Perspectives of support vector regression for static posturographic assessment of patients with cognitive impairment

Abstract

The study aims to assess and quantify the discriminate parameters of balance among patients affected by vestibular dysfunction. Several data were obtained using the Satel force plate. A total of 14 patients participated in the study. The postural strategies were studied from the trajectory of the Center Of Pressure (COP), in standing position and under the cognitive “open eyes (Cognitive-Task-Free - CTF)’’ and “open eyes with a cognitive task (Cognitive-Task-Active -CTA)’’. Experimental data were used for training of the intelligent soft computing scheme Support Vector Regression (SVR). In the present study, the polynomial and Radial Basis Functions (RBF) were applied as the SVR kernel function to predict the two cognitive positions. Performance of the proposed estimators was confirmed through the simulation results. According to our findings, a greater improvement in accuracy estimation can be achieved through the SVR with radial basis function compared to SVR with polynomial basis function. The SVR coefficient of determination (R 2 ) with radial basis function was found to be equal to 0.8613, while R 2 with the polynomial basis function was equal to 0.6437.