Studying the human connectome.

02 March 2013

Late last year I did an article about the simulation of parts of the the human brain on a massive scale called SPAUN that was implemented using software called Nengo. The basic concept behind SPAUN, as you may recall, is that it is a functional model of some aspects of the human brain which duplicate some of the neural networks as well as the myriad connections between them. What isn't obvious is that this connection model was developed in part through the microscopic examination of many human brains post mortem plus many different kinds of scans carried out on living people over many, many years. Also, a great deal of knowledge has come from projects which have analyzed partial connectomes of other species, such as the Open Connectome Project. In certain structural and functional respects, mammalian and reptilian brains are comparable to human brains and the knowledge gained from them has been applied to human neuroanatomy.

This body of knowledge is growing by leaps and bounds thanks to data from an ambitious effort called the Human Connectome Project, a $40mus undertaking which aims to build a much more comprehensive map of the human central nervous system at the cellular level than any previous effort to date. It is hoped that the data gathered by the HCP can be used to better understand certain neurological phenomena, such as dyslexia or autism. It is also hoped that certain cognitive proclivities will be better understood, such as having a talent for mathematics, music, or languages. In addition to the data collected (which you can apply to gain access to, though I have a gut feeling that not just anyone is going to be accepted) funds have been allocated to advance the state of the art in neuro-imaging technology, specifically toward increasing the resolution of scanners.

The techniques used to map the human connectome include EEG, functional MRI to measure oxygen utilization by neurons, and diffusion MRI to track flows of various molecules through neurons to record the patterns they make on the macroscale. The quantities of data collected by each scan are nontrivial to say the least, and require vast amounts of disk space and processing power to store and manage. For example, the Open Connectome Project built a partial connection map of the retina and visual cortex of a lab mouse, and it's about twelve terabytes in size. Human brains are orders of magnetude larger and more complex than a lab mouse's, and the scans are going to be at a much higher resolution, so I can't even begin to guess how big a full human connectome is going to be (though people much smarter than I have no doubt made estimates and back-of-the-envelope computations). The fascinating thing about these data sets is that they can be visualized so that you can see what's happening inside the mapped brains at a moment in time. Discrete functional groups stand out like lightning strikes, and by correlating those bundles of neuronal activity you can start to see how different parts of the brain cooperate (or not) to carry out certain tasks or kinds of thoughts.

So, what can this tell us? Of what use is the Human Cognome Project?

First, the HCP will help the related field of psychology better understand human cognition and behavior. While the chemoelectrical aspects of human thought under certain conditions are known, we don't yet have a great deal of knowledge about what the other mechanisms of the brain are up to at the same time. Electrical activity is not the end-all-be-all of thought; in fact, it seems as if it's a shadow of what is actually happening. Second, the complexity of the interconnections of neurons between one another, and the complexity of the interconnections between all of those networks of neurons is staggering. The human cerebral cortex - the outermost part of the brain - is between 1mm and 4mm thick and is comprised of around 100 billion neurons. The rest of the brain, the matter below the surface that you can't see, seems to be mostly comprised of the connections between that thin layer of neurons on the surface. I've no doubt that there is an amazing amount of emergent activity taking place in such a dense network, and the HCP will allow us to get a better picture of what those emergent patterns look like. Third, by characterizing and correlating what (relatively) baseline people's cortical activity and structures look like, we will gain better insight into how to treat (or at least manage) brain injuries and neurological abnormalities. At the very least, surgical intervention will probably become that much more precise.