Overcoming Barriers to Artificial Intelligence Adoption in Interventional Radiology: Current Challenges and Future Prospects

Monday, March 31, 2025

5:05 PM – 5:10 PM CT

Location: Music City Center, 204

Abstract Presenter(s)

Mohammad Ajwad Al Salkhadi, MD

Medical student
Department of Radiology, Jordan University of Science and Technology, Jordan

Author/Co-author(s)

HA

Hassan Al-Balas, MD

Vascular and Interventional Radiology
Michael E. Debakey VA Medical Center, United States
NO

Naser Obeidat, MD

Assistant Professor of Diagnostic Radiology , and a Musculoskeletal and Abdominal Radiologist
Department of Diagnostic Radiology, Jordan University of Science and Technology, Jordan
AA

Asham Al Salkhadi, MD

Physician
Mubarak Al-Kabeer Hospital Department of Medicine, Kuwait

Learning Objectives:

To review barriers to Artificial Intelligence (AI) adoption in interventional radiology (IR), focusing on procedural complexity, data standardization, and expertise gaps.

To discuss recent advancements in AI applications, such as intra-procedural image annotation, device tracking, and therapy planning, and offer solutions for AI integration in IR practice.

Background:

AI has advanced significantly in diagnostic radiology (DR), but its adoption in IR has been slower due to the complexity of real-time procedural guidance, tactile interactions, and dynamic decision-making {1}. However, developments, like AI-assisted catheter navigation and ablation, show promise {2}. We will discuss recent efforts to overcome the barriers to AI in IR and highlight successful examples.

Clinical Findings/Procedure Details:

IR involves complex procedures requiring precise navigation through blood vessels and quick decision-making. AI-enhanced systems, such as robotic catheter navigation tools, are being trialed to assist radiologists in complex procedures {2} {3}, providing real-time feedback that could potentially improve procedural outcomes. Tactile feedback remains crucial, as AI models incorporating tactile sensors are in early development to simulate the feedback that physicians depend on. For effective AI applications we need large sets of consistent data. However, unlike DR where data is usually more standardized, data in IR is often variable and hard to access. As a result, it is more challenging to create effective AI applications in IR {4}. Reporting templates projects like RadReport to standardize IR data and create the datasets required to train and validate AI models. There is a critical need for AI education among interventional radiologists. Programs such as the Clinical Radiology AI course by The Royal College of Radiologists are being developed to bridge knowledge gaps. Ethical concerns, including trust, fairness, and transparency, must be addressed to enable responsible AI implementation in IR, with initiatives like the Transparent-AI Initiative by the American College of Radiology promoting patient safety and trust in AI systems.

Conclusion and/or Teaching Points:

AI is making significant development in IR, with promising developments in intra-procedural guidance and therapy planning. Overcoming challenges like data standardization, tactile feedback, and AI education will be key for broader adoption. Ensuring transparency and ethical standards will result in public trust of AI integration, leading to improved patient outcomes.