SBIR Phase I: Improving Diagnosis of Epilepsy by Applying Network Analytics to Non-Seizure Scalp EEG Data

点击翻译按钮获取中文摘要

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is the development of a novel electroencephalogram (EEG) analytics tool that will improve the speed and accuracy of diagnosing epilepsy. The tool is an easy-to-use software package that utilizes scalp EEG data. It is being developed as a cloud-based application designed to integrate with existing software packages and to provide easy-to-read heatmaps available within minutes. Epilepsy centers and other settings where EEG diagnostics are used will benefit from improved accuracy in diagnosing epilepsy: Currently the accuracy is estimated at less than 60%, whereas the proposed tool can improve this figure by over 25%, more accurately distinguishing between epileptic and non-epileptic pathologies from EEG alone. Furthermore, the technology will increase the speed of epilepsy diagnosis: Currently, patients often require multiple EEGs, during which they are at high risk of further seizures. The proposed tool will provide a definitive diagnostic on the first visit. This Small Business Innovation Research (SBIR) Phase I project involves performing a retrospective study to validate a novel EEG analytics tool on 60 or more patients, developing an algorithm to automate artifact removal from scalp EEG data most appropriate for this clinical application, and developing the tool as a cloud-based service. These milestones will facilitate clinical adoption and easy integration into the clinical workflow, both of which are necessary for successful commercialization of the innovation. The tool will predict if a brain network is epileptic while a patient is monitored at rest when no seizure occurs. The key strengths are the use of a dynamic network model (DNM) to uncover connections in the brain that only exist in an epilepsy patient during rest. All other FDA proved tools are based on individual EEG channel properties rather than network-based properties. As a result, their utility is limited to identifying abnormal events (e.g., when an EEG spike occurs), potentially vulnerable to artifacts. In addition, the proposed tool is transformative because it captures how nodes in a network dynamically influence each other, while clinical approaches rely on reading EEG with naked eyes.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.