Abstract:Background: Eye movement tests remain significantly underutilized in emergency departments and primary healthcare units, despite their superior diagnostic sensitivity compared to neuroimaging modalities for the differential diagnosis of acute vertigo. This underutilization may be attributed to a potential lack of awareness regarding these tests and the absence of appropriate tools for detecting nystagmus. This study aimed to develop a nystagmus measurement algorithm using a lightweight deep-learning model that recognizes the ocular regions. Method: The deep learning model was used to segment the eye regions, detect blinking, and determine the pupil center. The model was trained using images extracted from video clips of a clinical battery of eye movement tests and synthesized images reproducing real eye movement scenarios using virtual reality. Each eye image was annotated with segmentation masks of the sclera, iris, and pupil, with gaze vectors of the pupil center for eye tracking. We conducted a comprehensive evaluation of model performance and its execution speeds in comparison to various alternative models using metrics that are suitable for the tasks. Results: The mean Intersection over Union values of the segmentation model ranged from 0.90 to 0.97 for different classes (sclera, iris, and pupil) across types of images (synthetic vs. real-world images). Additionally, the mean absolute error for eye tracking was 0.595 for real-world data and the F1 score for blink detection was ≥ 0.95, which indicates our model is performing at a very high level of accuracy. Execution speed was also the most rapid for ocular object segmentation under the same hardware condition as compared to alternative models. The prediction for horizontal and vertical nystagmus in real eye movement video revealed high accuracy with a strong correlation between the observed and predicted values ( r = 0.9949 for horizontal and r = 0.9950 for vertical; both p < 0.05). Conclusion: The potential of our model, which can automatically segment ocular regions and track nystagmus in real time from eye movement videos, holds significant promise for emergency settings or remote intervention within the field of neurotology.

Eye Tracking Using Neural Network and Mean-Shift

A Single-Camera Gaze Tracking System under Natural Light

Real-time Follow-Up Head Tracking in Dynamic Complex Environments

Eye-Gaze Tracking Research Based on Image Processing

Head Tracking by Means of Probabilistic Neural Networks

A CNN based real-time eye tracker for web mining applications

Classification of Eye Tracking Data Using a Convolutional Neural Network

Eye detection and tracking based on AdaBoost and Kalman algorithms

Image based Eye Gaze Tracking and its Applications

Eye movement analysis for real-world settings using segmented linear regression

Eye gaze tracking based on analysis of electrooculography and steady-state visual evoked potentials

Eye Tracking for Everyone

An integrated neural network model for eye-tracking during human-computer interaction

Eye tracking method based on mobile big data in computer environment

[Implementation of Eye Movement Tracking System Based on Camshift Algorithm].

Eye-Tracker to Control Pointer Using Machine Learning

Towards Hardware-Agnostic Gaze-Trackers

Convolutional Neural Network Implementation for Eye Movement Recognition Based on Video

A Novel Real-Time Eye Detection in Human-Computer Interaction

Bayesian Eye Tracking

Feasibility of video-based real-time nystagmus tracking: a lightweight deep learning model approach using ocular object segmentation