The demand for qualified medical practitioners and surgeons is expected to increase twofold within the next decade, as professionals are retiring faster than new ones enter the field, and the population continues to age. These factors place increasing stress on doctors and patients alike and have prompted a hunt for solutions.

New research co-authored by a University of Utah computer scientist explores how artificial intelligence (AI) can help alleviate this crunch by using it to deploy medical needles within the body.

For decades, surgeons have been guided by imagery produced by computed tomography, ultrasound and magnetic resonance as they operate, but rapid advances in AI are now opening new avenues for high-precision procedures where there is little room for error, according to Alan Kuntz, who helped write a comprehensive review of prior research, published this month in Science Robotics.

“AI in this context is not this monolithic intelligent agent that’s performing all of these tasks. It’s not an artificial surgeon. We look at targeted locations in which these kinds of technologies can be inserted into the surgical or interventional workflow in order to improve a piece of capability there,” said Kuntz, an assistant professor in the Kahlert School of Computing and a member of the U’s Robotics Center. “Where are the places in a typical needle-based intervention where these tools are appropriate? How have people used these tools in those places? And then what can we learn? What are the next steps?”

Kuntz’s lab is developing a suite of next-generation autonomous tools, designed to enhance the capabilities of surgeons—without replacing their role in the operating room. Such advances could ease burnout rates among medical professionals and reduce work-related stress. In particular cases, autonomous tools can outperform their human counterparts.

“What we’re trying to do is reduce the per-patient workload of surgeons and interventionalists,” Kuntz said. “That’s one advantage of AI in this space: expanding access in a world where we’re losing access.”

He and his co-authors, who include Ron Alterovitz of the University of North Carolina and Janine Hoelscher of Clemson University, established a four-part AI-guidance framework for deploying needles, among the most widely used medical instruments, needed for a vast array of procedures from targeted drug delivery to extracting biopsies.

“Breakthroughs in AI and robotics will continue to enable increasing degrees of AI guidance and robotic automation for medical procedures,” said Alterovitz in a UNC release. “AI and robotics can provide physicians with new tools to make challenging procedures safer and more effective.”

The framework’s components are perceiving anatomy, planning motions, perceiving instrument state and performing motions—each deployable across the autonomy spectrum, from basic support to full autonomy.

“Perceiving anatomy from an AI perspective is really about taking medical imaging or camera imaging and determining automatically what is the structure of this is geometrically and what is the structure of this is from an anatomy semantics perspective,” Kuntz said.

This framework pinpoints where AI-integrated tools add value to needle deployment and where human supervision remains essential. Applications span image segmentation, CT-based target and obstacle identification, manually drawn insertion paths and sampling-based planners for steerable needles following curvature-constrained, collision-free trajectories. The study identified performance benefits and potential risks, allowing engineers to build “guardrails” around each workflow.

“The first key point is that AI in needle-based interventions is much more nuanced than ‘give a robot a needle and let it do whatever,’” Kuntz said. “The second is that this nuanced approach lets us create intermediate steps where we have AI technology in pieces of it, but the whole system isn’t relying on that AI technology without any other boundaries.”

In his Utah lab, Kuntz and his collaborators explore ways to reduce the invasiveness of surgical tools while adhering to safety standards. Their work has largely focused on surgical needles but has expanded to other applications, including laparoscopic surgery.

These new tools utilize AI to enhance image processing and guidance, resulting in surgical instruments with greater autonomy, precision and efficacy. They are integrated into robotic platforms that can, in some cases, operate independently or be controlled by a surgeon through a console—always guided by an endoscopic camera.

“The device enters through a port in the body and is designed so that it doesn’t move its insertion point. It rotates around that point and has a small wrist inside the body to improve dexterity while manipulating tissue,” Kuntz said.

Their smaller size makes them more capable, less invasive and more convenient for surgeons; on top of this, they can be steered to avoid delicate areas within the patient’s body.

“We can access deep sites in the brain and lungs by using steerable needles,” Kuntz said. “These can curve through the body, going around blood vessels or other obstacles, and still reach the target.”

Overall, the research and development of AI-augmented surgical robotics aims to alleviate several challenges in the health care system—addressing workforce shortages, reducing burnout and enhancing patient safety—while providing safer, more effective treatments. By focusing on narrowly defined workflows and integrating human-in-the-loop verification, these AI-enhanced tools can be deployed responsibly across a range of procedures.

---

The study titled “Medical needles in the hands of AI: Advancing toward autonomous robotic navigation” was published July 9 in Science Robotics. Funding was provided by the National Institutes of Health and the National Science Foundation. The content is solely the responsibility of the authors and does not necessarily represent the official views of these funding agencies.

• Evan Lerner Director of communications, John and Marcia Price College of Engineering
  801-581-5911 evan.lerner@utah.edu