A Multicenter Prospective Observational Study of Computer-aided Risk Perception and Prognosis Prediction in the Whole Process of Laparoscopic Hepatobiliary and Pancreatic Surgery

Artificial intelligence technology is used to realize high-quality 3D scene reconstruction, whole process segmentation, scene activity understanding for common surgery guidance in hepatobiliary surgery, as well as intelligent identification, perception, early warning of key events in the whole process of endoscopic surgery (such as bleeding, blocking, tumor location, anastomosis, etc.), and decision-making assistance

Endoscopic surgery is the most important and commonly used minimally invasive surgery technology in the field of modern surgery, especially in hepatobiliary surgery, which has become one of the conventional diagnosis and treatment methods of surgery. Compared with traditional open surgery, it has smaller trauma, faster recovery time and lower complication rate. However, the limited visual field of surgical observation caused by the narrow surgical space and the difficulty of immediate identification of key events in the surgical scene greatly increase the difficulty and complexity of endoscopic surgery. Its safety and efficacy largely depend on the precise perception of the complex surgical field and the ability to handle key events during the operation. Therefore, combining modern image processing technology and machine learning algorithm, it is particularly urgent to develop a system that can provide real-time dynamic perception and safety warning of endoscopic surgery. Although domestic and foreign scholars have carried out a lot of research on endoscopic video dynamic perception and safety warning, the current research only focuses on local problems in the surgical process. Traditional image processing technology is often difficult to meet the needs of highly sensitive to real-time dynamic information of surgery, it is difficult to achieve efficient three-dimensional reconstruction of the surgical scene, it can not provide the key anatomical structure information of human organs, and it can not accurately detect the key events in the operation process. For endoscopic surgery, further research is urgently needed to realize the video dynamic perception and safety early warning system of endoscopic surgery, and assist doctors to achieve safe, accurate and efficient endoscopic surgery. In recent years, with the continuous progress of computer graphics and image technology and machine learning methods, the dynamic perception and safety early warning system of endoscopic surgery video will develop towards higher automation and intelligence. Future research may focus on improving the real-time and accuracy of the algorithm, as well as how to better integrate artificial intelligence technology into the clinical operation process, realize the real-time perception and safety warning of endoscopic surgery, and improve the efficiency and safety of surgery through the comprehensive analysis and understanding of endoscopic surgery process. Due to the complexity and variability of the endoscopic surgery environment, it is difficult to identify the key anatomical structures of organs during the operation, and it is very dependent on the subjective empirical judgment of the surgeon. There is a lack of objective instructions. It is particularly important to develop a machine learning method that can detect, perceive and recognize the key anatomical structures in real time during the operation. At the same time, the workflow of endoscopic surgery is fine and complex, so it is very necessary to comprehensively analyze and detect the key events and activity scenes in the video of endoscopic surgery through the AI auxiliary system. In addition, the realization of intraoperative hidden target area augmented reality surgery navigation needs to be carried out on the accurate dynamic organ surface reconstruction and non rigid registration results. However, the complex and narrow field of view endoscopic video further reduces the accuracy of non rigid registration, making augmented reality assisted endoscopic surgery extremely challenging. In conclusion, how to solve the problem of real-time dynamic perception and safety warning of endoscopic video is the key to achieve safe, accurate and efficient endoscopic surgery in clinic. Through the research and application of endoscopic video real-time dynamic perception and safety early warning technology, it can realize real-time dynamic perception, key event early warning, prediction of the location of invisible lesions and other decision-making information in various high-risk and difficult endoscopic surgery processes, and assist doctors to "see", "see clearly", "see accurately" in the operation process, so as to further improve the efficiency and safety of endoscopic surgery. At the same time, based on the above content, the success of the treatment of some key fields in surgery will greatly affect the prognosis of patients and the quality of life of patients. Another purpose of this study is to more comprehensively and objectively understand the incidence, risk factors, prevention and treatment of intraoperative and postoperative complications after evaluation combined with surgical video analysis, so as to provide clinicians with a more scientific treatment plan and guidance.

• Hepatocellular Carcinomas
• Cholecystectomy, Laparoscopic
• Hepatectomy
• Pancreatoduodenectomy

• NFEC-2024-403