- Researchers believe the system will be useful in separating ‘real pain from fake’
- People express pain differently, which makes it hard for doctors to gauge
- The system is personalized by taking into account age, sex and complexion
- Researchers hope the system will one day be available as an app on phones
- Past research suggests computer systems are 30% more effective than humans
A computer is being developed that rates how much pain someone is in by detecting small changes to their facial expressions.
Such a system could help doctors determine how to treat patients and may save the prescription of painkillers.
Study author Dr Jeffrey Cohn from the University of Pittsburgh, said: ‘These metrics might be useful in determining real pain from faked pain.’
People experience and express pain differently, which makes it difficult for doctors to gauge the extent patients are suffering based on self-reported scores.
The researchers hope the system could one day be available as an app that doctors have on their smartphones.
A computer is being developed that rates how much pain someone is in (stock image)
How the system was created
The researchers ‘trained’ a computer based on videos of people wincing and grimacing in pain.
Each of the videos featured a person with shoulder pain who performed various movements before rating their discomfort.
This helped to create a system where subtle differences in facial expressions can be used to determine how much pain a person is feeling.
Large movement around the nose and mouth is thought to be particularly indicative of pain.
Could determine ‘real pain from faked pain’
The system was made more accurate by accounting for people’s age, sex and complexion.
Personalized approaches were found to be more reliable than one-size-fits-all. Age in particular is thought to influence people’s pain perceptions.
The researchers hope the system could one day be available as an app that doctors have on their smartphones.
Dr Cohn said: ‘These metrics might be useful in determining real pain from faked pain.’
Previous research by the University of California in San Diego revealed computer systems weed out ‘pain fakers’ 85 per cent of the time, whereas trained humans are only accurate in 55 per cent of cases.