Abstract Details

Title Artificial Intelligence in Autonomous Navigation for Endovascular Neurointerventions: Current Landscape

Topic Cerebrovascular Disease and Interventional Neurology

Presentation(s) P6 - Poster Session 6 (5:00 PM-6:00 PM)

Poster/Presentation
Number 4-004

Objective

To describe current advances in the application of artificial intelligence (AI) for autonomous navigation during endovascular neurointerventions.

Background Endovascular neurointerventions are minimally invasive procedures performed within the cerebral vasculature to diagnose and treat conditions such as aneurysms, arteriovenous malformations, and ischemic stroke. They rely on real-time imaging and highly skilled manual catheter manipulation through complex vascular anatomy. Artificial intelligence (AI) refers to computational systems capable of learning from data and performing tasks that traditionally require human expertise. In recent years, AI and robotics have begun to converge within interventional neuroradiology, enabling automated perception, navigation, and procedural assistance. This integration seeks to enhance precision, efficiency, and safety while reducing operator fatigue and procedural variability.

Design/Methods A narrative review was conducted following a structured literature mapping approach informed by PRISMA principles. Publications indexed in PubMed, IEEE Xplore, Scopus, and arXiv were examined to identify trends in algorithm design, model architectures, validation environments, and translational readiness. Studies were thematically grouped into perception and mapping, path-planning and control, and robotic or teleoperated integration.

Results Reinforcement and imitation learning dominate algorithmic strategies, complemented by deep convolutional and graph neural networks for vascular segmentation, path estimation, and device localization. Most investigations remain confined to simulation or benchtop settings, with limited ex vivo validation and no human autonomous procedures to date. Persistent challenges include sim-to-real transfer, patient-specific adaptation, and safety validation. However, AI-assisted teleoperated systems demonstrate emerging translational feasibility.

Conclusions AI-driven autonomous navigation in neurointerventions is progressing from conceptual development to preclinical validation. Continued interdisciplinary collaboration and standardized evaluation frameworks will be critical to enable safe, reproducible, and clinically integrated deployment of autonomous neurointerventional systems.

Authors/Disclosures
Valentina Velasco PRESENTER	Miss Velasco has nothing to disclose.
Andres Ricaurte, MD (Weill Medical College of Cornell University)	Dr. Ricaurte has nothing to disclose.
Diana M. Riaño, MD	Dr. Riaño has nothing to disclose.
Andres Felipe Cardenas Cruz, Jr., MD (Javeriana)	Dr. Cardenas Cruz has nothing to disclose.
Kiara Torres Tello, Student	Miss Torres Tello has nothing to disclose.
Juliana Coral, MD	Dr. Coral has nothing to disclose.

��ɫ��

��ɫ��