Non-Eye Tracking, Deep Learning-enabled Detection of Nystagmus in Dizzy Patients

Narayani Wagle, John Morkos, Jorge Otero-Millan, Jingyan Liu, Joseph Greenstein, Kirby Gong, Daniil Pakhomov, Sanchit Hira, Indranuj Gangan, Oleg Komogortsev, Raimond Winslow, David Newman-Toker, David S Zee, Kemar E. Green

Keywords: deep learning, eye tracking, remote neurologic diagnosis, image analysis, action recognition

TL;DR: Deep learning detection of abnormal eye movements from videos

Abstract: Patients with dizziness related to disruption of the ear-brain-eye sensory and neural circuitry often present with a particular pattern of ocular instability called nystagmus. These subtle eye movements can be difficult to detect and interpret at the bedside, and usually require robust eye tracking devices for accurate quantification. Here, we adopted an image processing and deep learning approach to detect nystagmus directly from videos from a small clinical dataset without applying traditional eye tracking techniques. Classification with our best performing model resulted in an AU-ROC of 0.864. This method may have potential future applications in augmented/virtual reality (AR/VR) eye tracking for healthcare purposes

Registration: I acknowledge that publication of this at MIDL and in the proceedings requires at least one of the authors to register and present the work during the conference.

Authorship: I confirm that I am the author of this work and that it has not been submitted to another publication before.

Paper Type: both

Primary Subject Area: Detection and Diagnosis

Secondary Subject Area: Application: Ophthalmology

Confidentiality And Author Instructions: I read the call for papers and author instructions. I acknowledge that exceeding the page limit and/or altering the latex template can result in desk rejection.