Who is better at surgery: an experienced surgeon or a robot?
Typically, surgeons have to make relatively large incisions during surgery, while the robot’s small instruments can accommodate smaller incisions. Given this advantage of robotic systems, it is now common for surgeons to use remote-controlled robotic arms to perform surgeries — combining the precision of a seasoned human with the minimal invasiveness of a small robotic arm. However, in these cases, the surgeon still needs to control the robot, and a fully automated robotic system that outperforms the surgeon in terms of accuracy has not yet been realized.
However, a recent development suggests that robots may be able to demonstrate superhuman performance in the near future. In a paper published May 10 in the IEEE Transactions on Automation Science and Engineering, a multinational research team reports the results of a study in which robots were able to perform common procedures with the same precision as experienced surgeons. Training tasks while completing faster and more consistently.
Minho Hwang, assistant professor at Daegu Gyeongbuk Institute of Science and Technology, South Korea, participated in the study. He noted that many robotic systems currently rely on the automatic control of cables, which are subject to friction, cable coupling and stretching, all of which can make precise positioning difficult.
“When humans control the robot, they can compensate through the human’s visual feedback. But because of [these] positional errors, the automation of robot-assisted surgery is very difficult,” Hwang explained.
For their study, Hwang and colleagues took the standard da Vinci robotic system, a common model used in robot-assisted surgery, and strategically placed 3D-printed markers on its robotic arms. This allowed the team to track its movement using color and depth sensors. They then analyzed the arm’s movements using machine learning algorithms. The results show that the trained model can reduce the average tracking error by 78%, from 2.96 mm to 0.65 mm.
Next, the researchers tested their system against an experienced surgeon who had performed more than 900 surgeries, and nine volunteers with no surgical experience. Study participants were asked to complete the peg transfer task, a standardized test for trained surgeons that involves transferring six triangular blocks from one side of the pegboard to the other and back again. While the task sounds simple, it requires millimeter-level precision.
Study participants completed three different peg tasks using the da Vinci robotic system: unilateral (using one arm to move one peg), bilateral (using both arms to move two pegs simultaneously), and bilateral cross (using one arm The arm picks up the peg, transfers to the other arm, and places the peg on the board). Their robot-assisted performance was compared to a fully automated robotic system designed by Hwang’s team.
Using one arm, surgeons outperform autonomous robots in speed. But on more complex tasks involving two arms, the robot outperformed the surgeon.
For example, for the most difficult task (bilateral switching), the surgeon had a 100% success rate, with an average transfer time of 7.9 seconds. The robot had the same success rate, but the average transfer time was only 6.0 seconds.
Ken Goldberg, a professor of electrical engineering and computer science at UC Berkeley, also participated in the study. He said: “We were very surprised by the speed and accuracy of the robot, as it was difficult to surpass the skills of a trained human surgeon. We were also surprised by the robot’s consistency; it transferred 120 times perfectly, without A failure.”
Goldberg and Hwang note that this is an initial study in a controlled environment, and more research is needed to achieve fully automated robotic surgery. But to their knowledge, this is the first instance of a robot outperforming a human in a surgery-related training task.
“We have demonstrated that fast and accurate automation is feasible for a surgical task involving rigid objects of known shape,” Hwang said. “The next step is to demonstrate this in other tasks and in more complex human environments.”
In future work, the team plans to extend its approach to automated surgical subtasks, such as tissue suturing, and hopes to build on its methods of calibration, motion planning, visual servoing, and error recovery, he said.