ACM TechNews
Carnegie Mellon Study Identifies Where Thoughts of Familiar Objects Occur Inside Human Brain
Carnegie Mellon News (01/03/08) Spice, Byron; Watzman, Anne
Carnegie Mellon University computer scientists and cognitive
neuroscientists, using combined methods of machine learning and brain
imaging, have developed a way of identifying where people's thoughts and
perceptions of familiar objects originate in the brain by identifying brain
activity patterns associated with the objects. The technique was developed
over two years by neuroscience professor Marcel Just and computer science
professor Tom M. Mitchell. Study participants were shown line drawings of
10 different objects, five tools and five dwellings, and asked to think
about their properties while lying in a MRI scanner. The researchers were
eventually able to identify which picture the subject was looking at based
on their characteristic whole-brain neural activation patterns. Just and
Mitchell discovered that the activation patterns evoked by an object are
not located in a single place in the brain. The machine-learning part of
the study used a computer algorithm to extract patterns from a
participant's brain activation. Data collected in one part of the study
was tested against the algorithm on data from another part of the same
study so that the algorithm was never exposed to the same patterns it was
tested on. The algorithm was able to identify a participant's thoughts
based on the patterns extracted from other participant's, indicating that
different brains show the same activity patterns when viewing the same
object. "This first step using computer algorithms to identify thoughts of
individual objects from brain activity can open new scientific paths, and
eventually roads and highways" says University of South Carolina assistant
professor of psychology Svetlana Shinkareva, the study's lead author. "We
hope to progress to identifying the thoughts associated not just with
pictures, but with words, and eventually sentences."
http://www.cmu.edu/news/archive/2008/January/jan3_justmitchell.shtml
© Copyright 2008 Information, Inc. This service may be reproduced for internal distribution.