Document resource
Introduction Autonomous navigation of catheters and guidewires in endovascular interventional surgery has the potential to reduce procedure times, enhance intraoperative decision-making, lower radiation exposure, and expand access to treatment.Aim of Study To systematically review the literature and assess the impact, challenges, and opportunities of artificial intelligence (AI) in enabling autonomous navigation of catheters and guidewires during endovascular interventions.Method PubMed and IEEEXplore were searched for studies applying AI to autonomous navigation in endovascular surgery. Inclusion criteria focused on studies involving AI-enabled catheter or guidewire navigation. Articles were reviewed according to PRISMA guidelines and assessed using QUADAS-2. PROSPERO ID: CRD42023392259.Results Out of 462 identified studies, 14 met inclusion criteria. Reinforcement learning (RL) was used in 64% (9/14) and expert demonstration learning in 50% (7/14). Most studies used physical phantoms (10/14, 71%) or in-silico models (4/14, 29%). The majority focused on cardiac vessels (10/14), while three studies used simplified, idealized platforms, and one involved the porcine liver venous system. No studies involved human patients. Common limitations included lack of patient selection criteria, absence of clinical reference standards, and limited reproducibility, contributing to low clinical evidence and generalizability. Risk of bias was present across studies.Abstract A268 Figure 1Conclusion AI applications in autonomous endovascular navigation remain at a proof-of-concept stage, with a technology readiness level of 3. Clear reference standards and robust performance metrics are urgently needed to enable comparison and validation of future data-driven algorithms and support clinical translation.Conflict of Interest No