Dr. James Tudor, MD, spearheads the integration of AI into XCath’s robotics systems. Driven by a passion for the convergence of technology and medicine, he enthusiastically balances his roles as a practicing radiologist, Assistant Professor of Radiology at Baylor College of Medicine, and AI researcher.
Founded in 2017, XCath is a startup focused on advancements in medical robotics, nanorobotics, and materials science. The company develops next-generation endovascular robotic systems and steerable guidewires aimed at treating cerebrovascular disorders and other serious medical conditions.
Dr. Tudor, what initially sparked your interest in the intersection of AI and medicine, particularly in the field of radiology?
In 2016, as I was beginning my radiology residency, DeepMind’s AlphaGo defeated world champion Go player Lee Sedol. AlphaGo’s ability to compress and abstract the vast complexities of Go, a game with more possible board positions than atoms in the observable universe, captured my imagination. Excited about AI’s potential to transform radiology and medicine as a whole, I dove headfirst into AI. During residency, I’d spend my evenings and weekends doing AI projects.
Can you tell us more about your journey from medical school to becoming the VP of AI at XCath? What motivated you to pursue AI integration within healthcare robotics?
My career path has taken some unexpected turns. After finishing my radiology residency, I wanted to dedicate more time to AI and its commercial applications. I joined a fitness robotics startup, founded by Eduardo Fonseca, who is now XCath’s CEO. It was a formative experience, but I never anticipated it would lead down the path of treating acute stroke with endovascular telerobots.
Around a decade ago, a revolution occurred in acute stroke care. The standard of care used to be a medication called tPA that would break up the clot. In 2015, clinical trials demonstrated the superiority of directly removing the clot from the cerebral arteries by navigating tiny guidewires and catheters within the arterial vasculature, a procedure called mechanical thrombectomy. Despite the procedure being markedly effective for large vessel strokes, less than 40% of the US population has access to it. There are a limited number of stroke centers, generally limited to urban areas, that have specialists who can perform the procedure. Globally, the statistics are even more dismal: less than 3% of the world has access.
XCath’s mission is to increase access to mechanical thrombectomy with a hub-and-spoke model, where specialists can provide expert stroke care from a distance with endovascular telerobots deployed to regions without access.
Eduardo asked me how AI could augment the safety of the telerobotic system. I was so curious I spent a few weeks deep in research, having conversations with interventionalists and learning about the telerobot. The mission and potential humanitarian impact are so compelling I had to answer that call to arms.
How did your experiences as an academic radiologist shape your approach to integrating AI in medical devices?
Teaching radiology residents has sharpened my ability to explain complex ideas clearly, which is key when bridging the gap between AI technology and its real-world use in healthcare. It also keeps me grounded in the challenges clinicians face, which helps me design AI solutions that are clinically practical and user-friendly.
As the VP of AI at XCath, what are some of the key challenges you faced while integrating AI into XCath’s robotic systems? How did you overcome them?
Integrating AI into surgical robotics presents a U-shaped challenge. The greatest difficulties lie at the beginning—acquiring and managing data—and at the end—integrating it into an embedded software package. In comparison, the actual training of the AI models is relatively straightforward.
Acquiring medical data is challenging, but fortunately, we were able to establish excellent image-sharing partnerships. Implementing the models for clinical use requires orchestrating the efforts of various teams, including AI, Quality, Software, UI/UX, and Robotic engineers, all while constantly validating with the clinical team that the solution is useful and effective. With so many moving parts, success ultimately depends on having dedicated, high-performing teams that communicate frequently and effectively.
Could you elaborate on how AI enhances the capabilities of XCath’s endovascular robotic systems? What role does AI play in improving patient outcomes?
AI algorithms can serve as a constant teacher and assistant, decreasing the cognitive load and leveling up all providers to provide world-class care. AI can provide intraoperative and postoperative feedback, accelerating the training and adoption process of endovascular robotics. We aim to make the system so effective and accessible that other intravascular specialists such as interventional body radiologists and interventional cardiologists can be trained to provide acute stroke care with the robot.
Additionally, locally embedded algorithms can provide an extra level of safety from cyber-attacks and network failures as they anticipate the expected path of a procedure and can alert and pause the procedure in the case of the unexpected.
At the end of the day, we do not want to take control from the interventionalist, but augment their abilities so that every patient can be confident they are getting world-class care.
How does XCath’s AI-driven technology address the complexities of navigating the human vasculature during endovascular procedures?
XCath’s Endovascular Robotic System represents a major advancement in precision medicine, designed to navigate intricate human vasculature with sub-millimeter accuracy. Our system is designed to minimize procedural variability and enhances control over various endovascular devices through an intuitive control console.
Additionally, XCath’s ElectroSteer Deflectable Guidewire System, the world’s first electronically-controlled steerable smart guidewire, features a steerable tip engineered to navigate complex vascular anatomies and challenging vessel angulations.
AI will further enhance navigation capabilities with locally embedded computer vision and path planning models. These models play a crucial role in reducing the cognitive load on interventionalists during procedures by assisting with real-time image analysis and enhancements and providing safeguards through parallel autonomy.
XCath recently achieved a significant milestone with the world’s first telerobotic mechanical thrombectomy demonstration. Could you share your insights on the role AI played in this groundbreaking procedure?
We used an earlier version of the robot for that groundbreaking achievement, so AI did not play a role. However, it’s an incredible milestone that lays the foundation for future integration of AI into telerobotic procedures.
In this live demonstration, Dr. Vitor Pereira performed an MT procedure from Abu Dhabi on a simulated patient in South Korea, removing a blood clot in the brain in minutes. We were thrilled by the results of the telerobotic demonstration, which found low latency and reliable connection between the robotic controller located in Abu Dhabi and the robotic device in South Korea. We project regional robotic telestroke networks, but we went to an extreme to demonstrate the capabilities of the technology.
What do you believe is the future of telerobotic surgery in the treatment of acute neurovascular conditions, and how is XCath preparing to lead in this space?
Justifying the necessity of telerobotic surgery in many medical scenarios can be challenging, especially when a surgeon is readily available or patient transfer is feasible. However, in the context of stroke treatment, where every minute counts and neurons are rapidly lost, telerobotic interventions become crucial.
XCath is uniquely positioned to pioneer telerobotic surgery, initially focusing on stroke treatment. Our approach addresses the critical need for rapid intervention in areas with limited access to specialized care. Once we’ve successfully tackled this challenge, I believe it will pave the way for telerobotic solutions in other time-sensitive medical emergencies. Also, given the extreme precision of the robotic controls, there is potential for using the robot locally to perform technically difficult surgeries, such as aneurysm repairs.
Where do you see the future of AI in healthcare heading, particularly in relation to robotic systems and minimally invasive procedures?
AI has immense potential to revolutionize healthcare. The initial wave of AI applications has primarily focused on triage and efficiency improvements. We’ve seen significant advancements in radiology, particularly in flagging urgent cases or automating acquisition of measurements. I’m also excited about automated medical record documentation. A current challenge is that doctors often spend more time documenting in front of computers than interacting with patients. I anticipate the development of systems that can document patient interactions or surgeries in real-time, freeing up valuable physician time. In the realm of robotics, AI will play a crucial role in assisting and proctoring, thereby enhancing the consistency and quality of care.
In the foreseeable future, AI is going to augment, but not replace surgeons. The implementation of parallel autonomy in robotic systems will significantly improve both the safety and efficiency of procedures.
As someone deeply involved in AI research, what advancements in AI do you think will have the most significant impact on medical device development over the next decade?
In the last few years, we’ve witnessed a wave of supervised deep learning models receiving FDA approval and are just now starting to fulfill their promise of transforming healthcare. A wave of generative AI applications will likely dominate the next few years. Agentic AI, by comparison, is in its infancy, but holds much greater promise. As AI is rapidly evolving, it’s very likely we will see multi-agent systems that can diagnose and treat in real time. There will be additional regulatory hurdles for these agents whose actions are both opaque and probabilistic. However, global need will drive the demand for adoption. In Rwanda, the company Zipline is using flying drones to deliver vital medical supplies within minutes around the country. Similarly, in places that lack access to medical resources, the risk/benefit equation is very different and would likely prompt them to leapfrog the developed world in the deployment of multi-agentic AI medical devices.
Thank you for the great interview, readers who wish to learn more should visit XCath.
