Automatic Choroid Layer Segmentation from Optical Coherence Tomography Images Using Deep Learning.

Area of Science:

• Ophthalmology
• Medical Imaging
• Biomedical Engineering

Background:

• The choroid layer's thickness is crucial for diagnosing ophthalmic diseases like diabetic macular edema (DME).
• Accurate choroid layer segmentation in Optical Coherence Tomography (OCT) images is challenging due to image quality issues.
• Current segmentation methods are often manual or semi-automatic, limiting clinical application.

Purpose of the Study:

• To develop and implement an accurate, fully automatic method for choroid layer segmentation in OCT images.
• To segment Bruch's Membrane (BM) and the choroid layer for precise thickness mapping.
• To improve upon existing automated segmentation techniques for clinical utility.

Main Methods:

• A hybrid approach combining deep learning for choroid segmentation and morphological operations for Bruch's Membrane (BM) segmentation.
• Utilized deep learning to leverage extensive image statistics for accurate choroid layer identification.
• Employed morphological operations for precise segmentation of BM boundaries.

Main Results:

• The proposed method demonstrated a significant reduction in segmentation error rates compared to state-of-the-art techniques.
• Achieved high accuracy in segmenting both the choroid layer and Bruch's Membrane.
• Validated through comprehensive evaluation metrics against existing methodologies.

Conclusions:

• The developed automatic segmentation method offers improved accuracy and efficiency for choroid layer analysis in OCT images.
• This technique holds potential for enhanced clinical diagnosis and monitoring of ophthalmic conditions.
• The combination of deep learning and morphological operations provides a robust solution for challenging OCT image segmentation.