Artificial Intelligence and Digital Tools in HPB Surgery: From Imaging to Intraoperative Guidance – A PRISMA-Guided Systematic Review

Abstract: Hepato-pancreato-biliary (HPB) surgery involves complex liver, pancreatic, and biliary procedures where outcomes depend on precise imaging, careful planning, and intraoperative decisionmaking. Artificial intelligence (AI) and digital tools have rapidly emerged to assist in these domains. We conducted a PRISMA 2020–compliant systematic review of current applications of AI in HPB surgery – spanning preoperative imaging analysis to intraoperative guidance – and evaluated their impact on surgical planning, execution, and outcomes. Comprehensive searches of PubMed, Embase, Web of Science, and Scopus (through June 2025) identified studies on AI or digital technologies in HPB surgical care. Inclusion criteria encompassed any AI-driven tool (machine learning, deep learning, augmented reality (AR), etc.) applied to imaging interpretation, operative planning, intraoperative guidance, or outcome prediction in HPB surgery. Two reviewers independently screened and selected studies; 38 studies met inclusion criteria. AI models in preoperative imaging (radiology and ultrasound) achieved high accuracy in detecting and characterizing HPB tumors – for example, differentiating malignant from benign lesions with area under ROC curves (AUCs) of ~0.80–0.98. Several machine learning algorithms outperformed traditional risk scores in predicting postoperative complications such as pancreatic fistula (with AUCs ~0.80 vs < 0.60 for conventional scores). AI assisted 3D reconstruction tools improved surgical planning by enhancing visualization of tumor location and anatomy, while AR systems enabled real-time intraoperative navigation of liver and pancreatic resections in pilot studies. Computer vision techniques have been applied to laparoscopic videos to identify critical anatomical landmarks – notably, an AI system recognized the “critical view of safety” in laparoscopic cholecystectomy with ~83% accuracy, suggesting potential to reduce bile duct injury. Early implementations of combined AI and AR have shown feasibility of real-time hazard detection (e.g. highlighting surgical bleeding) displayed through smart glasses to alert surgeons. Conclusions: AI and digital tools are increasingly permeating HPB surgery, demonstrating promise from improved preoperative diagnostics to enhanced intraoperative guidance. Current evidence is largely observational and proof-of-concept, but consistently shows that AI can augment surgeons’ decision-making (e.g. by more accurately stratifying tumors and risks, guiding resection planes, or monitoring for intraoperative events). To translate these advances into routine care, prospective clinical studies are needed to validate efficacy and safety, and surgical teams must integrate AI-driven workflows into practice. With continued refinement and multidisciplinary collaboration, AI has the potential to improve the precision and outcomes of HPB surgery across the entire treatment continuum.