosurgical Neuronavigation Using Resting State MRI and Machine Learning

点击翻译按钮获取中文摘要

Abstract. Long-term survival of patients with glioblastomas (GBM) are associated with two competing priorities: 1) gross total resection and 2) preservation of the patient’s function. Stereotactic navigation, in which reconstructed magnetic resonance images (MRI) of the brain are used for real-time intraoperative anatomic guidance, has become an essential tool for tumor resection. Further, there are emerging insights that glioma- specific perturbations of the functional organization of the brain impact the patient’s survival. However, the current barrier is that there is no FDA approved navigation system that enables the surgeon to visualize the functional architecture of the brain and the impact a tumor has on the brain’s network organization to inform prognosis. Resting state functional MRI (rs-fMRI) has emerged as a powerful tool for mapping clinically relevant brain networks and defining critical glioma-neuronal interactions. rs-fMRI is highly efficient, task independent, and multiple resting state networks (RSNs) can be mapped simultaneously. With this in mind, the long-term goal of our research is to improve treatment, survival, and quality of life for patients with brain tumors by improving the identification of eloquent cortex and providing actionable metrics for survival prognosis to best tailor a patient’s care. In our first Academic Industry Partnership between Washington University and Medtronic we were extremely productive in creating an integrated brain-mapping navigation technology using rs-fMRI. Specifically, we created a robust image acquisition/analysis pipeline that includes pre-processing of raw data, quality control analytics, and clinical validation demonstrating superior performance over task-based fMRI. We have also been leaders in deriving prognostic radiomic biomarkers from rs-fMRI. In this continuation, we will build on these successes. The overall objective is to create advanced rs-fMRI machine learning (ML) tools to more efficiently and accurately define functional cortex and provide preoperative prognostic metrics of survival as a comprehensive surgical/care navigation system. We have the expertise, infrastructure, and data, to advance rs- fMRI to be a powerful tool for neurosurgical decision support. The proposal entails three specific aims: 1) Advance an ML algorithm to enable more accurate and data efficient rs-fMRI brain-mapping software, 2) Create an rs-fMRI ML algorithm to preoperatively predict survival in glioblastoma (GBM) patients, and 3) Validate impact of mapping and prognostic algorithms on clinical decision making in prospective feasibility clinical trial. The expected outcome of this work will be an integrated imaging/surgical navigation technology using rs-fMRI for clinical decision support with defined performance, clinical validation, and a regulatory path for FDA clearance. Thus, this proposal is innovative because 1) the software will map networks with substantially shorter image acquisition times, thus enabling more widespread adoption and 2) provide critical pre-operative survival insights to inform surgical decisions. This work is significant because it will disseminate technology that fundamentally enhances more tailored approaches to improving patient outcomes and quality of life.