The paper presents LiViT-Net, a lightweight Transformer-based network designed for retinal vessel segmentation. The model integrates CNN and Transformer components to capture both local spatial details and long-range dependencies, enhancing edge detection and vessel segmentation accuracy. A novel MobileViT+ block is introduced, which combines convolutions and Transformers to improve the model's sensitivity to vascular edges and capture intricate image structures. A remapped, weighted joint loss function is designed to address pixel imbalances and complex vascular structures, leveraging the strengths of weighted cross-entropy and Dice loss. Extensive experiments on three retinal image datasets (DRIVE, CHASEDB1, and HRF) demonstrate the model's robustness and generalizability, outperforming other methods in complex scenarios. LiViT-Net is also optimized for efficiency, showing fast performance on devices with limited computational power. The paper concludes by discussing future directions, including advanced optimization techniques and real-time data augmentation.The paper presents LiViT-Net, a lightweight Transformer-based network designed for retinal vessel segmentation. The model integrates CNN and Transformer components to capture both local spatial details and long-range dependencies, enhancing edge detection and vessel segmentation accuracy. A novel MobileViT+ block is introduced, which combines convolutions and Transformers to improve the model's sensitivity to vascular edges and capture intricate image structures. A remapped, weighted joint loss function is designed to address pixel imbalances and complex vascular structures, leveraging the strengths of weighted cross-entropy and Dice loss. Extensive experiments on three retinal image datasets (DRIVE, CHASEDB1, and HRF) demonstrate the model's robustness and generalizability, outperforming other methods in complex scenarios. LiViT-Net is also optimized for efficiency, showing fast performance on devices with limited computational power. The paper concludes by discussing future directions, including advanced optimization techniques and real-time data augmentation.