A 2D feature space representation of the optokinetic velocity signal

Abstract

Early detection of visual dysfunction can lead to improved treatment outcomes. However, current standard visual acuity (VA) tests are subjective and impractical to apply in young children. Optokinetic nystagmus (OKN) is an involuntary reflexive movement of the eyes induced by a dynamic, patterned visual stimulus. OKN is characterized by alternating phases of slow stimulus tracking (pursuit), followed by quick refixation movements in the opposite direction to stimulus motion. The presence or absence of OKN may be used to indicate visual function. Previously we investigated an approach to measure the horizontal velocity component of the eye, in which OKN was indicated by large amplitude peaks corresponding to the characteristic quick refixation movements. In this paper, we used pre-peak and post-peak amplitudes to from a convenient and simple 2D feature space representation of the optokinetic signal. Video data (each lasting around 5 seconds) was obtained as a subject (N=1) viewed drifting patterns (randomly assigned directions) designed to elicit OKN under the conditions of visible drifting patterns (5 trials with LogMar = 1.0) and difficult to see drifting patterns (5 trials LogMar = 0.0). The resulting eye movement videos were subjectively analyzed by trained observers, which gave annotations indicating intervals containing OKN, eye blinks, and head movements. Results showed that the detected eye quick phases, represented by velocity peaks, were robustly correlated with the OKN annotations. A Wilicoxon test showed that the velocity of OKN peaks were statistically different from sources of noise in the videos. Trials without OKN signals, indicated that head movement and eye searching trials, can be differentiated from OKN trials. This feature based representation of optokinetic signals can be used as a basis for future work in which we will separate OKN signals, as we work towards approaches to measure visual acuity in young children.