Abstract
INTRODUCTION: The retinal microvasculature has been definitively linked to a variety of diseases, such as ophthalmological, cardiovascular, and other medical conditions. Precisely identifying the retinal microvasculature is crucial for early detection and monitoring of these diseases. While the majority of existing neural network-based research has primarily focused on utilizing the green channel of fundus images for vessel segmentation, it is important to acknowledge the potential value of other channels in this process. METHODS: This study introduces RetinalVasNet, a new method aimed at enhancing the accuracy and effectiveness of retinal vascular segmentation by implementing a sophisticated neural network architecture and incorporating multi-channel fundus images. RESULTS: Our experimental results demonstrate that RetinalVasNet outperforms previous research in most performance metrics. DISCUSSION: The findings suggest that each channel provides unique contributions to the vascular segmentation process, emphasizing the importance of incorporating multiple channels for accurate and comprehensive segmentation.