A neuromorphic spiking neural network detects epileptic high frequency oscillations in the scalp EEG

Abstract

Interictal High Frequency Oscillations (HFO) are measurable in scalp EEG. This development has aroused interest in investigating their potential as biomarkers of epileptogenesis, seizure propensity, disease severity, and treatment response. The demand for therapy monitoring in epilepsy has kindled interest in compact wearable electronic devices for long-term EEG recording. Spiking neural networks (SNN) have emerged as optimal architectures for embedding in compact low-power signal processing hardware. We analyzed 20 scalp EEG recordings from 11 pediatric focal lesional epilepsy patients. We designed a custom SNN to detect events of interest (EoI) in the 80–250 Hz ripple band and reject artifacts in the 500–900 Hz band. We identified the optimal SNN parameters to detect EoI and reject artifacts automatically. The occurrence of HFO thus detected was associated with active epilepsy with 80% accuracy. The HFO rate mirrored the decrease in seizure frequency in 8 patients (p = 0.0047). Overall, the HFO rate correlated with seizure frequency (rho = 0.90 CI [0.75 0.96], p < 0.0001, Spearman’s correlation). The fully automated SNN detected clinically relevant HFO in the scalp EEG. This study is a further step towards non-invasive epilepsy monitoring with a low-power wearable device.

The present and future of neural interfaces

“…Serving as key hardware for artificial neural networks, platforms like SpiNNaker, TrueNorth, and Loihi support closed-loop computing with low power consumption and miniaturized devices. For example, spiking neural networks have been used to continuously monitor brain activity and detect epileptic high-frequency oscillations with a low-power wearable device, paving the way to cheaper and less invasive epileptic monitoring (Burelo et al, 2022 ). Nevertheless, despite being low power, these neuromorphic platforms can do enough computations to accurately decode brain patterns typically associated with BCI paradigms, such as motor imagery (Behrenbeck et al, 2019 ).…”

“…In this scenario, organic neuromorphic platforms recently arose as tools to directly interface biological and artificial neurons to form functional biohybrid synaptic connections (Keene et al, 2020 ). Based on biocompatible organic semiconductors, these devices are capable of mixed ionic-electronic (trans)conduction that very accurately resembles the complexity of the neuronal electrochemical environment in which the neuronal bidirectional communication takes place (Burelo et al, 2022 ). Furthermore, their long-term potentiation and short-term depression (Tuchman et al, 2020 ) have been exploited to ultimately interface robotic actuators to comply with basic tasks in a closed-loop manner (Krauhausen et al, 2021 ).…”

Automatic Detection of High-Frequency Oscillations With Neuromorphic Spiking Neural Networks

“…In an improved approach to detect HFO in iEEG we designed an SNN in software and validated it on hardware (Sharifshazileh et al, 2021). We next adapted the SNN to detect HFO in intraoperative ECoG (Burelo et al, 2021) and scalp EEG (Burelo et al, 2022). Different from the original publications, we compare here in detail the multiple analyses conducted in the different recording modalities (iEEG, ECoG and scalp EEG).…”

“…These inputs project to a second layer of neurons (yellow) with different synaptic parameters. Panels modified from Burelo et al (2021Burelo et al ( , 2022 and Sharifshazileh et al (2021). For HFO detection in iEEG, we used the neurons in rows 1 and 2.…”

“…We did not find a case where the scalp EEG recording showed HFO in a patient who achieved seizure freedom (FP = 0). Therefore, the SNN associated the HFO rate with active epilepsy with 80% accuracy and 100% specificity (Burelo et al, 2022; Table 3).…”