StrokeSight: An Intelligent EEG-Based Approach to Rapid Stroke Diagnosis Using Spectral Biomarkers towards Precision Medicine Approaches

Overview

A stroke is defined as a neurologic deficit arising from an interruption in blood supply to the brain. According to the World Health Organization, over 15 million people suffer from strokes annually, of which almost 70% die or are permanently disabled. Effective treatment must be administered within one hour to prevent irreversible brain damage. Currently, there is no cost-effective, fast, and portable diagnostic tool for strokes. This research proposes StrokeSight, a novel pipeline that provides a comprehensive assessment of ischemic and hemorrhagic strokes in under 50 seconds using electroencephalograms (EEGs). Spectral biomarkers were computed using the averaged power spectral densities for 132, 60-second EEG readings. One-way ANOVA tests validated the effectiveness of such biomarkers for stroke classification. These were first used to train three deep neural networks that respectively predict a stroke’s type (none/ischemic/hemorrhagic), location (left/right hemisphere), and severity (small/large) with accuracies of 97.5%, 94.4%, and 100%. StrokeSight also uses these biomarkers in a novel process to visualize spectral abnormalities caused by strokes. Azimuthal equidistant projection and multivariate spline interpolation are used to project 3D electrodes onto a 2D plane, and a contour map of relative frequency band power is created, allowing neurologists to quickly and accurately interpret EEG data. StrokeSight could drastically improve the speed and accessibility of stroke diagnosis by making EEG-based diagnosis more cost-effective and easier to interpret. This research also sets up the potential for precision medicine approaches to stroke treatment by automating comprehensive analyses of strokes. BOREAS: Innovating Respiratory Care with Telemedicine Samvrit Rao Thomas Jefferson High School for Science and Technology, Alexandria, VA Telemedicine, as a revolutionary innovation, has the potential to remedy healthcare disparities on a global scale. The COVID-19 pandemic has facilitated the adoption of telemedicine as the primary mode of care for medical consultations in the United States, particularly for the diagnosis of respiratory diseases. However, a significant limitation of telemedicine is its inability to transmit breath sounds, a crucial clinical data metric, essential for the effective diagnosis of respiratory diseases. After extensive brainstorming, research and consultation with physicians, I conceptualized and designed BOREAS- an integrated hardware-software solution that can capture and transmit breath sounds, optimized for telemedicine. A prototype for BOREAS was built, consisting of a lapel microphone capture device and a smartphone app that utilizes a Javascript-based framework. It was designed, built, and tested using multiple breath sound libraries. The results of the waveform analysis of the transmitted breath sounds via the BOREAS platform revealed a high level of accuracy. Furthermore, the integration of machine learning, utilizing the 3M Littmann Library data enabled the recognition of breath sound patterns, thus validating the efficacy of the BOREAS platform and facilitating the rapid and accurate diagnosis of respiratory diseases. In summary, this integrated hardware- software solution can capture and transmit breath sounds, a crucial diagnostic parameter that is lacking in existing telemedicine solutions. It holds the potential to significantly improve healthcare delivery and reduce the morbidity/mortality associated with respiratory diseases, helping millions of patients worldwide.

Competition history

• JSHS 2023 Category not listed

Resources

• National JSHS abstract book ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.