The spine is a vital structure responsible for supporting and protecting the human body, making it a key focus in recent research to deepen our understanding of spinal diseases. With the rapid advancements in deep learning applied to medical image analysis, significant progress has been made in demonstrating the power of deep convolutional networks for the fast and accurate interpretation of spinal radiographs (X-rays). In medical image analysis, pre-trained image models are also widely applied to deal with various image-based disease/abnormality prediction problems, such as spine lesions. Inspired by recent achievements of previous studies within spinal X-ray image analysis and classification, we proposed a novel fuzzy-enhanced multi-view spinal X-ray image embedding network named FSpineNet in this paper. Unlike the previous approach, our proposed FSpineNet model supports capturing and representing rich information about the patient’s spine conditions through X-ray images in multi-view analysis. We evaluated the effectiveness of our proposed FSpineNet model within a real-world spinal X-ray dataset, which was constructed by collecting the front-view and side-view spine X-ray images of 1,456 people from Nguyen Tri Phuong Hospital, Ho Chi Minh City, Vietnam. The experimental results demonstrated the significance of our proposed FSpineNet model in improving the performance of multi-view X-ray image representation learning and classification for dealing with spinal lesion prediction problems.