Analyzing selected visual anomaly through ST-based multi-resolution VEP decomposition

Abstract

Visual anomaly interferes with the normal sight, and therefore can cause discomfort. Ophthalmologist describes a vision anomaly as any kind of vision loss, be it partial or total vision loss. The cause of vision loss can be traced back to genetic conditions, improper lifestyle, medical conditions and accidental causes. Visually evoked potentials (VEP) are often used in cases where the medical experts might have difficulties in determining the cause of the problem. VEP has often been reported in the literature as a reliable method for detecting visual abnormalities. By analyzing the recorded brain responses toward a visual stimulus, it is possible for the ophthalmologist to make an appropriate estimation. Since the analysis of VEP by the medical experts is still subjective, there is a need for a more comprehensive analysis method. Researchers are now trying to explore the spectral information of the VEP for easier diagnosis. In this regard, this work explores the use of Stockwell transform (ST) for the analysis of VEP signals. Via ST-based multi-resolution decomposition of the signals, the time-frequency information matrix (ST-matrix) is obtained. Statistical features extracted from ST-matrix were classified using least square support vector machine (LSSVM) and probabilistic neural network (PNN). The investigation results suggest that the LSSVM classifier was able to provide 96.07% of accurate prediction for selected visual anomaly investigated.