Visualization of Surgical Needle Tips Hidden Inside Organs Using Generative Adversarial Networks.

Postoperative complications in surgery are particularly prevalent in laparoscopic procedures, which are difficult for physicians to perform. One complication that is related to the surgical procedure itself is anastomotic leakage, in which the sutures do not fully attach and the contents leak out. A factor that contributes to anastomotic leakage is uneven needle suture spacing. Once the needle is inserted, the physician relies on intuition and experience to carry it through to the needle exit, which makes it difficult to maintain uniform suture spacing. For this reason, to reduce the risk of anastomotic leakage, it is effective to visualize the position of the needle tip in the organ. Therefore, in this study, we constructed a model to visualize the tip of a suture needle hidden inside an organ to improve surgical accuracy. Using the model, we tested the inference of images and achieved a real¬time speed of 33.4 fps, and the average estimated misalignment of the needle tip was 1.03 mm, which is less than 1.8 mm (10% of the total needle length). Thus, we showed that the proposed model effectively estimated needle tips hidden in organs. Therefore, we expect that this model will lead to improved surgical accuracy.