Brainwave Monitoring in Epileptic Pediatric Patient using MSAR-Normal, MSAR-Azzalini, and MSARExponential Power

This study aims to model brain wave data in children with special needs, specifically focusing on epilepsy, using the Markov Switching Autoregressive (MSAR) model. Given the substantial prevalence of children suffering from epilepsy in Indonesia, the implementation of precise interventions and monitoring is critical to fostering their cognitive and psychological well-being. Electroencephalograph (EEG) data were acquired from the Airlangga University Hospital, specifically focusing on the temporal region at the T3Cz channel. The MSAR-Normal model was selected for its capacity to effectively modeling the shifts in pattern dynamics and temporal dependencies. In nonlinear condition, challenges were encountered in satisfying the residual white noise assumption. The Bayesian approach was employed, to address these challenges by comparing the performance of the added distributions, Azzalini and Exponential Power (EP), into the MSAR model which are then called MSAR-Azzalini and MSAR-EP. The performance of this approach is evaluated against MSAR-Normal. The findings indicate that the MS(2)AR(1) model, particularly when utilizing the MSAR-EP, best represents the T3-Cz Channel. It achieves the AIC value of -1096.9, lower than the MSAR-Normal and MSAR-Azzalini model of 21813.2 and 20915.56. Furthermore, highly effective in detecting abnormal waves before and during seizures, and provides accurate personalized treatments.