Online Epileptic Seizure Detection in Long-term iEEG Recordings Using Mixed-signal Neuromorphic Circuits

Olympia Gallou,Jim Bartels,Saptarshi Ghosh,Johannes Sarnthein,Kaspar Schindler,Giacomo Indiveri

DOI: https://doi.org/10.1101/2024.06.13.24308876

2024-06-13

Abstract:Seizure detection stands as a critical aspect of epilepsy management, which requires continuous monitoring to improve patient care. However, existing monitoring systems face challenges in providing reliable, long–term, portable solutions due to the computational expense and power demands of continuous processing and data transmission. Edge computing offers a viable solution by enabling efficient processing locally, close to the sensors and without having to transmit the sensory signals to remote computing platforms. In this work, we present a mixed–signal hardware implementation of a biologically realistic SNN for always–on monitoring with on–line seizure detection. We validated the hardware system with wideband EEG signal recordings with over 122 continuous hours of data, without pre–filtering. The network was tested with a cohort of 5 patients and a total number of 22 seizures including generalized and focal onsets. Our system effectively captures spatiotemporal features based on synchronized multichannel intracranial EEG activity, achieving 100% sensitivity across all patients and near–zero false alarms. Remarkably, inference across patients required only calibrating the parameters of the network's output layer on a single recorded seizure from the patient.

Neurology

What problem does this paper attempt to address?

This paper mainly discusses how to use mixed-signal neuromorphic circuits to solve the problem of long-term online seizure detection. Existing monitoring systems face challenges in providing reliable, long-term, and portable solutions because of their high computational cost, large power consumption, and the need for continuous processing and data transmission. Edge computing, as a feasible solution, can perform local and efficient processing near the sensors without the need to transmit sensory signals to remote computing platforms.

In this study, the authors proposed a hardware implementation of a biologically realistic Spiking Neural Network (SNN) for real-time online seizure detection. They validated it using wideband electroencephalogram (EEG) signals recorded for more than 122 hours without pre-filtering. The network was tested on 5 patients and a total of 22 epileptic events, including generalized and focal seizures, achieving 100% sensitivity and close to zero false positives. It is noteworthy that cross-patient inference only requires calibrating the network's output layer parameters based on a single seizure recording from each patient.

The paper also introduces an Asynchronous Δ Modulator (ADM) with adaptive threshold encoding to address the impact of different noise levels on the detection performance of biomarkers. Additionally, a mixed-signal neuromorphic processor (DYNAP-SE) was configured for the SNN, which has low power consumption and low sensitivity to artifacts, making it suitable for long-term epilepsy monitoring.

The experimental results demonstrate that the hardware SNN successfully detected all seizure events of the patients, achieving zero false positives for 3 patients. Despite an average detection delay of 18.8 seconds, combined with a clinically low false positive rate, it paved the way for realizing closed-loop neural modulation SoC. Future work will focus on reducing the detection delay to support real-time closed-loop feedback strategies for seizure suppression.