Jan 3, 2018

What is the likelihood of medical robots making mistakes on their own rather than because of human error?

Copyright 2017 by Tom Wiklund. Property of UL, all rights reserved.Out-of-control robots (aka cyborgs) are the fodder of science fiction movies and television shows. Glowing red eyes, a flicker of sentience, and the pursuit of revenge against humans are fundamental elements of the good ones.

In reality, most robots are benign devices that do not look human and function only as they are programmed. They help automate warehouses, assist with manufacturing tasks, and even do vacuuming. Any “mistakes” committed by one of these can most likely be traced to some sort of programming error committed by a human. Similarly, medical robots are more often than not machines under direct human control, or at most, partial human control. Therefore, we would be scapegoating if we blamed a robot for a mistake.

Examples of medical robots

One such medical robot that is not “autonomous” is the da Vinci surgical system. The $2 million-dollar system enables a surgeon to work at a standalone console with hand controls that move separate mechanical arms and end effectors that perform tissue cutting, coagulation, and suturing tasks. Yes, the system includes a large tower with four arms that hovers over the patient and can have the appearance of a cyborg performing surgery. But this system includes a human in the control loop; the machine is not performing actions on its own.

The same can be said of CorPath’s angioplasty “robot” used by interventional cardiologists, which consists of a stand-alone, mechanical arm that advances a catheter into a coronary artery. But this action is only in direct response to the inputs of an operator seated at a nearby console that offers protection against repeated exposure to the radiation associated with fluoroscopy.

In some sense, an artificial pancreas is a much better example of a medical robot. One FDA-approved solution from Medtronic--the MiniMed 670G--senses a patient’s blood glucose (BG) level, and varies the infusion of insulin to keep the patient’s BG in the healthy range. There is normally no need for human intervention during the therapeutic phase of use.

Human input still a key factor

Is it possible for the latter type of robot to make a mistake? I don’t think so. A human still has to program the device to function properly in all possible use scenarios. Mistakes (i.e., use errors) remain a human ability, or more accurately, disability. Until medical robots start to rely on some elevated form of artificial intelligence to make their own decisions based on unforeseen conditions, it is up to medical robot manufacturers to do all they can to minimize the impact of these mistakes. But, the day will come when there are autonomous medical robots that do their own “thinking.” When we get to this point, perhaps we will need to include the robots in a transformed version of a usability tests.

Michael Wiklund is General Manager of Human Factors Engineering at UL. Read more about medical device design and usability on the UL Wiklund website.

Additional medical device regulatory resources:

Image ©2017 by Tom Wiklund. Property of UL, all rights reserved.


  • Michael Wiklund