Abstract
Purpose: :
Optokinetic nystagmus (OKN) is a useful tool for assessing specific aspects of visual function, particularly in young children. However, objective measurement of OKN often requires a head-mounted eye-tracker or specialized video recording equipment. The purpose of this study was to develop and test software for extracting OKN information from video footage obtained using low cost off-the-shelf video equipment.
Methods: :
Software was developed: (1) to determine directly from video footage of the face the velocity of the limbal edge, and (2) to estimate from trials containing the velocity signal, the direction and consistency of OKN (i.e., an OKN "strength" measure).A novel approach based on optical flow computation was used for the velocity estimation step. The OKN "strength" was determined by a normalized average of high velocity peaks within the velocity record. The resulting measure was used to predict the direction (by the sign) and consistency (magnitude) of OKN within a trial.The software was tested on video footage of participants’ eyes (N = 6, average age 25) as they viewed random dot kinematogram stimuli presented either at 100% motion coherence (73 trials) or 30-37.5% motion coherence (42 trials) with each trial lasting 8 seconds. The coherently moving dots in the stimulus moved to the left or right on each trial. For both the low and high coherence trials the direction estimated by the OKN detector was compared with the known direction of the stimulus pattern, as well as the subjective direction determined from the video footage by an experienced observer.
Results: :
The OKN detector predicted the veridical direction of the stimulus with 93% accuracy overall (96% for the high coherence group, and 88% for the low coherence group). A slightly lower accuracy (90%) was obtained when compared against the subjectively assessed data (96% for the high coherence group, and 83% for the low coherence group). The difference arose as two trials were incorrectly classified by the assessor. The incorrect detections (<10%) were correlated with low values of OKN strength measure, which was consistent with averaging out of leftward and rightward velocity features present in a trial.
Conclusions: :
Our software system is able to extract OKN from video footage, and predicts direction/strength with good accuracy (>90%). Our approach will be a useful tool for studies involving OKN analysis, where only off the shelf equipment is available and may provide the basis for the objective measurement of OKN in a range of clinical environments without the need for costly eye-tracking equipment.
Keywords: eye movements: recording techniques • nystagmus • image processing