A Multi-Modal Non-Invasive Deep Learning Framework for Progressive Prediction of Seizures

TL;DR

This paper presents a non-invasive, multi-modal deep learning framework that predicts seizures with high accuracy and granularity, providing critical lead time for intervention while ensuring privacy and real-time processing on edge devices.

Contribution

It introduces a novel multi-modal deep learning approach for progressive seizure prediction using non-invasive sensors, optimized for real-time edge deployment and personalized accuracy.

Findings

Achieved 95% sensitivity in seizure prediction.

Achieved 98% specificity and 97% accuracy.

Predicted seizure countdown up to one hour in advance.

Abstract

This paper introduces an innovative framework designed for progressive (granular in time to onset) prediction of seizures through the utilization of a Deep Learning (DL) methodology based on non-invasive multi-modal sensor networks. Epilepsy, a debilitating neurological condition, affects an estimated 65 million individuals globally, with a substantial proportion facing drug-resistant epilepsy despite pharmacological interventions. To address this challenge, we advocate for predictive systems that provide timely alerts to individuals at risk, enabling them to take precautionary actions. Our framework employs advanced DL techniques and uses personalized data from a network of non-invasive electroencephalogram (EEG) and electrocardiogram (ECG) sensors, thereby enhancing prediction accuracy. The algorithms are optimized for real-time processing on edge devices, mitigating privacy concerns and minimizing data transmission overhead inherent in cloud-based solutions, ultimately preserving battery energy. Additionally, our system predicts the countdown time to seizures (with 15-minute intervals up to an hour prior to the onset), offering critical lead time for preventive actions. Our multi-modal model achieves 95% sensitivity, 98% specificity, and 97% accuracy, averaged among 29 patients.