Reliable feature selection for automated angle closure glaucoma mechanism detection

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Glaucoma is a leading cause of vision loss due to optic nerve damage, often linked to high intraocular pressure.
• Primary angle-closure glaucoma (PACG) encompasses multiple underlying mechanisms, including pupil block, plateau iris, peripheral iris roll, and lens-related issues.
• Accurate identification of these PACG mechanisms is crucial for effective diagnosis and treatment.

Purpose of the Study:

• To develop a novel method for selecting reliable discriminant features from anterior segment optical coherence tomography (AS-OCT) images for PACG mechanism detection.
• To enhance the accuracy of classifying the four primary mechanisms of PACG.

Main Methods:

• Feature selection algorithms were employed to identify discriminant features in AS-OCT images for PACG detection.
• A novel criteria combining top-ranked features from multiple algorithms was proposed for enhanced feature selection.
• The selected features were used with an AdaBoost classifier to differentiate PACG mechanisms.

Main Results:

• The proposed feature selection method outperformed individual algorithms in identifying reliable features for PACG detection.
• The combination of selected features led to superior classification accuracy compared to features selected by single methods.
• The AdaBoost classifier achieved high accuracy in classifying the four PACG mechanisms using the optimized feature set.

Conclusions:

• The developed feature selection approach provides a more reliable set of discriminant features for PACG analysis.
• This method significantly improves the accuracy of classifying the specific mechanisms underlying primary angle-closure glaucoma.
• The findings contribute to more precise diagnostic tools for PACG using AS-OCT imaging.