An effective vessel segmentation method using SLOA-HGC.

Area of Science:

• Medical Imaging
• Computer Vision
• Ophthalmology

Background:

• Accurate segmentation of retinal blood vessels is vital for diagnosing various ophthalmic diseases.
• Existing methods face challenges in microfine vessel extraction and edge clarity.

Purpose of the Study:

• To develop and validate a novel retinal blood vessel segmentation algorithm.
• To enhance the accuracy and robustness of retinal vascular feature extraction.

Main Methods:

• Utilized Channel-aware self-attention (CAS) for improved microfine vessel segmentation sensitivity.
• Employed Heterogeneous adaptive pooling (HAP) for multi-scale feature extraction and accurate edge segmentation.
• Incorporated a ghost fully convolutional Rectified Linear Unit (GFCReLU) module for deep semantic information capture.
• Optimized neural network training using the Sparrow-Integrated Lion Optimization Algorithm (SLOA).

Main Results:

• Achieved Mean Intersection over Union (MIoU) scores ranging from 74.11% to 80.61% across multiple datasets.
• Obtained Dice coefficients between 68.93% and 78.97%.
• Demonstrated high accuracies, with results up to 96.67% on various datasets.
• The algorithm effectively segments retinal blood vessels, including microfine structures.

Conclusions:

• The proposed algorithm demonstrates superior performance in retinal blood vessel segmentation.
• This advancement holds significant potential for the early detection and diagnosis of ophthalmic diseases.
• The developed model offers a robust tool for analyzing retinal vascular patterns.