Skip to content

Into the nitty-gritty: gradients in brain networks

June 24, 2011

The world of functional imaging research is on fire right now with connectivity studies. (See my post here for an introduction to the domain of functional connectivity as a tool for studying the brain.) Although we have miles to go before we sleep, the study of distributed networks in the human brain is the forefront right now in bridging the field of psychology with the discipline of neuroscience…a bridge which science will be trying to build in a comprehensive way for the foreseeable future.

The most recent work that I will be presenting at the Human Brain Mapping (HBM) conference in Quebec addresses the relationships between two of the major functional networks in the human brain. Namely, the default mode network, and the attention control system of networks.

It has been observed that these networks are anti-correlated. In other words, as one network increases its activity, it is accompanied by diminished activity in the opposing network. The “how” and the “why” behind this anti-correlated activity is an open question…the former being a physical question, and the latter a question pertaining to the psychology correspondent to the system.

The findings I will present next week at HBM principally address topics related to the former set of questions, i.e., how these networks are physically interacting with one another. There are also insights, though, that may be gleaned about the psychological phenomena that are emergent from the mechanical dynamics in the brain.

In short, each node in these networks has a distinct hub that is the hot-spot of anti-correlation between the networks, and there is a gradient of diminishing anti-correlations fanning out from each respective hub. This means that these hot-spot points in each node of the network are likely the signal sources for turning one network down when the other network becomes active, and that the on/off signals spread outward from these points to the rest of the network.

Interestingly, these gradients become sharper with age, meaning that the network boundaries and the anti-correlations between the networks become more pronounced along the course of brain development. This may ultimately implications for why focused attention on a task increases naturally with age, in addition to possibly providing insights into what can go wrong in the development of attention-switching in the brain.

It’s an exciting time to be researching the brain–stay tuned for more to come!


Figure details: The highlighted regions in panel A are the key regions in the default mode network (DMN). The color gradients represent the correlation of the DMN with the attention control network (ACN). As can be seen in the figure, each hub of the DMN has a gradient of connectivity to the ACN, with a core of minimal correlation that is the likely source of inhibition for the network. Panel B illustrates the same idea for the ACN relative to its correlation with the DMN.

Figure details: (Acronyms: DMN=default mode network; ACN=attention control network; ROI=region of interest.) For DMN ROIs, the more strongly connected an ROI was to the network, the more strongly its connectivity to the DMN increased with age (r = 0.57, p = 0.8 * 10-72) consistent with a “sharpening” of boundaries during development.
A) Scatter plot of DMN ROIs comparing mean correlation to DMN (x-axis) to change in DMN correlation with age.
B) Mean correlation of DMN ROIs to the DMN vs. change in correlation with age to the ACN.
C) Mean correlation of ACN ROIs vs. change in correlation to ACN with age. Insula and anterior cingulate ROIs are shown in red and blue, respectively.
D) Mean correlation of ACN ROIs to DMN vs. change in correlation with age to ACN.


From → Uncategorized

  1. Bert permalink

    Hello Positive

    Why do you assume that the gradient change with age is the result of anything but vascular changes?

    • permalink

      Thanks for the question, Bert. The changes in network gradients with age correspond to whole brain findings wherein regional homogeneity is likewise altered across the lifespan. Given corroborating structural DTI data that track changes in axonal development with increasing age, the more likely explanations for homogeneity and gradient modifications seem to be connectivity modifications, rather than vascular dynamics. A vascular argument would require brain-wide decreases in blood flow efficiency in a peculiar way that would result in both the decreased local homogeneity in the BOLD signal–it seems like a bit of a stretch.

  2. Bert permalink

    Hello Positive

    I don’t know the fMRI and aging literature but there are definitely global changes in perfusion with aging. There are changes in the BOLD response with age as well, correct. So couldn’t these gradients in connectivity actually be related to gradients in BOLD response (for example a diminished and sluggish BOLD response) and have nothing to do with connectivity? This is not my area of expertise so please forgive my ignorance.

    • permalink

      Not a problem, Bert!–I appreciate the thoughtful conversation. Your question is a good one. Increased delays in microvascular response across the brain in conjunction with aging would not result in the decreases that are observed in functional connectivity: functional connectivity is a metric of how closely correlated different regions of the brain are to one another, and the correlation measurement is made across a window of BOLD frequencies (between 0.001 and 0.1 Hz). This means that the maximum period tested for correlation is 1000 seconds. Any BOLD signal delays caused by microvascular slowness with aging would still fit well within this long of a maximum period. Even if the microvasculature slows unevenly throughout the brain with aging, it would only introduce a phase lag between the peaks of the BOLD signal for any two regions. However, it would not change the degree of correlation between the BOLD signals from the respective regions.

      A diminished BOLD response–in contrast to a sluggish one–would affect the amplitude of the peaks that comprise the time series of any given region of interest in the brain. Because the amplitude of the BOLD time series of every region of interest is normalized before correlation is calculated, this means that the absolute value of the BOLD signal is likewise irrelevant–along with phase–in the ultimate determination of the functional connectivity.

      Not trying to be erudite in writing tone–let me know if that’s a bit too jargony, and we can certainly transfer over to email. But I think that is a solid and correct answer to your question.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: