News
Louis Dor
Dec 07, 2017
Stephen et al. 2017
A new international study has developed a computer model that can determine information about a person's health by analysing their face.
Published in the journal Frontiers in Psychology, the study says the computer model works in much the same way our brains do when analysing a person's health through their face.
The model, the study claims, is capable of detecting information about a person's Body Mass Index (BMI), body fat and blood pressure levels.
Study author Dr Ian Stephen explained:
We have developed a computer model that can determine information about a person’s health simply by analysing their face, supporting the idea that the face contains valid, perceptible cues to physiological health.
First, we used photos of 272 Asian, African and Caucasian faces to train the computer to recognise people’s body fat, BMI (a measure of body size) and blood pressure from the shape of their faces.
We then asked the computer to predict these three health variables in other faces, and found that it could do so.
The researchers then tested whether humans could do the same by asking 26 participants to change the appearance of faces in an app to make them look as healthy as possible.
We found that the participants altered the faces to look lower in fat, have a lower BMI and, to a lesser extent, a lower blood pressure, in order to make them look healthier.
This suggests that some of the features that determine how healthy a face looks to humans are the same features that the computer model was using to predict body fat, BMI and blood pressure.
The authors say the findings "provide strong support for the hypothesis that the face contains valid, perceptible cues to physiological health". While their computer model remains in an early stage at the moment, they hope it "could be used to help diagnose health problems in the future".
Time to hit the treadmill, it seems.
HT Daily Mail
More: These incredible photos show that you can look better after gaining weight
Top 100
The Conversation (0)