The end detection of guide wire in interventional surgery plays a vital role in ensuring the accurate control and safety of surgery. In this article, a method of end detection of guide wire based on the improved YOLOv4Tiny network is proposed for the clinical demand of end detection of guide wire during surgery. The YOLOv4Tiny network architecture is utilized in the proposed method. By optimizing the residual structure in the feature extraction network, and enhancing the attention mechanism and the hybrid expansion convolution network, the small target feature extraction ability and detection accuracy are greatly improved. The receptive field is also expanded, which ensures the image resolution without increasing the computation amount. To evaluate the effectiveness of the improved algorithm, it is tested in the constructed dataset and the actual surgical dataset. According to the experimental results, the average accuracy of the improved algorithm in the constructed dataset reaches 97. 6% , with the detection error of the guide wire end be less than 5% , and the average accuracy in the actual surgical dataset is 92. 8% . The improved algorithm is of great reference significance for the end detection of interventional surgical guide wire, which has broad application prospects in fields related to biomedical robots.