Retinal Blood Vessel Segmentation using The U-NetModel with the XAI Technique

Segmenting retinal blood vessels is important because the retinal microvasculature is a vital biomarker for severalclinical disorders. On the other hand, the conventional approach is thought to be a laborious and time-consuming process. The potential of deep learning and artificial intelligence in particular has created new opportunities for effectively segmenting retinal arteries. In our research, we provide a strategy to extract retinal blood vessels from fundus visuals using the U-Net model, whichis commonly used in semantic segmentation. A dataset for the study is obtained from the Kaggle resource. We performed pre-processing before feeding the images into the models based onthe need for deep-learning models. Moreover, we applied a U-Netmodel to segment the stereophotogrammetric converted images. Concerning the quality of the approach, the Intersection of Union (IoU), accuracy, and F1-Score measurements were employed. Furthermore, to enhance the grain of the segmentation results and to build confidence in the method we discuss the incorporation of eXplainable AI known as Grad-CAM. Therefore, this work shows that deep learning can be highly effective for retinal vessel segmentation and at the same time underlines the need for interpretability issues in medical image analysis.