Convolutional Neural Network Classification of Rest EEG Signals among People with Epilepsy, Psychogenic Non Epileptic Seizures and Control Subjects

Graph neural networks for electroencephalogram analysis: Alzheimer's disease and epilepsy use cases

Electroencephalography (EEG) is widely used as a non-invasive technique for the diagnosis of several brain disorders, including Alzheimer's disease and epilepsy. Until recently, diseases have been identified over EEG readings by human experts, which may not only be specific and difficult to find, but are also subject to human error. Despite the recent emergence of machine learning methods for the interpretation of EEGs, most approaches are not capable of capturing the underlying arbitrary non-Euclidean relations between signals in the different regions of the human brain. In this context, Graph Neural Networks (GNNs) have gained attention for their ability to effectively analyze complex relationships within different types of graph-structured data. This includes EEGs, a use case still relatively unexplored. In this paper, we aim to bridge this gap by presenting a study that applies GNNs for the EEG-based detection of Alzheimer's disease and discrimination of two different types of seizures. To this end, we demonstrate the value of GNNs by showing that a single GNN architecture can achieve state-of-the-art performance in both use cases. Through design space explorations and explainability analysis, we develop a graph-based transformer that achieves cross-validated accuracies over 89% and 96% in the ternary classification variants of Alzheimer's disease and epilepsy use cases, respectively, matching the intuitions drawn by expert neurologists. We also argue about the computational efficiency, generalizability and potential for real-time operation of GNNs for EEGs, positioning them as a valuable tool for classifying various neurological pathologies and opening up new prospects for research and clinical practice.

An update on psychogenic nonepileptic seizures

PURPOSE OF REVIEW

The understanding of psychogenic nonepileptic seizures (PNES) has advanced steadily over recent decades. This update summarizes new insights from the last three years.

RECENT FINDINGS

The process of diagnosing PNES has shifted from the exclusion of epilepsy to one based on the recognition of typical clinical features. While the diagnosis cannot rely on any single feature in isolation, a range of semiological features characterising PNES are now recognised and a number of studies hint at the potential for machine learning and AI to improve the diagnostic process. Advances in data processing and analysis may also help to make sense of the heterogeneity of PNES populations demonstrated by recent studies focussing on aetiology and patient subgroups. It is now clear that PNES are associated with high rates of mental and physical comorbidities and premature death, highlighting that they are only one manifestation of a complex disorder extending beyond the nervous system and the seizures themselves.

SUMMARY

PNES are now understood as a manifestation of dysfunction in interacting brain networks. This understanding provides an explanation for the psychopathological and semiological heterogeneity of PNES patient populations. New insights into medical comorbidities and increased rates of premature death call for more research into associated pathological processes outside the nervous system.

Classification of Epileptic and Psychogenic Non-Epileptic Seizures Using Electroencephalography and Electrocardiography

Patients with psychogenic non-epileptic seizures (PNES) may exhibit similar clinical features to patients with epileptic seizures (ES). Misdiagnosis of PNES and ES can lead to inappropriate treatment and significant morbidity. This study investigates the use of machine learning techniques for classification of PNES and ES based on electroencephalography (EEG) and electrocardiography (ECG) data. Video-EEG-ECG of 150 ES events from 16 patients and 96 PNES from 10 patients were analysed. Four preictal periods (time before event onset) in EEG and ECG data were selected for each PNES and ES event (60-45 min, 45-30 min, 30-15 min, 15-0 min). Time-domain features were extracted from each preictal data segment in 17 EEG channels and 1 ECG channel. The classification performance using k-nearest neighbour, decision tree, random forest, naive Bayes, and support vector machine classifiers were evaluated. The results showed the highest classification accuracy was 87.83% using the random forest on 15-0 min preictal period of EEG and ECG data. The performance was significantly higher using 15-0 min preictal period data than 30-15 min, 45-30 min, and 60-45 min preictal periods ( [Formula: see text]). The classification accuracy was improved from 86.37% to 87.83% by combining ECG data with EEG data ( [Formula: see text]). The study provided an automated classification algorithm for PNES and ES events using machine learning techniques on preictal EEG and ECG data.