Thickness profiles of retinal layers by optical coherence tomography image segmentation

Area of Science:

• Ophthalmology and Vision Science
• Medical Imaging Analysis
• Biomedical Engineering

Background:

• Accurate measurement of retinal layer thickness is crucial for diagnosing and monitoring various eye diseases.
• Existing methods for retinal layer thickness measurement using optical coherence tomography (OCT) can be time-consuming and subjective.
• Development of automated, quantitative methods is needed to improve efficiency and reproducibility in retinal imaging analysis.

Purpose of the Study:

• To develop and report an image segmentation algorithm for quantitative thickness measurement of six specific retinal layers.
• To assess the accuracy and reliability of the automated algorithm by comparing its measurements to manual expert delineations.
• To generate normative thickness profiles for individual retinal layers in healthy subjects using OCT.

Main Methods:

• A prospective, cross-sectional study involving 15 normal subjects (time-domain OCT) and 10 normal subjects (spectral-domain OCT).
• A novel 2D edge detection algorithm was developed to enhance retinal layer boundaries and reduce speckle noise in OCT images.
• Automated thickness measurements were compared against manual measurements from three independent observers to evaluate performance.

Main Results:

• The algorithm successfully identified seven boundaries, enabling thickness measurement of six distinct retinal layers.
• Mean absolute differences between automated and manual thickness measurements ranged from 3-4 µm, comparable to interobserver variability.
• Generated thickness profiles revealed characteristic patterns for inner retinal layers, outer plexiform layer, outer nuclear layer and photoreceptor inner segments, and photoreceptor outer segments, with specific variations at the fovea.

Conclusions:

• The developed image segmentation algorithm provides accurate and reproducible quantitative thickness measurements of multiple retinal layers from OCT images.
• This automated technique shows significant potential for clinical applications, including the assessment of disease progression and treatment efficacy.
• Further application of this segmentation method can aid in understanding subtle changes in retinal layer thickness related to ocular pathologies.