Automated anterior chamber angle localization and glaucoma type classification in OCT images

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Glaucoma diagnosis relies on identifying the anterior chamber angle (ACA) in optical coherence tomography (OCT) images.
• Accurate and efficient classification of ACA is crucial for determining glaucoma type.

Purpose of the Study:

• To develop an automated image processing and machine learning framework for accurate ACA localization and classification in OCT images.
• To differentiate between open-angle (OA) and angle-closure (AC) glaucoma using visual features.

Main Methods:

• Automated localization of the ACA region in digital OCT images.
• Extraction of visual features from the ACA region for OA/AC classification.
• Comparison of high-dimensional visual features against low-dimensional clinical features.

Main Results:

• The framework achieved an Area Under the Curve (AUC) of 0.921 ± 0.036.
• Balanced accuracy reached 84.0% ± 5.7% at 85% specificity.
• The method processed images at 0.26s per image on a dataset of 2048 images.

Conclusions:

• The proposed automated framework accurately classifies glaucoma type using visual features from OCT images.
• High-dimensional visual features provide superior classification accuracy compared to traditional clinical features.
• This method offers an efficient and automated approach for glaucoma diagnosis.