Convolutional neural networks for quantitative smartphone video nystagmography: ConVNG

Abstract

Background Eye movement abnormalities are paramount in neurological disorders. However, unaided eye movement assessments lack granularity. Although videooculography (VOG) improves diagnostic accuracy, resource intensiveness preclude its broad use. To bridge this care gap, we here validate a framework for smartphone video-based nystagmography capitalizing on recent computer vision advances. Methods A recurrent convolutional network was fine-tuned for pupil tracking using >550 annotated frames: ConVNG. Slow phase velocity of optokinetic nystagmus was calculated in 10 subjects using both ConVNG and VOG. Equivalence of accuracy and precision was assessed using the two one-sample t-test (TOST) and Bayesian interval-null approaches. Results ConVNG tracking accuracy reached 9-15% of an average pupil diameter. SPV measurement accuracy was equivalent to VOG (p< .017; Bayes factors (BF) > 24). Average precision was 0.30 deg. for ConVNG and 0.12 deg. for VOG. Conclusions ConVNG enables smartphone video nystagmography with an accuracy comparable to VOG and precision approximately one order of magnitude higher than comparable ARKit applications. This serves as a blueprint for highly accessible tools with potential to accelerate progress toward precise and personalized Medicine.