Twitter may already be used to plan social lives, interact with celebrities and communicate with friends.

But now a researcher believes the social networking site could have a far more serious use - tracking disease.

Researchers have already used the site to track flu as it spreads through New York using a 'heatmap' of users who complain of being ill.

Adam Sadilek at the University of Rochester and his team analyzed 4.4 million GPS-tagged Tweets from over 600,000 users in New York City over the course of one month in 2010.

They trained their artificial intelligence algorithm to ignore tweets by healthy people such as those claiming they were 'sick' of a particular song, and trained it to find those who were really ill.

Sadilek says the key to his system is friendships.

'Given that three of your friends have flu-like symptoms, and that you have recently met eight people, possibly strangers, who complained about having runny noses and headaches, what is the probability that you will soon become ill as well?' he said.

'Our models enable you to see the spread of infectious diseases, such as flu, throughout a real-life population observed through online social media.'

The tweets were plotted on a map, and used to predict when a particular users was at high risk of getting ill.

'We apply machine learning and natural language understanding techniques to determine the health state of Twitter users at any given time,' Mr Sadilek said.

'Since a large fraction of tweets is geo-tagged, we can plot them on a map, and observe how sick and healthy people interact.

'Our model then predicts if and when an individual will fall ill with high accuracy, thereby improving our understanding of the emergence of global epidemics from people's day-to-day interactions.'

The heatmaps show a city going through a flu epidemic.

The more red an area is, the more people are afflicted by flu at that location.

'We show emergent aggregate patterns in real-time, with second-by-second resolution,' boasted Sadilek.

'By contrast, previous state-of-the-art methods (including Google Flu Trends and government data) entail time lags from days to years.'

The algorithm looked not just at users' friends' health, but also strangers in the same area.

The algorithm was correct 90 percent of the time and about eight days in advance, the team said.

In unpublished findings described to New Scientist during an interview at the Conference on Artifical Intelligence in Toronto, Canada,the team also revealed that people who go to the gym regularly are moderately less likely to get sick.

People with low socio-economic status, on the other hand, are much more likely to become ill.