Automatic Choroidal Segmentation in Optical Coherence Tomography Images Based on Curvelet Transform and Graph Theory

Area of Science:

• Ophthalmology
• Medical Imaging
• Biomedical Engineering

Background:

• Automatic segmentation of the choroid in optical coherence tomography (OCT) images aids in diagnosing eye pathologies, offering speed and consistency over manual methods.
• Speckle noise in OCT images presents a significant challenge to accurate automatic segmentation and interpretation.
• A novel curvelet transform-based K-SVD method is proposed to address speckle noise in OCT imaging.

Purpose of the Study:

• To develop and evaluate a new method for automatic choroidal segmentation in OCT images.
• To reduce speckle noise in OCT images using a curvelet transform-based K-SVD approach.
• To compare the performance of the proposed automatic segmentation technique against manual segmentation.

Main Methods:

• Speckle noise was reduced using curvelet transform-based K-SVD dictionary learning and the Lucy-Richardson algorithm.
• Outer/Inner Choroidal Boundaries (O/ICB) were identified using graph theory.
• The choroidal region was defined as the area between the inner and outer choroidal boundaries.

Main Results:

• The proposed method achieved an average Dice Similarity Coefficient (DSC) of 92.14% ± 3.30% when compared to manual segmentation.
• In contrast, the latest open-source algorithm by Mazzaferri et al. achieved a mean DSC of 55.75% ± 14.54% on the same dataset.
• A high degree of similarity was observed between the automatic and manual segmentations.

Conclusions:

• The proposed automatic segmentation method demonstrates significant similarity to manual segmentation.
• This technique can be valuable for large-scale quantitative studies of the choroid.
• The developed method shows promise for improving the interpretation of OCT images in clinical practice.