Article intro - The Guardian on the autonomy of surgical robots
A nice article appeared from Charlie Metcalfe in the Guardian recently: "Robot surgeons provide many benefits, but how autonomous should they be?"
"Combined with AI and other novel technologies, engineers are developing advanced robotics to herald another new era for surgery – and this time, the surgeon’s role in the operating theatre may change altogether.
Although robots are put to a variety of tasks in surgery, their use as a tool in performing laparoscopy – otherwise known as keyhole surgery – has attracted the most attention within and outside medicine. Keyhole surgery reduces the time patients need to recover by operating through smaller incisions. This subsequently reduces the chance that patients catch infections, and so accelerates their recoveries.
Without robots, keyhole surgery requires a very high level of skill. Surgeons need to operate at awkward angles, moving their hands in the opposite direction to that in which they want their instruments to move inside the body. With robots, surgeons can perform more complex operations that might otherwise have demanded open surgery, they suffer less physical strain, and they require less training time. Moreover, they are getting better at using the robots.
“Some of those patients who have ultra-advanced diseases involving blood vessels at the back of the pelvis might still get an open operation,” says Deena Harji, a colorectal surgeon in Manchester, “but we’re starting to see some very early case studies coming out where they’re starting to have robotic approaches applied to them, at least in part. When robotics started 20 years ago, that group would not have been eligible for robotic operation. But as we have developed experience and knowledge, we can offer really complex patients robotic surgery.”
Surgeons are limited by their physical capacity, and their minds are limited in their potential to learn and improve. That’s why engineers are hoping robotic systems combined with AI might be able to surpass the skills of human surgeons to produce more consistent results, with fewer errors.
Last year, engineers at Johns Hopkins University in the US came one step closer to realising that goal. In what they described as one of the most delicate procedures in the practice of surgery, their Smart Tissue Autonomous Robot (Star) sutured the ends of a severed intestine together in four pigs – while they were under anaesthetic. According to the engineers, it performed better than a human surgeon would have. “Our findings show that we can automate one of the most intricate and delicate tasks in surgery,” Axel Krieger, an assistant professor of mechanical engineering, and the project’s director, said at the time.
The Star’s procedure was not the first time a robot had performed with a level of autonomy in surgery. The TSolution-One device (formerly called RoboDoc), for example, is FDA-approved to prepare human limbs for joint replacements according to a surgeon’s plan. What makes the Star’s procedure special was that it performed its task using keyhole surgery – a world first.
Surgical robotics presents a good opportunity for engineers to introduce autonomy because of the vast volume of data that devices can collect. An intelligent system, once developed, can use this data to teach itself. In theory, it could become better with each operation that it performs as it gathers more and more data. This could help healthcare organisations “standardise” the results of operations.
Mark Slack, the chief medical officer at CMR Surgical, which manufactures another surgical robot, Versius, says that manufacturers have failed to exploit this data until now. That’s why they and researchers such as those involved in the Star project are scrambling to collect and process as much as possible. “Data, data, data,” Slack says. “This data has had significant untapped potential.”
Despite the Star team’s success, it’s still too early to forecast autonomous surgery in hospitals any time soon. Engineers talk about “levels of autonomy”. For a robotic device, the question is not whether it is autonomous or not; the question is how autonomous it can be. And the Star system performed only a small section of a complete surgery without human help. In fact, it even needed humans to apply a fluorescent marker to guide its movements. “You’re not supposed to call it autonomous surgery,” Tamás Haidegger, an associate professor of intelligent robotics at Óbuda University in Budapest, says. “This is automating one particular surgical subtask.”
Haidegger makes what he believes is another important distinction – between the kind of complexity required for a system like the Star and the devices used in hospitals. Standard laboratory best practice in research environments often falls short of the safety and design standards of clinical settings, he says.
For use in clinical environments, manufacturers need to be able to explain exactly how their devices work, which continues to prove a challenge for people who develop AI. There’s also the impending introduction of AI-specific regulation that governments across the world, including the UK and EU, are developing. Autonomous surgical robots will need to comply with those too.
According to Haidegger, this all amounts to a very expensive process for manufacturers to prove that their devices meet the regulatory requirements. Each device needs to gain approval for each new field of surgery, one at a time, which has already decelerated the adoption of the human-operated robots used today. It will take a lot more work for a commercial manufacturer to decide that the potential profit justifies the cost of research and development. “It’s not going to radically change medical devices overnight,” Haidegger says.
Comments