By: Doug Teany, Innovation Advisor and Former COO of Corindus
AI and its applications in healthcare have matured considerably over the last several years. AI technologies are more widely available to healthcare providers than ever before and they present tremendous opportunities across healthcare, but especially in highly complex robotic-assisted procedures that demand an exceptional amount of insight and foresight from physicians. Unfortunately, there is still a belief that AI, as a result of its maturation, will inevitably replace the role of the physician, though this couldn’t be further from the truth.
In fact, AI empowers physicians with new knowledge and capabilities that improve clinical decision making and help to achieve more positive patient outcomes. It transfers their attention from thinking technically about maneuvering a wire through the heart to thinking more holistically about case strategy and how to achieve the best possible outcome. When deployed in robotic-assisted procedural settings, AI can help to build an encyclopedia of historical clinical and procedural knowledge that any physician can access to solve any case. Incorporating capabilities such as computer vision will revolutionize the current standard of care, equipping physicians with a powerful tool that can respond to and prevent complications before they arise. All of this is done without replacing the key driver in the process – the physician.
Laying the Foundation with Data-Driven Insight and Machine Learning
Today, physicians primarily rely on personal experience and individual decision-making skills that can vary based on their training, number of procedures they’ve completed, and other factors. At its core, AI introduces physicians to new data and informational insights previously unavailable to them, which can be particularly useful in complex procedural settings utilizing robotics. In these instances where robotic devices are connected to historical data stores, AI offers unparalleled value by filling knowledge gaps with insight from thousands of similar cases. By empowering a physician with access to historical case data beyond his or her own experience, they can mitigate some of the risk associated with these complicated procedures.
Machine learning enables the robotic system to learn from experience and adjust to new inputs. In turn, the robot can automate its movements to mirror the same techniques used by the best physicians in the world, which increases the odds of solving challenges that arise during a case. Ultimately, this allows physicians to focus their attention on case strategy and respond to complications more effectively. This type of data-driven insight resembles how data drastically changed the landscape of professional baseball in the early 2000’s, highlighted by the release of the book Moneyball. The book followed the Oakland Athletics and manager Billy Beane, who took a data-driven, evidence-based approach to building a team, rather than relying on outdated methods that lacked data. In turn, this model was a huge success and effectively spurred a new era of the game driven by data.
In healthcare, a database of procedural knowledge powered by machine learning creates similar insights for physicians, equipping them with valuable information previously unavailable. The ability to view a patient against the background of thousands of similar cases enables physicians to make better, more informed decisions for how they tackle a case.
The Next Step: Computer Vision and High Automation
The full potential of AI in robotic-assisted procedural settings is fully realized when data-driven insight and machine learning are combined with computer vision, enabling the robotic system to intelligently identify and respond to situations based on rich imaging data. In cardiovascular interventions, a significant percentage of procedure time is dedicated to wire manipulation, much of which is a ‘‘trial and error’’ method for navigating through the vessels. This methodology can potentially lead to misplaced medical devices, such as stents, or result in longer procedure times that increase radiation exposure for physicians and patients.
But, as automated systems continue to learn effective movements and create knowledge-based rules, we can allow computer vision to automate larger portions of the procedure. Computer vision also enables the system to detect complications before they occur, such as when a guidewire prolapses, and can automatically perform corrective actions to deliver the device and prevent harm to the patient.
Importantly, automation, regardless of the industry, is on a spectrum, and the automotive industry offers a solid example of this. Many of today’s cars have some degree of low-level automation, such as brake assist, adaptive cruise control and lane centering. On the other end of the spectrum, manufacturers like Tesla automate all safety-critical functions, requiring less action from drivers, though drivers are still necessary. The same is true for automation in healthcare. Ultimately, AI-powered automation keeps physicians in the driver’s seat while automating certain tasks that take physicians’ attention away from critical elements, such as case strategy and responding to complications.
AI Puts Physicians at the Center
Concerns about AI and automation replacing physicians simply don’t hold any merit and aren’t grounded in a full understanding of how healthcare is deploying AI and machine learning. Just as autonomous cars need a driver, the automated interventional procedures we are working toward will need the guidance of a physician, but with the added expertise of a digital encyclopedia built on the experience of other highly skilled physicians from across the globe. Automation will standardize the way procedures are done to a very high level of quality, which gets to the overarching goal of robotic automation – making cases safer, faster, and more effective while reducing trauma on the patient.
- NFL and AWS Close Out AI Safety Challenge
- IBM Acquires Envizi, Looks Toward Sustainability and Environmental Initiatives
- Overinterpretation May Be a Bigger and More Intractable Threat Than Overfitting
- Navrina Singh, CEO and Founder of Credo AI – Interview Series
- BioNTech, InstaDeep Develop Early Warning Detection System for COVID Variants