A fuzzy set theoretical approach to automatic analysis of nystagmic eye movements

A method for computer analysis of nystagmic eye movements in vestibulo-ocular (VOR) and optokinetic (OKN) reflexes is developed. A fuzzy set theoretical approach is used to construct the slow cumulative eye position (SCEP) curve by eliminating fast components (saccades) from the eye movement signal. These procedures are able to perform automatically some pattern recognition tasks traditionally used in classical interactive programs when human operators distinguish between fast and slow phases of eye movements. The structure of the algorithm is as follows. A fuzzy clusters of slow and fast phases is made. An iterative method is used to refine the membership function of slow-phases, step by step, until a sufficiently discriminating membership function is obtained. Saccades are detected and removed from the eye position signal. SCEP is then built by interpolating between slow phases. A weighted least-squares curve fitting is made. Weighting coefficients are obtained from the last membership function resulting from the iterations. The curve fitting is referenced to the SCEP, and the parameters of VOR and OKN are calculated using this last curve.