An explainable RSPWVD based EEG microstate sequence approach for scalp connectivity estimation and channel selection in patients with epilepsy.

BackgroundElectroencephalography (EEG), a noninvasive technique for recording the brain's electrical activity, has been widely utilized to investigate neurological disorders.ObjectiveEEG recordings can estimate scalp connectivity and select representative channels, which reveal network connectivity and associated brain regions. These details are considered essential for understanding the characteristics of neurological disorders.MethodsThis work proposes an explainable Reassigned Smoothed Pseudo Wigner-Ville Distribution (RSPWVD) based EEG microstate sequence approach to achieve scalp connectivity estimation and channel selection. Epilepsy, one of the most frequently studied neurological disorders using EEG, has been selected for method validation. Receiver Operating Characteristic (ROC) curve analysis and consistency analysis with conventional techniques are performed to specify key parameters such as connection thresholds and time durations, ensuring the reliability of the outcomes.ResultsThe experimental results of the clinical Karunya dataset indicate that the proposed microstate sequence compressed from the EEG contains sufficient information to estimate scalp connectivity and select representative channels. The scalp connectivity results reveal differences between focal and generalized seizures, where focal seizures exhibit more localized connectivity and generalized seizures display a widespread distribution. Moreover, statistical results demonstrate that the F4, C4, T4, and P4 channels present a higher rate of being representative channels in this dataset.ConclusionsThe proposed approach offers valuable characteristics, indicating brain networks that assist in epilepsy analysis by focusing on the most informative scalp locations and reducing computational complexity. It lays the groundwork for investigating various neurological disorders through scalp behaviors from EEG, guiding personalized diagnostics and therapeutic strategies.