Computers 'decode' the human brain

Scientists can predict activity patterns in the human brain when prompted by certain words

Scientists have come up with a computer-based system which can predict activity patterns in the human brain when prompted by certain words.

The results of a study published today in the journal Science show how scientists can use magnetic resonance imaging (MRI) techniques to monitor blood flow patterns in the brain.

This has allowed them to forecast the response to around 60 nouns associated with senses, including sight, touch, taste or smell.

Using a computer-based system, researchers at Pittsburgh-based Carnegie Mellon University were able to use MRI scans to prove a 72 per cent accuracy rate in predicting which verbs the brain would associate with a given noun.

"The bottom line, and this is what's really new here, is that nobody had previously even tried to build a theory or computational model that would predict neural activity for arbitrary words," said co-author of the study Tom Mitchell, from the university's computer science's machine learning department.

The researchers are hoping to extend the scope of their work to include using their brain-scan based program to garner greater understanding of medical conditions such as autism or schizophrenia.