Machine Learning Analysis of Spectro-Graphic Data for Biomarker Identification

Abstract

Epilepsy is a chronic neurological disorder affecting over 50 million people worldwide. Current methods for monitoring and diagnosing epilepsy are conducted through EEG data collection and manual analysis in a clinical setting. This method can be both stressful and costly for patients as well as time-consuming and inefficient for clinicians. Our team aims to produce a machine-learning EEG analysis program coupled with a wearable continuous EEG monitoring system that allows patients and clinicians to collect critical data points while patients perform their day-to-day activities. This product will reduce patient costs by reducing office visits and increase efficiency and accuracy for clinicians in data analysis and interpretation.

The wearable continuous monitoring EEG device is a battery-powered cap with approximately 15-20 medical-grade 0.5mm diameter electrodes evenly embedded throughout a soft-silicone material on the scalp side and a knit material around the top outside of the cap for a more discrete design. EEG data is filtered and amplified through hardware circuitry and wirelessly transmitted using a Bluetooth-enabled microcontroller unit to a post-processing analysis and interpretation software program. The software program feeds spectrographic data in both the time and frequency domains into a machine-learning algorithm to establish an individualized baseline signal and extract key features indicating an epileptic event. Clinicians can then access this data to aid in monitoring and diagnosis.

The global market for epilepsy monitoring devices market size was estimated to be USD 559.0 million and is expected to grow at a CAGR of 6.6% from 2024-2030. Additionally, the AI in healthcare market size was estimated at USD 19.27 billion in 2023 and is expected to grow at a CAGR of 38.5% from 2024-2030, indicating a very profitable market for our device.

Video