Deep Learning Approaches for Automated Seizure Detection from Scalp Electroencephalograms

Abstract

Scalp electroencephalograms (EEGs) are the primary means by which physicians diagnose brain-related illnesses such as epilepsy and seizures. Automated seizure detection using clinical EEGs is a very difficult machine learning problem due to the low fidelity of a scalp EEG signal. Nevertheless, despite the poor signal quality, clinicians can reliably diagnose illnesses from visual inspection of the signal waveform. Commercially available automated seizure detection systems, however, suffer from unacceptably high false alarm rates. Deep learning algorithms that require large amounts of training data have not previously been effective on this task due to the lack of big data resources necessary for building such models and the complexity of the signals involved. The evolution of big data science, most notably the release of the Temple University EEG (TUEG) Corpus, has motivated renewed interest in this problem. In this chapter, we discuss the application of a variety of deep learning architectures to automated seizure detection. Architectures explored include multi-layer perceptrons, convolutional neural networks (CNNs), long short-term memory networks (LSTMs), gated recurrent units, and residual neural networks. We use the TUEG Corpus, supplemented with data from Duke University, to evaluate the performance of these hybrid deep structures. Since TUEG contains a significant amount of unlabeled data, we also discuss unsupervised pre-training methods used prior to training these complex recurrent networks. Exploiting spatial and temporal context is critical for accurate disambiguation of seizures from artifacts. We explore how effectively several conventional architectures are able to model context and introduce a hybrid system that integrates CNNs and LSTMs. The primary error modalities observed by this state-of-the-art system were false alarms generated during brief delta range slowing patterns such as intermittent rhythmic delta activity. A variety of these types of events have been observed during inter-ictal and post-ictal stages. Training models on such events with diverse morphologies has the potential to significantly reduce the remaining false alarms. This is one reason we are continuing our efforts to annotate a larger portion of TUEG. Increasing the data set size significantly allows us to leverage more advanced machine learning methodologies.