HybridVesselNet: A Hybrid Approach for Retinal Vessel Segmentation and Biomarker Extraction for Alzheimer's Disease

Alzheimer's disease (AD) is caused by cognitive decline and cerebral microvascular dysfunction, which is often reflected in alterations of retinal vascular structure. This study proposes HybridVesselNet, a scratch trained hybrid deep learning framework developed to achieve precise retinal vessel segmentation and extract morphological biomarkers that may aid early AD detection. By combining residual blocks with Efficient Channel Attention (ECA), a Swin Transformer-based self-attention bottleneck, and a Parallel Atrous Pyramid (PAP) module, the model effectively captures both local and global vessel structures. A tailored preprocessing pipeline incorporating Contrast limited adaptive histogram equalization (CLAHE), gamma correction, and brightness-contrast enhancementwas applied to improve image quality and vessel visibility. The model was trained on the DRIVE dataset and evaluated on CHASE DB1, attaining satisfactory…