Towards the automated detection of interictal epileptiform discharges with magnetoencephalography

Abstract

The analysis of clinical magnetoencephalography (MEG) in patients with epilepsy traditionally relies on the visual identification of interictal epileptiform discharges (IEDs), which is time consuming and dependent on (subjective) human criteria. Data-driven approaches enabling both spatial and temporal localization of epileptic spikes would represent a major leap forward in clinical MEG practice. Here, we explore the ability of Independent Components Analysis (ICA) and Hidden Markov Modeling (HMM) to automatically detect and localize IEDs. Combined with kurtosis mapping, we developed a fully automated identification of epileptiform independent components (ICs) or HMM states. We tested our pipeline on MEG recordings at rest from 10 school-age children with either focal or multifocal epilepsy and compared results with the traditional MEG analysis performed by an experienced clinical magnetoencephalographer. In patients with focal epilepsy, both ICA- and HMM-based pipelines successfully detected visually identified IEDs with high sensitivity, but also revealed low-amplitude IEDs unidentified by the visual detection. Success was more mitigated in patients with multifocal epilepsy, as our automated pipeline missed IED activity associated with some foci--an issue that could be alleviated by post-hoc manual selection of epileptiform ICs or HMM states. Therefore, IED detection based on ICA or HMM represents an efficient way to identify spike localization and timing, with heightened sensitivity to IEDs compared to visual MEG signal inspection and requiring minimal input from clinical practitioners.