Automatic detection of epilepsy from EEGs using a temporal convolutional network with a self-attention layer.

BACKGROUND: Over 60% of epilepsy patients globally are children, whose early diagnosis and treatment are critical for their development and can substantially reduce the disease's burden on both families and society. Numerous algorithms for automated epilepsy detection from EEGs have been proposed. Yet, the occurrence of epileptic seizures during an EEG exam cannot always be guaranteed in clinical practice. Models that exclusively use seizure EEGs for detection risk artificially enhanced performance metrics. Therefore, there is a pressing need for a universally applicable model that can perform automatic epilepsy detection in a variety of complex real-world scenarios. METHOD: To address this problem, we have devised a novel technique employing a temporal convolutional neural network with self-attention (TCN-SA). Our model comprises two primary components: a TCN for extracting time-variant features from EEG signals, followed by a self-attention (SA) layer that assigns importance to these features. By focusing on key features, our model achieves heightened classification accuracy for epilepsy detection. RESULTS: The efficacy of our model was validated on a pediatric epilepsy dataset we collected and on the Bonn dataset, attaining accuracies of 95.50% on our dataset, and 97.37% (A v. E), and 93.50% (B vs E), respectively. When compared with other deep learning architectures (temporal convolutional neural network, self-attention network, and standardized convolutional neural network) using the same datasets, our TCN-SA model demonstrated superior performance in the automated detection of epilepsy. CONCLUSION: The proven effectiveness of the TCN-SA approach substantiates its potential as a valuable tool for the automated detection of epilepsy, offering significant benefits in diverse and complex real-world clinical settings.

Diagnosis of Autism Spectrum Disorder by Dynamic Local Graph-Theory Indicators Based on Electroencephalogram.

BACKGROUND: Autism Spectrum Disorder (ASD) is a complex neurodevelopment disease characterized by impaired social and cognitive abilities. Despite its prevalence, reliable biomarkers for identifying individuals with ASD are lacking. Recent studies have suggested that alterations in the functional connectivity of the brain in ASD patients could serve as potential indicators. However, previous research focused on static functional-connectivity analysis, neglecting temporal dynamics and spatial interactions. To address this gap, our study integrated dynamic functional connectivity, local graph-theory indicators, and a feature-selection and ranking approach to identify biomarkers for ASD diagnosis. METHODS: The demographic information, as well as resting and sleeping electroencephalography (EEG) data, were collected from 20 ASD patients and 25 controls. EEG data were pre-processed and segmented into five sub-bands (Delta, Theta, Alpha-1, Alpha-2, and Beta). Functional-connection matrices were created by calculating coherence, and static-node-strength indicators were determined for each channel. A sliding-window approach, with varying widths and moving steps, was used to scan the EEG series; dynamic local graph-theory indicators were computed, including mean, standard deviation, median, inter-quartile range, kurtosis, and skewness of the node strength. This resulted in 95 features (5 sub-bands × 19 channels) for each indicator. A support-vector-machine recurrence-feature-elimination method was used to identify the most discriminative feature subset. RESULTS: The dynamic graph-theory indicators with a 3-s window width and 50% moving step achieved the highest classification performance, with an average accuracy of 95.2%. Notably, mean, median, and inter-quartile-range indicators in this condition reached 100% accuracy, with the least number of selected features. The distribution of selected features showed a preference for the frontal region and the Beta sub-band. CONCLUSIONS: A window width of 3 s and a 50% moving step emerged as optimal parameters for dynamic graph-theory analysis. Anomalies in dynamic local graph-theory indicators in the frontal lobe and Beta sub-band may serve as valuable biomarkers for diagnosing autism spectrum disorders.