Neurophysiological assessment using a cloud computing platform

Abstract

A methodology for the identification of the human eye movement system (EMS) using integral Volterra models in the form of first- and second-order transient characteristics and the formation of diagnostic features based on these characteristics for the classification of the neurophysiological state is proposed. Identification was carried out using experimental “input-output” data obtained under visual test stimuli with different distances from the initial position on the monitor screen and the corresponding responses recorded by the Tobii Pro TX300 eye tracker. The least squares method was applied for model construction. A feature space of heuristic features and a feature space derived from wavelet decomposition coefficients were formed. To assess the fatigue state, the Support Vector Machine method with a Gaussian kernel was used. The reliability of the assessment was determined by the probability of correct recognition (PCR) on constructed datasets through an exhaustive search of all possible pairs of features. The computations were performed on a cloud computing platform combining PaaS and SaaS services, which enables effective work in research and educational tasks both with program code in multiple programming languages and with implemented identification methods via GUI interfaces. The research results demonstrate a maximum PCR of 93.75% when using features obtained through wavelet decomposition, confirming the practical applicability of the proposed approach for the intelligent diagnosis of the human neurophysiological state.