A topological-aware automatic grading model corneal epithelial damage evaluation from full Corneal Fluorescence...

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Corneal Fluorescence Staining (CFS) imaging is vital for assessing corneal epithelial damage.
• Automating CFS image grading reduces subjectivity and improves diagnostic efficiency.
• Current methods often overlook spatial distribution of stained regions, limiting accuracy.

Purpose of the Study:

• To develop a three-stage automatic model for corneal epithelial damage assessment using CFS images.
• To optimize grading by integrating topological features of stained regions.
• To enhance the accuracy of corneal epithelial injury evaluation.

Main Methods:

• Accurate corneal localization using intensity and phase information.
• Detection of stained regions via multi-scale morphological top-hat operator.
• Construction of a multi-scale graph with topological, textural, and morphological features for an ensemble grading model.

Main Results:

• The proposed model achieved high Accuracy (0.7589), F1 score (0.7449), and AUC (0.9335) on an in-house dataset.
• Topological features significantly outperformed other individual features in grading.
• The model demonstrates effective assessment of corneal epithelial damage.

Conclusions:

• The developed model effectively grades corneal epithelial damage using CFS images.
• Incorporating topological features is crucial for accurate assessment of spatial staining patterns.
• This approach holds potential for improved disease classification and clinical decision-making in dry eye management.