Automated Optokinetic Nystagmus Detection for Use with Young Subjects

Abstract

There are currently no objective tests of visual function for young children and the tests that do exist are subjective, can only be applied reliably in children over 3 years of age and require high levels of specialist clinical training. To overcome these problems, we have developed a low cost, portable, clinically applicable system that assesses visual function by inducing and measuring an involuntary, reflexive eye movement known as optokinetic nystagmus (OKN). This eye movement only occurs when a moving target is visible. First, we developed a novel semi-automated method for OKN detection able to separate eye movements from the raw head/eye signal using stabilization of the eye region. The stabilization method was validated using direct subtraction of head motions from combined head-eye movements. The detection of OKN-like peaks agreed well with subjective judgements provided by an experienced observer (accuracy=85%). Secondly, we assessed whether frame rates achievable with a low cost commercially available camera were sufficient to measure key parameters of OKN. We used a webcam (30 Hz) and a research grade eye tracker (420 Hz) to record OKN in 10 adults. Downsampling was used to simulate eye tracker data at 30, 60, 120, 240 Hz. We found that for OKN detection, there was no need to use high frame rate cameras, but for accurate measurement of OKN parameters higher frame rate (an average of 255 Hz) was required. Finally, we applied our methods to provide a fully automated estimation of global motion coherence thresholds. We combined our existing methods with the Visage SDK head and eye tracking software (Visage SDK). The improved automated OKN detection system was used as an objective method for estimation of global motion coherence thresholds in 20 two-year old children. We found good agreement between the objective and subjective (gold standard) measurements of global motion coherence threshold (83.01% 10.45% and 83.36 11.90% respectively). Overall, we found that OKN detection does not require any calibration procedures, head rests or special hardware. Our new algorithms could be a core component of a device that provides measures of visual functions such as visual acuity, contrast sensitivity and global motion coherence in children as young as two years old.