Brainwave Monitoring in Epileptic Pediatric Patient using MSAR EM-Gaussian and MSAR Bayesian-Exponential Power

The complex neurological condition of brain development in pediatric patients with epilepsy is critically important to be observed individually using EEG recording in order to facilitate better personalized treatment. In this study, the Markov Switching Autoregressive (MSAR) approach is developed by enabling Bayesian in estimating autoregressive Exponential Power distribution, namely MSAR Bayesian-Exponential Power. The development of this study is combining the MSAR coupled with Bayesian-Exponential Power to capture the fluctuating and nonlinear patterns within brain EEG. This approaches is achieved by incorporating the Expectation Maximization (EM) and Bayesian-Exponential Power distributions inside MSAR model. The performance of this approach is evaluated againts the MSAR EM-Gaussian. The model will calculate the transition probabilities of each state of temporal lobe, T6-Cz channel, which contains two waves before (Pre-Ictal) and during seizure (Ictal) events. This approach succeed to illustrates the dynamic switching system within latent brain states of the epilepsy patient more accurately than MSAR EM-Gaussian. It achieves the AIC value of -793.896, lower than the MSAR EM-Gaussian method of 27979.51. This method is highly effective in detecting abnormal waves before and during seizures, and provides accurate personalized treatments.