Real-time epilepsy seizure detection and brain connectivity analysis using electroencephalogram

Real-time epilepsy seizure detection and brain connectivity analysis using electroencephalogram

This thesis integrates signal processing techniques, functional brain connectivity analysis, and artificial intelligence methods to address the challenges of real-time epilepsy seizure detection and provide insights into complex brain disorders, such as alcoholism and schizophrenia (ScZ). By leveraging electroencephalogram (EEG) data and advanced analysis methods, the research aims to achieve high accuracy detection and identification of biomarkers for these conditions. The real-time epilepsy seizure detection involves three main steps: pre-processing, feature extraction, and machine learning (ML) or deep learning (DL) models for classification and detection. Various signal processing methods, such as discrete wavelet transform, tunable-Q wavelet transform, and short-time Fourier transform, are applied in the pre-processing stage to decompose the EEG signals into time domain, frequency domain, and time-frequency domain data. Feature extraction techniques, such as statistical moments, entropy algorithms, and power spectrum analysis, are used to detect the sharp waves indicative of seizure activity. ML and DL models, such as support vector machines and convolutional neural networks (CNN), are employed to address the robustness challenges and provide high accuracy classification results for seizure detection. The EEG brain connectivity analysis focus on the correlations between different EEG channel signals within the complex brain network. Alcoholism and ScZ datasets are utilized in three experiments. The signal processing methods include continuous wavelet transform and multivariate autoregressive models to extract relevant features from the EEG signals. Functional brain connectivity is then calculated using mutual information and coherence algorithms. DL models, particularly CNN, are employed to classify patients and healthy control subjects based on the calculated functional connectivity. Furthermore, statistical analysis of the entire brain connectivity is conducted to identify biomarkers of abnormal connectivity patterns associated with complex brain disorders.

File reproduced in accordance with the copyright policy of the publisher/author.