Pathological-Corneas Layer Segmentation and Thickness Measurement in OCT Images

Area of Science:

• Ophthalmology and biomedical imaging.
• Medical image analysis and computational pathology.

Background:

• Accurate corneal thickness measurement is crucial for diagnosing and managing pathological conditions.
• Existing segmentation methods struggle with the irregular layers often found in diseased corneas.

Purpose of the Study:

• To introduce a new algorithm for precise segmentation and thickness measurement of pathological corneas with irregular layers.
• To leverage a two-stage graph search and ray tracing approach for improved accuracy.

Main Methods:

• A two-stage graph search algorithm was developed for segmenting corneal layers from optical coherence tomography (OCT) images.
• The first stage segmented air-epithelium and endothelium-aqueous boundaries, while the second stage refined segmentation using gradient, directional, and multiplier costs.
• Ray tracing was employed to correct inter-boundary distances for accurate thickness measurements.

Main Results:

• The algorithm achieved segmentation accuracy comparable to manual operators, with mean unsigned errors of 0.89 ± 1.03 pixels.
• Thickness measurements demonstrated high precision, with mean unsigned errors of 3.62 ± 3.98 µm.
• Performance was validated on 190 manually segmented OCT images.

Conclusions:

• The proposed algorithm provides accurate segmentation and thickness measurements for pathological corneas.
• The method demonstrates potential for clinical utility in ophthalmology practices.