MAU-Net: A Retinal Vessels Segmentation Method.

Area of Science:

• Medical Imaging
• Computer Vision
• Biomedical Engineering

Background:

• Accurate retinal vessel segmentation is crucial for diagnosing various eye conditions.
• Existing methods struggle with the complex morphology and deformations of retinal vessels.

Purpose of the Study:

• To develop an advanced deep learning model for precise retinal vessel segmentation.
• To improve the accuracy and robustness of retinal image analysis.

Main Methods:

• Proposed MAU-Net, a U-Net based architecture incorporating Modulated Deformable Convolution (MDC) blocks.
• Integrated dual attention modules (Position Attention Module - PAM and Channel Attention Module - CAM) for enhanced feature representation.
• Validated the method on publicly available DRIVE, STARE, and CHASEDB1 datasets.

Main Results:

• MAU-Net demonstrated superior performance compared to existing algorithms on all tested datasets.
• Achieved high accuracy in segmenting retinal vessels, capturing diverse shapes and deformations.
• Quantitative and qualitative results confirmed the method's effectiveness and precision.

Conclusions:

• MAU-Net offers a robust and accurate solution for retinal vessel segmentation.
• The proposed architecture effectively models complex vessel structures, aiding clinical applications.
• This method advances automated analysis of retinal images for improved diagnostic capabilities.