Multiple layer segmentation and analysis in three-dimensional spectral-domain optical coherence tomography volume...

Area of Science:

• Ophthalmology
• Medical Imaging
• Computational Biology

Background:

• Spectral-domain optical coherence tomography (SD-OCT) is crucial for visualizing retinal structure.
• Understanding retinal layer morphology is key to diagnosing various eye diseases.
• Accurate segmentation of retinal layers is essential for quantitative analysis.

Purpose of the Study:

• To develop and validate an automated algorithm for segmenting 11 retinal surfaces in SD-OCT volumes.
• To assess the accuracy of the automated segmentation compared to manual segmentation.
• To establish a baseline for analyzing retinal layer properties in normal subjects.

Main Methods:

• A three-stage automated graph search algorithm was developed for retinal surface segmentation.
• The algorithm utilized downsampled images and morphological shape models for refinement.
• Segmentation accuracy was evaluated using 20 macular SD-OCT scans from normal subjects.

Main Results:

• The automated algorithm achieved high accuracy in segmenting 11 retinal surfaces, with a mean absolute difference of 3.19 ± 2.46 μm.
• The study quantified intensity/reflectivity and thickness properties of retinal layers.
• Results provide a quantitative reference for normal retinal morphology.

Conclusions:

• The developed automated graph search algorithm offers precise and reproducible retinal surface segmentation in SD-OCT images.
• This method serves as a valuable tool for quantitative analysis of retinal morphology.
• The findings establish a normative dataset for future studies on ocular diseases.