Wednesday, March 19, 2014

The virtual brain





The Glass Brain is a lovely visualization of commutations between brain regions. It's composited from ongoing EEG scans measuring electrical transmissions, and mapped in three dimensions onto a structural model of the same brain generated by MRI, as well Diffusion Tensor Imaging used to trace out the major transmission tracts.

Yes, scanning and creating a real time visualization of brain activity is hugely complex at present. Building a quality visual model like the one above is matter of fusing models which detect very different features and structures. My hat's off to the University of San Francisco's Neuroscience Imaging Center for tackling this almost baroque undertaking, and coming up with something so strikingly modern and almost minimalist in the end.

It's also got me thinking about how extraordinarily well engineered the brain is when it comes to minimizing the kinds of transmissions depicted in the Glass Brain. Or at least it does so whenever it can.

Long-range inter-region transmissions consume a lot of power. They generate a good deal of heat. The infrastructure to move them is bulky. The more inter-region chatter and required bandwidth, the less horizontal and vertical real estate there is for things like regulating hormones and keeping the lungs and heart pumping.

So, such communications are costly, but at times they're also very necessary. Creating things like, a coherent field of vision, or an abstract comprehension, involve lots of regions talking to one another. Recalling an idea means the associative prefrontal cortex communicating with sensory cortices in order to access memories and visual models to knit together, as well as the areas outside the neocortex involved in real time working memory, motivation, and focus. Incorporating that idea into a model of possible futures means the prefrontal cortex also interfacing with the limbic system to mediate the emotions associated with the concept and the potential outcomes.

How does the brain keep those transmissions or their costs down when it can? In part by doing some virtualization and modeling of its own. Regions simulate one another. They read what are for the most part slow, noisy, short, and therefor cheap-to-send incoming transmissions, and use that constrained data to update those simulations of what the other connected areas are up to. Then they respond in kind with a minimum of processed information, selected in part based on the data requirements of the other participants, as they see them.

Think of two people navigating their way across the countryside, separated several hundred feet by a ravine filled with a loud river. They can shout at one another, and gesture, but here are all kinds of difficulties associated with those means of transmission. So rather than try and communicate every little detail necessary to get from parallel points A,B to C,D, they use what data is moving back and forth to figure out each other's intentions.

In a sense you're something similar: In part a collection of generalist virtual machines running on specialist hardware modules. Software talking with one another in what is for the most part a low-bandwidth, long-range transmission environment, with a low signal to noise ratios.

Then there is the associative prefrontal cortex that is the seat of our decision-making executive. This area of the frontal lobe is a continual generator of large-scale virtual machines. It sets up multiple feedback loops networks, which each makes use of a slice of several regions, creating competing models of the future or present when thinking. Models that incorporate everything from simulated motions, vision, audio, and even emotional reactions.

In someways, this makes the perfrontal cortex an exception to the rule of low cost transmissions. It's communication requirements are high enough that it rates its own super highway or privileged transmission network of fiber tracts.

All of this together, local and brain-wide virtualization creates the rich and textured models of the world and self that we call consciousness and being.

No comments: