Automated anterior segment OCT image analysis for Angle Closure Glaucoma mechanisms classification.

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Angle closure glaucoma (ACG) is a global eye condition with four primary causative mechanisms.
• Accurate identification of ACG mechanisms is crucial for tailored treatment strategies.
• Current diagnostic methods rely on manual analysis of anterior segment optical coherence tomography (AS-OCT) images, requiring segmentation and clinician interpretation.

Purpose of the Study:

• To develop a fully automated method for classifying ACG mechanisms using AS-OCT images.
• To eliminate the need for manual image segmentation and parameter measurement.
• To provide a less subjective and more efficient diagnostic tool for ACG.

Main Methods:

• A novel approach extracts morphological features directly from raw AS-OCT images using compound image transforms.
• A Naïve Bayes Classifier (NBC) is trained on a selected subset of informative, low-redundancy features.
• The method operates without manual intervention or image segmentation.

Main Results:

• The automated method achieved high classification accuracies: 89.2% with leave-one-out cross-validation and 85.12% with 10-fold cross-validation.
• Demonstrated a fully automated classification of ACG mechanisms.
• Offered a less subjective alternative to existing diagnostic techniques.

Conclusions:

• The proposed method automates ACG mechanism classification directly from AS-OCT images.
• It bypasses the need for image segmentation and manual parameter extraction.
• This approach offers a completely automated and efficient diagnostic solution for ACG.