TECH

AI-Powered Camera Can Track When Hospital Employees Haven't Washed Their Hands

Is this the way to combat superbugs or is it a step too far?

15/08/2017 12:07 BST

Tackling the issue of hygiene in hospitals is a difficult proposition. While you can provide staff and visitors with the tools they need, ultimately it’s up to them to make sure they use them.

Well now a joint research team from the University of Stanford and Swiss Federal Institute of Technology have developed an AI-powered camera that can actually track people around the hospital and make sure they’re washing their hands.

Department of Computer Science Stanford University

To make sure none of the images it takes are identifiable, the team developed something known as a depth camera.

What this camera does is instead of assigning a pixel to a colour, it assigns it a distance from the camera. What you then have is an equally accurate image but one that’s made up of distances instead of colours.

This is perfect for anonymous monitoring because it can still assign an identity to a person but not one that can then be used by hackers or criminals.

To test their technology the team were able to install their cameras in two participating hospitals.

Cameras were set up around an acute care paediatric unit and an adult intensive care unit.

The cameras then tracked people both entering and leaving those units and measured how many followed the correct hand-cleaning protocol.

Loop Images via Getty Images

What they found was that of the 170 people that were tracked entering the room, just 30 were found to be washing their hands correctly.

At the end of the experiment the team had collected over 150,400 images, 12,292 of which contained pedestrians using the dispenser.

It should be noted that the system wasn’t flawless, in their study the team noted that when then system was double-checked using human monitoring their system had 75% accuracy.

While it’s not perfect, they argue that 75% is still considerably higher than any other form of observational method that’s currently in use.

If monitoring staff in hospitals sounds concerning, then you might be surprised to learn that it has actually been happening since 2012 when a number of hospitals started trialling a human-based version of this system.

Instead of using machine learning to anonymously track staff, the original system used camera footage that was then sent to human analysts who would then monitor how often staff were washing their hands and then send the results to the hospitals themselves.

As you can see this system offers a number of benefits not least because it’s anonymous but also because it removes the human element entirely.