E-POSTER DETAIL | 11th Asia Pacific Tele Ophthalmology Society Symposium

Title
A Desktop-Based Tele-Ophthalmology Tool for Remote Blink Rate Monitoring and Intelligent Reminder Delivery Using Webcam Video Processing

Authors
KARPAGAM DAMODARAN, Manasha B.B, Jemima Hubert, Maheshwari Srinivasan, Sairam M.R, Alfred Philomin, Jeevitha A, Philo Chamberline, Antony Nikhil

Presenting
KARPAGAM DAMODARAN

PURPOSE:
To develop and validate a desktop-based tele-ophthalmology application capable of remotely monitoring real-time blink rates using live webcam video processing.

METHODS:
The application was built with a React–Tauri–Vite frontend and a Python (Flask) backend using OpenCV and FaceMeshModule for real-time landmark detection. Blink data were stored in SQLite, with thresholds updated via a Flask API. The system computed per-minute blink rates, 10-minute aggregates, and daily summaries, triggering a subtle alert when rates fell below seven blinks/min. This cross sectional comparative study included student volunteers.Manual and app-recorded blink counts were compared at 30, 40, and 50 cm and at ±10° gaze angles.

METHODS:

The application was built with a React–Tauri–Vite frontend and a Python (Flask) backend using OpenCV and FaceMeshModule for real-time landmark detection. Blink data were stored in SQLite, with thresholds updated via a Flask API. The system computed per-minute blink rates, 10-minute aggregates, and daily summaries, triggering a subtle alert when rates fell below seven blinks/min. This cross sectional comparative study included student volunteers.Manual and app-recorded blink counts were compared at 30, 40, and 50 cm and at ±10° gaze angles.

RESULTS:
Validation included 50 participants (18 males, 32 females; 18–25 years) with >4 hours of daily screen time. Regression showed significant accuracy at 30 cm (R²=0.456, p<0.001) and 40 cm (R²=0.341, p<0.001), but not at 50 cm (R²=0.034, p=0.199). Accuracy was strong with upward gaze (R²=0.401, p<0.001) and weaker but significant with downward gaze (R²=0.122, p=0.013), indicating optimal performance at closer distances and upward gaze angles.

RESULTS:

Validation included 50 participants (18 males, 32 females; 18–25 years) with >4 hours of daily screen time. Regression showed significant accuracy at 30 cm (R²=0.456, p<0.001) and 40 cm (R²=0.341, p<0.001), but not at 50 cm (R²=0.034, p=0.199). Accuracy was strong with upward gaze (R²=0.401, p<0.001) and weaker but significant with downward gaze (R²=0.122, p=0.013), indicating optimal performance at closer distances and upward gaze angles.

CONCLUSIONS:
The blink-tracking application is adaptive, accurate, and user-friendly. By providing intelligent, real-time reminders only when blink rates drop below physiological norms, it supports tele-ocular wellness programs for dry eye patients.