Fundamentally, neural activity is electical potentials and chemical signals. fMRI doesn't measure these directly. Rather, it measures changes in the oxygen content of blood in different parts of the brain.
The more the brain cells are firing, the more oxygen they use up, although oxygenation actually increases as a kind of compensation for the activity and this increase is what gets measured. The oxygenation changes associated with neural firing is called the BOLD response.
Using fMRI you can measure BOLD and end up with some pretty blobs of activation. But what does it mean for a region of the brain to be activated? Just as no man is an island, no brain region can do anything on its own. Every area gets inputs from other areas, and sends outputs as well.
So if an area gets more active, that could mean one or more of three things:
- It's sending more outputs
- It's getting more inputs
- It's doing more "internal" processing within that area - "talking to itself".
They took some rats, and stimulated their whiskers. Using electrodes, they measured blood oxygenation changes in an area called the barrel cortex, which is known to deal with whisker-based sensations (they didn't actually use fMRI, but this would be seen as a BOLD signal if they had.)
In many ways this is not surprising: it would be weird if the cortex were just picking up signals and doing nothing with them. However, it's good to be able to put a figure on just how much intra-cortical processing contributes to the fMRI signal. In rats, at any rate.
Harris S, Jones M, Zheng Y, & Berwick J (2010). Does neural input or processing play a greater role in the magnitude of neuroimaging signals? Frontiers in neuroenergetics, 2 PMID: 20740075
8 comments:
This highlights concerns that I have had with correlation analysis (aka connectivity) in fMRI which does not separate input, output and internal dynamics and can only measue correlations on the long (0.5 seconds or more) time scale associated with fMRI.
It would appear that the brain acts more like an OS controlling peripheral hardware: Specialized functional areas of the brain (like the video and audio chipsets of a computer) storing information and dumping that information in a temporal manner (like the duty cycle of computer chipsets) associated with the functional area onto communication channels (like a computer's bus system) after receiving a very simple signal (like a simple one-bit "play" instruction sent to a register of an audio chipset).
But neuroskeptic are you sure you want to write this:
"they didn't actually use fMRI, but this would be seen as a BOLD signal if they had."
MIGHT be seen as BOLD activity if they had.
The change in response seems to make sense given the anatomical data about thalamo-cortical connections and cortico-cortical connections (if we're assuming synaptic activity is the main BOLD correlate). Don't know about rat barrel cortex, but in cat V1 (area 17), <10% synaptic inputs to layer 4 are from the thalamus ( http://www.jneurosci.org/content/24/39/8441.long , http://www.ncbi.nlm.nih.gov/pubmed/19632814 ). I suspect if you repeated the experiment and somehow only stimulated e.g. hippocampus, which has a different connectivity profile, you would get different results.
Something else to consider, energy is not only required by neurons but also by astrocytes for instance which form part of the neurovascular coupling. Astrocytes by themselves account for a good portion of energy consumption by the brain--one that is not necessarily correlated with neurones. Therefore, regions of activation are not only areas of neuronal activity but also astrocytic activity.
It might be worthwhile to mention here what 'blobs' do represent, most explicitly: model fit. Researchers predict a pattern of activation based on the time course of cognitive events (i.e. their model), and the blobs represent where their model was most consistent with the fluctuations in fMRI signal.
Richard: Right, it fits with what you'd expect from the anatomy.
Interestingly though this suggests that the source of the BOLD signal in areas which mainly receive & relay input i.e. subcortical structures, could be different from in the cortex.
I.e. I believe that most synapses in the thalamus are the end of inputs from either below or above (cortico-thalamic descending projections) - not sure what the % is but I don't think the thalamus "talks to itself" in the same way that the cortex does - if so, the BOLD signal in the thalamus might be a mere echo, as it were, of activity elsewhere.
nice one, neuroskeptic! I usually end up having my masters students explain the Logothetis papers to me (again!) and this looks like a welcome addition to the syllabus.
what do you think happens when input to a region (globus p/Snpr) is mostly gabaergic & results in local inhibition? In this case, the thalamus could be an 'echo' (good word-choice there): patterns of activation/inhibition might be a good read-out of basal ganglia processing. Also do you then think that so-called BOLD 'de-activation' = local neuronal inactivation or is de-activation an artefact of subtraction?
Neuroskeptic: "Interestingly though this suggests that the source of the BOLD signal in areas which mainly receive & relay input i.e. subcortical structures, could be different from in the cortex."
Indeed. I guess this would be the next thing to test.
Khalil A.: true; not sure how well correlated glial activity and synaptic activity are, but I'm guessing reasonably strongly as they're the cells clearing up the ions and neurotransmitters as well as helping to supply the oxygen from the blood stream.
Hi Lavana, don't forget that the baseline we're measuring against is often ~80% of ceiling, so getting deactivation from reduced activity is quite straightforward. Indeed, Bojana Stefanovich and Bruce Pike showed several years ago what I think is the simplest and most elegant demonstration of this effect: contrast left-hand grip (right motor activation) against right-hand grip (left M1) and rest. Activating left M1 causes right M1 to deactivate. It must be due to inhibition, causing reduced activation. There are some more recent publications showing similar deactivation effects in other brain regions, especially visual areas. (Disclaimer: I'm on one of them!)
As to the bigger question of what exactly is in the neurovascular transfer function, I'm staying out of it!! Too many variables and too much biology for my wee noggin! Though I do agree Khalil A. that astrocytes are important. Lactate shuttling is an elegant suggestion (Pelegrin and Magistretti for brain metab., I believe), as it neatly explains the mismatch between glycolitic and oxygenation rates that causes the conc. of Hb02 to increase in venous blood following "activations." (Perhaps we should consider all "activations" and "deactivations" in quotes! They all need a baseline to be meaningful!)
Post a Comment