August 9, 2013
New Research Sheds Light on How The Brain Processes Visual Information
|New research coordinated by International School for Advanced Studies of Trieste in collaboration with the Polytechnic University of Turin, studied the "code" used by the brain to categorize objects we see and sheds light on a scientific debate: When, in the process of the brain analysis of visual stimulus, does "shape become meaning?"|
Opening our eyes and seeing the world before us, full of objects, is a simple action we may take for granted. Yet our brain is constantly carrying out a huge analysis only to let us see a flower, a pen, the face of our children.
Where exactly in our brain does shape become meaning?
A group of scientists coordinated by Davide Zoccolan of SISSA of Trieste, in collaboration with the team headed over by Riccardo Zecchina of Polytechnic University of Turin, studied a specific area of the brain that is located precisely halfway between the visual and the semantic analysis, shedding light on its function. The study has been just published in PLOS Computational Biology.
|Image Source: PLOS Computational Biology|
At least this is what the majority of scientists had been assuming until a few years ago, when in 2007 and 2008 two studies raised doubts about such theories, suggesting that the IT may be implied in functions of semantic processing of visual objects.
In other words, scientists until then had thought that such brain area represented objects mainly according to their visual properties (such as, for instance, shape), while the new studies claimed that the meaning of the objects had in this case a predominant role.
Zoccolan analyzed the electrophysiological data he gathered on primates a few years ago, while he was working in the laboratory of James DiCarlo, at MIT in Boston. The data have been processed employing various machine learning computational techniques, including a clustering algorithm recently developed by Zecchina's team.
Such analysis methods have helped verify whether the categorization of the objects in the IT occurred according to hierarchies based on the similarity at the level of shape or at the level of meaning. In short, the researchers asked if for the IT an orange looked more similar to a ball children play with (they are both round) or to a banana (they both belong to the 'fruit' category).
"Our data indicate that most objects are categorized according to their visual similarity, while semantic membership seemed strongly represented only in the case of a specific class of objects: four-legged animals," explains Zoccolan. "The traditional IT model, considered as the area which encodes mainly visual information, remains still valid although our study does not rule out the notion that certain semantic classes which are particularly relevant to primates may be also represented here."
Overall, these findings reassert the primary function of IT as a conveyor of explicit visual shape information, and reveal that low-level visual properties are represented in IT to a greater extent than previously appreciated. The work demonstrates how combining a variety of state-of-the art multivariate approaches can substantially advance our understanding of neuronal coding of visual objects in cortex.
"Besides this confirmation we have also observed something unexpected", adds Zoccolan. "It has been known for a long time that the IT is a processing station of complex information, " explains the scientist, "basically, the most advanced stage of visual information processing, where objects are categorized explicitly, which means that the entire structure of complex objects, rather than single parts of them, is encoded. What we have observed instead is that the IT retains also a 'raw', lower-level coding, for instance whether an object's color is dark or light, if it's big or small, and so on. This is basically an innovative assumption that changes our interpretation of the inferotemporal area's function."
|By 33rd Square||Subscribe to 33rd Square|
Topics - artificial intelligence , Davide Zoccolan , International School for Advanced Studies , IT , machine learning , neuroscience , Polytechnic University of Turin , semantic processing , Visual processing
33rd Square explores technological progress in AI, robotics, genomics, neuroscience, nanotechnology, art, design and the future as humanity encroaches on The Singularity.