A sparse multiscale nonlinear autoregressive model for seizure prediction

Efficacy of novel attention-based gated recurrent units transformer for depression detection using electroencephalogram signals

Purpose

Depression is a global challenge causing psychological and intellectual problems that require efficient diagnosis. Electroencephalogram (EEG) signals represent the functional state of the human brain and can help build an accurate and viable technique for the early prediction and treatment of depression.

Methods

An attention-based gated recurrent units transformer (AttGRUT) time-series model is proposed to efficiently identify EEG perturbations in depressive patients. Statistical, spectral and wavelet features were first extracted from the 60-channel EEG signal data. Then, two feature selection techniques, recursive feature elimination and the Boruta algorithm, both with Shapley additive explanations, were utilised for selecting essential features.

Results

The proposed model outperformed the two baseline and two hybrid time-series models-long short-term memory (LSTM), gated recurrent units (GRU), convolutional neural network-LSTM (CNN-LSTM), and CNN-GRU-achieving an accuracy of up to 98.67%. Feature selection considerably increased the performance across all time-series models.

Conclusion

Based on the obtained results, novel feature selection greatly affected the results of the baseline and hybrid time-series models. The proposed AttGRUT can be implemented and tested in other domains by using different modalities for prediction.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13755-022-00205-8.

Epileptic seizure prediction using successive variational mode decomposition and transformers deep learning network

As one of the most common neurological disorders, epilepsy causes great physical and psychological damage to the patients. The long-term recurrent and unprovoked seizures make the prediction necessary. In this paper, a novel approach for epileptic seizure prediction based on successive variational mode decomposition (SVMD) and transformers is proposed. SVMD is extended to multidimensional form for time-frequency analysis of multi-channel signals. It could adaptively extract common band-limited intrinsic modes among all channels on different time scales by solving a variational optimization problem. In the proposed seizure prediction method, data are first decomposed into multiple modes on different time scales by multivariate SVMD, and then, irrelevant modes are removed for preprocessing. Finally, power spectrum of denoised data is input to a pre-trained bidirectional encoder representations from transformers (BERTs) for prediction. The BERT could identify the mode information related to epileptic seizures in time-frequency domain. It shows fair prediction performance on an intracranial EEG dataset with the average sensitivity of 0.86 and FPR of 0.18/h.

Epileptic seizure prediction using spectral width of the covariance matrix

Epilepsy is a common neurological disorder in which patients suffer from sudden and unpredictable seizures. Seizures are caused by excessive and abnormal neuronal activity. Different methods have been employed to investigate electroencephalogram (EEG) data in patients with epilepsy. This paper introduces a simple yet accurate array-based method to study and predict seizures. We use the CHB-MIT dataset (all 24 cases), which includes scalp EEG recordings. The proposed method is based on the random matrix theory. After applying wavelet decomposition to denoise the data, we analyze the spatial coherence of the epileptic recordings by looking at the width of the covariance matrix eigenvalue distribution at different time and frequency bins. We train patient-specific support vector machine (SVM) classifiers to distinguish between interictal and preictal data with high performance and a false prediction rate as low as 0.09/h. The proposed technique achieves an average accuracy, specificity, sensitivity, and area under the curve (AUC) of 99.05%, 93.56%, 99.09%, and 0.99, respectively. Our proposed method outperforms state-of-the-art works in terms of sensitivity while maintaining a low false prediction rate. Also, in contrast to neural networks, which may achieve high performance, this work provides high sensitivity without compromising interpretability.