An automated and multiparametric algorithm for objective analysis of meibography images

Area of Science:

• Ophthalmology and optometry
• Medical imaging analysis
• Biomedical engineering

Background:

• Meibography is a non-contact imaging technique crucial for evaluating meibomian glands and diagnosing meibomian gland dysfunction (MGD).
• Current manual analysis of meibography images lacks repeatability and efficiency.
• Automated, quantitative, and multiparametric analysis is needed for comprehensive MGD assessment, especially for detecting subtle gland changes.

Purpose of the Study:

• To develop a fully automated algorithm for multiparametric analysis of meibography images.
• To enable objective and repeatable quantification of meibomian gland morphology and function.
• To improve the diagnostic accuracy and efficiency of MGD evaluation.

Main Methods:

• Developed a novel algorithm for fully automatic and repeatable segmentation of meibography images.
• The algorithm includes tarsal conjunctiva segmentation, meibomian gland identification, and quantitative multiparametric analysis.
• Key parameters include gland area ratio (GA), length (L), width (D), diameter deformation index (DI), tortuosity index (TI), and signal index (SI).

Main Results:

• High similarity (k=0.94±0.02 for ROI, k=0.87±0.01 for glands) and low segmentation errors (false-positive rate < 7%, false-negative rate < 19%) were achieved compared to manual segmentation.
• The algorithm successfully processed meibography images from subjects with varying MGD statuses.
• Quantitative parameters (GA, L, D, DI, TI, SI) were generated for objective gland assessment.

Conclusions:

• A fully automated algorithm for meibography image analysis has been successfully developed.
• The algorithm demonstrates high accuracy and offers multiple quantitative parameters for objective evaluation of meibomian gland morphology and function.
• This tool provides a significant advancement for the objective assessment and diagnosis of MGD.