Let us level with you…
As should be obvious, surgery requires extensive training. In the US at least, students have to successfully get through university, postgraduate medical school, and years of practical training at a surgical residency program before they can be certified by the American Board of Surgery. They also have to learn to navigate the sensitive nature of patient-doctor relationships.
Robots, on the other hand, can skip school. Machines don’t need to accumulate a wealth of biology lessons in order to perform certain surgeries. And unlike their human counterparts they do not get bogged down by stress or fatigue.
Guang-Zhong Yang, professor and director of the Hamlyn Centre for Robotic Surgery at Imperial College London, believes a classification system is necessary to overcome the “regulatory, ethical and legal barriers” as medical robots get more autonomous.
In other words, these classifications formally define the capabilities of each machine. This makes it a lot easier to work out how much scrutiny each robot requires before it is allowed to work on human flesh in the field: the more complex the design, the more reviews and testing it should go through by regulators.
The six levels are:
- No autonomy – device is controlled by user, like prosthetic limbs.
- Robot assistance – robot provides some mechanical assistance such as helping patients move and supporting their balance.
- Task autonomy – robot can do certain tasks autonomously, such as a mechanical arm sewing stitches.
- Conditional autonomy – a system can generate its own tasks but mostly relies on humans to decide, and can perform tasks independently.
- High autonomy – robot can make medical decisions but under the supervision of qualified doctors.
- Full autonomy – robot can perform the entire surgery as well as a human general surgeon, without supervision.
America’s Food and Drug Administration (FDA) reviews and scrutinizes medical devices before they enter the market, a process that takes ten months on average – or as much as 54 if the agency thinks the product is potentially high risk. These gadgets and tools fit within the proposed levels zero to three.
According to Yang, autonomous robotic surgeons – while still science-fiction – come in at levels four and five, and fall outside the grasp of the FDA. That’s because they are no longer just bits of equipment used in theatre: they are decision-making, medicine-practicing computers. In the future, the difference between an AI system and a surgeon will be that a patient’s life is held in the balance not by human hands, but by actuators and software code.
The fear is that the FDA will give the green light to a level four or five machine without considering the medical skill of the robot. Think of it this way: the drug agency can approve a new scalpel for use on the basis that whoever ends up using the blade is suitably qualified. The skill of the doctor using the tool is something for practitioner panels to worry about.
Now imagine an intelligent machine with software, sensors, and a scalpel on the end of a robot arm. The system, as a device, can be physically safe and clean, but can the drug watchdog assess the medical skill? Apparently not.
The next step, we’re told, will be to get the American Board of Surgery on the case to test the ability of level four and five robots, just as if they were real surgeons, while the FDA checks that the physical design is safe and up to scratch.
“Unlike autonomous cars, the spectrum of tasks, environments, technology, and risk is practically limitless,” said Yang and his colleagues.
Crucially, the levels will also help decide the different roles and regulations for different types of robots. For example, a care assistant robot at level four will not need to be scrutinized as heavily as a robot surgeon at level five.
The recent burst of interest in deep learning has spurred a wealth of machine learning algorithms that help agents navigate their environments. As they improve, robots have the potential to be more autonomous and could steal more practical jobs, while medical experts focus more on “diagnosis and decision-making,” the authors argue.
“As the autonomous capabilities of medical robotics grow, most of the role of the medical specialists will shift toward diagnosis and decision-making. This shift may mean that dexterity and basic surgical skills may decline as the technologies are introduced, with implications for training and accreditation.
“At the same time, pattern recognition and self-learning algorithms will improve, allowing medical robotics an increasingly larger role in higher levels of autonomy.” ®