Monday, July 7, 2008

Your Brain On Graph Theory

Using diffusion imaging, a group led by Indiana University has completed a connectivity map of the brain. They then applied all kinds of fun graph-theoretic operations on the data set. (Here's a paper from a few years ago (PDF), which in turn references a set of matlab functions that allow the dataset to be browsed. Can't seem to find the new research on the site for the lead researcher, Olaf Sporns.)

We've known for a long time that the cerebral cortex organizes itself into various regions and that those regions project forward and backward to each other, as well as to other non-cortical structures. But determining those projections isn't easy, as they're buried in a confusing tangle of white matter. Functional MRI studies could tell what regions were active during what activities, and gross dissection could pick out some of the obvious projections, but diffusion imaging, which can detect the drift of water molecules along axons, was needed to make sense of the subtler projections in the white matter. This appears to be the most comprehensive mapping done so far.

Top-line discovery: There's a core set of regions that seem to be connected to everything:
The most highly connected node, which Sporns dubbed the core, is located at the back of the head, in parts of the brain known as the posterior medial and parietal cerebral cortex. The node lies on the shortest path between many different parts of the neural network. "It's highly connected amongst itself, but also highly central with respect to the rest of the brain," says Sporns. "Network studies in other fields, from the Internet to protein interaction networks, suggest that these kinds of highly connected nodes tend to be very important for determining what the network does as a whole."

Previous functional brain-imaging studies have also highlighted this region: it's one of the most metabolically active parts of the brain, particularly when people are cognitively at rest, meaning that they are awake and alert but not engaging in any particular task. "People refer to this as the resting state, daydreaming, or self-referential processing," says Sporns. As part of the new study, Sporns and his colleagues also used functional magnetic resonance imaging to measure blood flow to different parts of the brain. They found that it correlated with the level of white-matter connectivity in the individual subjects.
So we now have a decent model for a cortical column. We're developing a connectivity map of the cortex. Presumably, we'll also have a complete map of how the cortex maps to other structures in the brain. At some point, this all becomes simulatable. (We're not quite there on the computer hardware front, but we'll get there.) From there, we'll finally be able to extract enough information from those simulation to make some inferences on how intelligent, even conscious, processing occurs. After that, we're into the realm of engineering, and the world will be a very different place.

No comments: