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

Abstract

Interictal high-frequency oscillations (HFO) detected in electroencephalography recordings have been proposed as biomarkers of epileptogenesis, seizure propensity, disease severity, and treatment response. Automatic HFO detectors typically analyze the data offline using complex time-consuming algorithms, which limits their clinical application. Neuromorphic circuits offer the possibility of building compact and low-power processing systems that can analyze data on-line and in real time. In this review, we describe a fully automated detection pipeline for HFO that uses, for the first time, spiking neural networks and neuromorphic technology. We demonstrated that our HFO detection pipeline can be applied to recordings from different modalities (intracranial electroencephalography, electrocorticography, and scalp electroencephalography) and validated its operation in a custom-designed neuromorphic processor. Our HFO detection approach resulted in high accuracy and specificity in the prediction of seizure outcome in patients implanted with intracranial electroencephalography and electrocorticography, and in the prediction of epilepsy severity in patients recorded with scalp electroencephalography. Our research provides a further step toward the real-time detection of HFO using compact and low-power neuromorphic devices. The real-time detection of HFO in the operation room may improve the seizure outcome of epilepsy surgery, while the use of our neuromorphic processor for non-invasive therapy monitoring might allow for more effective medication strategies to achieve seizure control. Therefore, this work has the potential to improve the quality of life in patients with epilepsy by improving epilepsy diagnostics and treatment.

Robust compression and detection of epileptiform patterns in ECoG using a real-time spiking neural network hardware framework

“…The Spectrum detector was developed on UMCU intraoperative ECoG recordings 11 and then applied to the USZ intraoperative ECoG data 9 , 10 . The SW-SNN detector was then applied to the same dataset for the patients whose resection was guided by hd-ECoG 15 , 16 .…”

“…Due to their event-based processing, spiking neural networks (SNN) are particularly well-suited for bio-signal analysis. Recent work has been done in the detection of HFO with SNN in intracranial EEG 14 , ECoG 15 , 16 , and scalp EEG 17 . In particular, HFO detection in the presurgical intracranial EEG 14 has been performed using ultra-low-power mixed-signal neuromorphic hardware (DYNAP-SE) 18 .…”

“…Moreover, the implemented SNN was performing single-channel analysis and did not account for possible artifacts in the EEG that can produce high-amplitude activity in the HFO band. The SNN tested on ECoG data 15 , 16 introduced an artifact rejection stage, but the network was only simulated in software and therefore it was not suitable for real-time applications.…”

Scalp high‐frequency oscillations differentiate neonates with seizures from healthy neonates

“…These approaches, drawing from intracranial human and animal data, will, however first have to be adapted and validated on scalp EEG. Furthermore, dedicated HFO detection devices 46 may facilitate future long‐term scalp HFO monitoring.…”

Recent advances in clinical electroencephalography

“…Apart from innovative approaches to optimize the recording of HFOs, various groups focused on the development of more comprehensive detection algorithms as well as the use of new recording technologies using high-resolution cortical arrays with large coverage. In contrast to traditional detectors based on time frequency or signal amplitude features alone [32], the latest HFO detector generation is focusing on not only HFO detection but also on de-noising true HFO detections from artifacts as well as real-time HFO detection, something that is pertinent in intraoperative settings [33,34 ▪ ,35]. The benefit of such systems was shown in a recent work that demonstrated that a deep learning-based classification was able to distill pathological HFOs, regardless of the initial HFO detection methods; this is promising, as it might aid to increase specificity of HFOs for localization of the epileptogenic zone [36].…”