A deep learning architecture to detect events in EEG signals during sleep
This addresses the need for automated, less tedious event detection in sleep medicine for clinicians, but it is incremental as it builds on existing deep learning approaches.
The paper tackles the problem of automatically detecting both macro- and micro-events in EEG signals during sleep, which is time-consuming and variable when done manually, by proposing a deep learning method that jointly predicts event locations, durations, and types, achieving efficiency compared to state-of-the-art event-specific algorithms.
Electroencephalography (EEG) during sleep is used by clinicians to evaluate various neurological disorders. In sleep medicine, it is relevant to detect macro-events (> 10s) such as sleep stages, and micro-events (<2s) such as spindles and K-complexes. Annotations of such events require a trained sleep expert, a time consuming and tedious process with a large inter-scorer variability. Automatic algorithms have been developed to detect various types of events but these are event-specific. We propose a deep learning method that jointly predicts locations, durations and types of events in EEG time series. It relies on a convolutional neural network that builds a feature representation from raw EEG signals. Numerical experiments demonstrate efficiency of this new approach on various event detection tasks compared to current state-of-the-art, event specific, algorithms.