Demodicosis Mite Detection in Eyes with Blepharitis and Meibomian Gland Dysfunction Based on Deep Learning Model

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Medical Imaging

Background:

• Demodex mites cause underdiagnosed ocular surface diseases like blepharitis and meibomian gland dysfunction (MGD).
• Current diagnostic methods rely on manual microscopy, which is time-consuming and labor-intensive.

Purpose of the Study:

• To develop and evaluate a deep learning system for automated detection and quantification of Demodex mites.
• To improve the efficiency and accuracy of diagnosing ocular demodicosis.

Main Methods:

• Collected 1610 microscopic eyelash images from patients suspected of ocular demodicosis.
• Trained and evaluated two deep learning models, YOLOv11 and RT-DETR, on 665 annotated images.
• Utilized Grad-CAM for model visualization and quantitative analysis for mite counting.

Main Results:

• Both YOLOv11 and RT-DETR models successfully detected Demodex mites.
• YOLOv11 achieved high performance with an average precision of 0.9441, sensitivity of 0.9478, and F1-score of 0.9459.
• RT-DETR showed strong sensitivity (0.9389) and stable quantitative analysis results.

Conclusions:

• The AI-based system shows significant potential for efficient and accurate diagnosis of ocular demodicosis.
• This technology can reduce reliance on manual microscopy, leading to faster clinical decisions.
• Automated detection assists ophthalmologists in managing Demodex-related eye conditions.