Advanced Analysis of OCT/OCTA Images for Accurately Differentiating Between Glaucoma and Healthy Eyes Using Deep...

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Glaucoma diagnosis relies on detecting structural and functional changes.
• Optical coherence tomography (OCT) and OCT angiography (OCTA) provide detailed retinal imaging.
• Deep learning (DL) shows promise in analyzing complex medical images for disease detection.

Purpose of the Study:

• To assess the discriminative ability of OCT and OCTA images.
• To identify the optimal combination of OCT and OCTA imaging for differentiating glaucoma from healthy eyes.
• To utilize a convolutional neural network (CNN) for image analysis and classification.

Main Methods:

• A cross-sectional study involving 157 subjects and 1,106 eye scans.
• Analysis of en-face images from superficial and choroid layers for vessel density (OCTA) and structural thickness (OCT) of the macula (M) and optic disc (D).
• A two-component CNN architecture for feature extraction and classification of eyes as healthy or glaucomatous, with performance evaluated by accuracy, sensitivity, specificity, and AUC.

Main Results:

• For OCT images, combining macula and disc (D+M) outperformed individual macular or disc imaging.
• OCTA D+M imaging demonstrated superior performance over D or M alone across all metrics.
• In disc imaging, OCTA outperformed OCT and combined OCT+OCTA in accuracy, sensitivity, and specificity.
• OCTA D+M combination showed the best overall performance, followed by OCT+OCTA D+M.

Conclusions:

• The OCTA D+M combination is the most effective for glaucoma diagnosis.
• When both OCT and OCTA en-face images are available, OCTA is preferred.
• Integrating both disc and macula imaging is crucial for achieving optimal diagnostic accuracy in glaucoma detection.