Computer scientists studying the stability of Internet connections may someday find their research used to help patients suffering from schizophrenia, Alzheimers disease or stroke.
Recently published research indicates, for the first time, that networks in the human brain work similarly to those in the World Wide Web and other apparently unrelated networks. Thus, techniques to optimize one kind of network could potentially be applied to another.
Characteristics of individuals, such as height, weight and intelligence, tend to be distributed like a bell curve, with most close to some median value. Surprisingly, this is often not the case for the number of links of any individual component, or node, in a networked system. Networks often have very many nodes with very few links, and very few nodes with very many links.
The differences are often extreme; while the vast majority of nodes might have five or fewer links, the super-linked nodes, or hubs, could have as many as a million. In addition, very few “jumps” may be necessary to get from one node to another. Practitioners in the field describe such networks as scale-free (many hubs) and small-world (tightly connected).
This conception of networks is relatively new. Until the late 1990s, people who thought about networks assumed that links were distributed more or less randomly between nodes, and that all nodes in one network would have about the same number of links.
A study of the World Wide Web in 1998 refuted this notion and shocked researchers, as two authors of that study recounted in a 2003 Scientific American article. Subsequently, all sorts of networks were recognized as belonging to the new category, including social systems and protein interactions within a cell.
Now, the human brain joins this group. Dante Chialvo of Northwestern University and colleagues analyzed brain scans of subjects performing tasks including tapping their fingers in response to verbal or visual cues or simply listening to music.
They divided the brain into more than 30,000 cubes, called “voxels,” and examined which ones had correlated activity. The characteristics were clearly those of a scale-free, small-world network.
The new insight will be particularly useful for understanding the effect of brain damage, Chialvo told eWEEK.com. Currently, he said, people assume that brain networks are somewhat like road networks, in which local damage has a relatively local effect, but the airline industry is actually a better model.
“Scale-free networks are very particular in how they respond to damage. If you shut down Chicago, the number of flights you must take to get from city to city could go from two to 10 immediately.”
Chialvo said a confluence of advances had enabled the research. “Five years ago, there was no activity in network theory,” he said. In any case, the brain scans could not have provided fine enough resolution, and computers would have worked too slowly a few years ago.
“Were using 35,000 by 35,000 matrices,” he said. “Before, those calculations would have taken a week; now they take half an hour.”
In addition, he said, the tools of physics are increasingly being applied to understanding biology.
“Overall, our initial results indicate that the brain networks share these two fundamental properties, implying that the underlying properties can be understood using the theoretical framework already advanced in the study of other, disparate networks,” Chialvo said.
The work is published this month in the journal Physical Review Letters. The research group included scientists from the IBM T.J. Watson Research Center, in Yorktown Heights, N.Y.; and the University of Islas Baleares, in Mallorca, Spain.
Check out eWEEK.coms for the latest news, views and analysis of technologys impact on health care.