Artificial intelligence-driven real-time assistance in minimally invasive surgery: a technology-oriented narrative review

Although minimally invasive surgery is a standard surgical approach because of its proven benefits to patient outcomes, it imposes increased cognitive demands on surgeons because of limited visualization, lack of tactile feedback, and constraints in instrument manipulation. These challenges have been partially mitigated by robot-assisted surgery, while recent advances in artificial intelligence (AI) have established intraoperative AI as a key technology enabling real-time support of surgical perception, decision-making, and instrument control. Here, we review the key AI technologies used during the intraoperative phase of endo-laparoscopic and robotic surgery, including anatomical and lesion recognition, instrument detection and tracking, surgical phase and workflow analysis, real-time tissue characterization, image-guided and augmented navigation, AI-assisted instrument control, and multimodal event detection. We also discuss key clinical integration challenges and future research directions focused on foundation and self-supervised learning paradigms, and human-AI collaborative system design.