Corneal Layer Segmentation in Healthy and Pathological Eyes: A Joint Super-Resolution Generative Adversarial Network...

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Accurate segmentation and thickness measurement of corneal layers are crucial for diagnosing and monitoring various eye conditions.
• Existing methods for corneal layer analysis using optical coherence tomography (OCT) may lack precision, especially for pathological cases.

Purpose of the Study:

• To enhance the accuracy of corneal layer segmentation and thickness measurement in ultra-high axial resolution OCT images.
• To develop a novel approach combining super-resolution generative adversarial network (SRGAN) and adaptive graph theory for improved corneal analysis.
• To evaluate the method's performance in healthy eyes and eyes with conditions like Fuchs endothelial corneal dystrophy (FECD) and after Descemet's membrane endothelial keratoplasty (DMEK).

Main Methods:

• A fine-tuned SRGAN was employed to improve the contrast and visibility of corneal layer interfaces in OCT images.
• Adaptive graph theory was utilized for layer segmentation, with search spaces adjusted based on layer contrasts.
• The method was applied to volumetric high-resolution corneal OCT images from healthy individuals and patients with specific corneal conditions.

Main Results:

• The combined SRGAN and adaptive graph theory approach achieved improved segmentation accuracy for five distinct corneal layers.
• Thickness maps generated showed high reproducibility for the whole cornea and stroma (ICC=0.97), and moderate reproducibility for other layers.
• Average thickness errors were within acceptable ranges, with the total cornea error at 3.5 µm.

Conclusions:

• The proposed method demonstrates superior performance compared to conventional graph search techniques for corneal layer segmentation.
• This advanced segmentation and measurement capability holds significant potential for the diagnosis and monitoring of corneal diseases.
• The technique offers precise thickness measurements, which can aid in monitoring DMEK and diagnosing FECD.