Automatic Choroidal Layer Segmentation Using Markov Random Field and Level Set Method.

Area of Science:

• Ophthalmology
• Medical Imaging
• Biomedical Engineering

Background:

• The choroid, a vascular layer, provides essential oxygen and nutrients to the retina.
• Alterations in choroidal thickness are implicated in the pathophysiology of various retinal diseases.
• Accurate segmentation of the choroid is crucial for understanding these disease processes.

Purpose of the Study:

• To develop and validate an automatic method for segmenting the choroidal layer from macular optical coherence tomography (OCT) images.
• To improve the accuracy and robustness of choroidal layer segmentation.

Main Methods:

• Utilized a level set framework for choroidal layer segmentation.
• Employed a 3D nonlinear anisotropic diffusion filter to reduce OCT imaging artifacts (speckle noise) and enhance contrast.
• Integrated distance regularization and edge constraint terms to refine segmentation boundaries and prevent irregular regions.
• Incorporated a Markov random field model to address regional texture inhomogeneity and prevent segmentation leakage caused by vascular shadows.

Main Results:

• The proposed method successfully segmented the choroidal layer in macular OCT images.
• Demonstrated high accuracy in estimating the posterior choroidal boundary.
• Outperformed other segmentation methods when compared against manually labeled ground truth.

Conclusions:

• The developed automatic level set-based method provides an effective and accurate approach for choroidal layer segmentation.
• This technique holds potential for clinical applications in diagnosing and monitoring retinal diseases associated with choroidal changes.