Feb 10 2011

Evidence for Neural Networks

It is common knowledge that the human brain is horrifically complicated – perhaps the single most complex thing of which humans are aware. I am often asked if we understand how the brain works, often phrased to imply a false dichotomy, a yes-or-no answer. Rather, we understand quite a bit about how the brain is organized, what functions it has and how they work and connect together, and we know quite a bit about brain physiology, biochemistry, and electrical function. But there is also a great deal we do not know – layers of complexity we have not yet sorted out. I would not say that the brain is a “mystery” – but rather that we understandĀ  a lot, but we also have much to discover.

One aspect of brain function that is an active area of investigation is the overall organization of brain systems. Specifically – to what degree is the brain organized into discrete modules or regions that carry out specific functions vs distributed networks that are carrying out those functions? IĀ have written about this debate before, concluding that the answer is both. As is often the case in science, when there are two schools of thought, each with compelling evidence in their favor, it often turns out that both schools are correct.

I would summarize our current knowledge (I would not call this a consensus as there is still vigorous debate on this issue, but this is how I put the evidence together) as the brain being comprised of identifiable regions that are specialized for a specific type of information processing. These regions, or modules, connect and communicate with other regions in networks. Some of these networks represent discrete functions themselves, but they may also just be ways for different modules to communicate the results of their processing to other modules. A given module may participate in many networks, although they will tend to cluster around the same theme.

In the last decade fMRI technology as steadily improved giving researchers the ability to visualize modules and networks in action. Studies with fMRI are still very tricky, and it seems that there are a lot of false positives out there, but there are some quality studies also. The view of brain function emerging from these studies supports the general notion of modules involved in networks.

A new study reports on efforts to map neural networks in the brain and finds that they can be correlated with specific cognitive functions – lending support to the role of networks in brain function. From the abstract:

The cognitive domains included processing speed, memory, language, visuospatial, and executive functions. We examined the association of these cognitive assessments with both the connectivity of the whole brain network and individual cortical regions. We found that the efficiency of the whole brain network of cortical fiber connections had an influence on processing speed and visuospatial and executive functions. Correlations between connectivity of specific regions and cognitive assessments were also observed, e.g., stronger connectivity in regions such as superior frontal gyrus and posterior cingulate cortex were associated with better executive function. Similar to the relationship between regional connectivity efficiency and age, greater processing speed was significantly correlated with better connectivity of nearly all the cortical regions. For the first time, regional anatomical connectivity maps related to processing speed and visuospatial and executive functions in the elderly are identified.

Essentially they found that the strength and speed of connections among different regions of the brain (networks) were the best predictor of cognitive function in the elderly.

These results make sense from my summary above – brain functions depend upon specialized regions engaging in specific networks, and further on communication among various regions and networks. As overall processing speed decreases we would predict to see functional decline in many areas, especially global functional areas like executive function (the ability to strategically plan and control our own behavior).

While this study, combined with other studies, does support the role of networks it does not eliminate the role of modules in brain function. Atrophy or damage to specific brain regions also correlates with impaired function. So again we keep coming back to the conclusion that brain function is a combination of modules engaging in networks – and damage to either will reduce function.

I suspect that as our tools improve and our mapping and models become more details we will discover further layers of complexity only hinted at now. For example, we may find that some cognitive functions require combinations of networks (networks of networks of modules). The number of possible patterns of activity increases exponentially with each layer of complexity, so researchers definitely have their work cut out for them.

While research such as this demonstrates that we are making steady progress in understanding the brain, we are also getting to point where we have a better idea of how much we still do not know. Also, progress in neuroscience has been clearly accelerating recently, boosted by new technologies such as fMRI, EEG mapping, and transcranial magnetic stimulation. Because these technologies are fundamentally computer-based, Moore’s law is in full effect, and so they are likely to continue to improve geometrically (doubling every 18 months or so) along with other computer-based technology.

Further, there are parallel research programs that are attempting to create virtual models of the brain, and they are making steady progress. While at the same time computer scientists are trying to develop better artificial intelligence systems which are drawing from neuroscience and potentially might feed back into our understanding of neuroscience.

In short – we are at the beginning of what may be the progressive merger of neuroscience and computer science. And this process is accelerating.

5 responses so far