Deep Learning with Skip Connection Attention for Choroid Layer Segmentation in OCT Images.

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Accurate choroid layer segmentation is crucial for diagnosing ophthalmic diseases.
• Segmentation faces challenges due to ambiguous boundaries and lack of context.
• Inconsistent predictions arise from limited context or large segmented regions.

Purpose of the Study:

• To propose a novel Skip Connection Attention (SCA) module for improved choroid layer segmentation in OCT images.
• To address challenges of ambiguous boundaries and insufficient context information in segmentation tasks.
• To enhance the precision and consistency of choroid layer segmentation.

Main Methods:

• Integration of a novel Skip Connection Attention (SCA) module into U-Shape architectures like U-Net and CE-Net.
• The SCA module captures global context at the highest level to guide the decoder.
• Stage-by-stage guidance from the SCA module aims to extract more context and ensure consistent predictions.

Main Results:

• The proposed SCA module significantly improves the accuracy of choroid layer segmentation.
• Integration into U-Net and CE-Net architectures demonstrated enhanced segmentation performance.
• The module effectively addresses issues related to ambiguous boundaries and context deficiency.

Conclusions:

• The Skip Connection Attention (SCA) module is an effective enhancement for choroid layer segmentation in OCT images.
• The SCA module's ability to capture global context improves segmentation precision and consistency.
• This approach offers a promising solution for accurate ophthalmic disease diagnosis through improved OCT image analysis.