Making Sense of fMRI

Blogging on Peer-Reviewed ResearchFunctional Magnetic Resonance Imaging (fMRI) studies have rapidly become the bread and butter of cognitive neuroscience research. They are equally popular with science journalists who, with the zeal of phrenologists past, gleefully report that “Optimism Lights Up a Brain Part” or “Brain Circuits That Control Hunger Identified.” The only problem with these studies is that nobody knows what the fMRI signal actually means.

Well, that’s an exaggeration. We do know that fMRI signal reflects regional fluctuations in blood oxygenation, but we have yet to figure out what this metabolic activity means in terms of neural activity. In a recent Nature Neuroscience paper, Ahalya Viswanathan and Ralph Freeman make substantial progress toward answering this question.

Increased oxygen consumption might reflect neural inputs (e.g. synaptic activity at dendrites) or it might reflect neural outputs (e.g. action potentials at axons). Viswanathan et al. used a creative experimental design to distinguish between these possibilities. They used dual microelectrodes to simultaneously measure local oxygen levels, spiking activity and local EEG signal (or field potentials) in the cat visual cortex. They found that oxygen levels correlated most closely with the local EEG signal, which is known to reflect dendritic inputs rather than axonal outputs. For a review of this connection please check out my Method of the Month feature.

One challenge in determining whether tissue oxygenation reflects dendritic or axonal activity is that they usually go hand in hand. If a neuron is receiving a flurry of synaptic inputs, it usually starts firing action potentials as well. Viswanathan et al. sidestepped this problem by taking advantage of the fact that the visual system is composed of modules that are arranged in series. Moreover, these modules respond differently to the same visual stimuli. In this study, the experimenters used a drifting pattern of black and white lines to stimulate the visual system. When the grating drifts slowly (e.g. 4 Hz), the visual stimulus activates the thalamus, which goes on to activate the visual cortex. When the grating drifts quickly (e.g. 30 Hz), the stimulus activates the thalamus, but thalamic projections fail to activate the visual cortex. While the cortex receives inputs corresponding to the high frequency visual stimulus, it does not fire action potentials (or spikes) in response.

By recording from the visual cortex, Viswanathan et al. were able to show that both frequencies of visual stimuli resulted in local field potentials and changes in tissue oxygenation, even though only the low fequency stimuli resulted spiking activity. This dissociation strongly suggests that fMRI signals, like EEG signals, reflect synaptic inputs rather than axonal outputs.

Here is one of the most important figures from the paper, which pretty much sums up their findings:


On the left, the dark blue histograms show spiking activity in response to visual stimulus (shown by square wave at top) while the waveform shows fluctuations in tissue oxygenation. Note the initial dip in oxygenation followed by a subsequent increase, a response pattern that is typical of regional fMRI signals. On the right, the red areas correspond to increased EEG signal upon stimulus presentation. The 30 Hz signal alters blood oxygenation and local field potentials without resulting in action potentials.

This study clarifies one of the major questions obscuring the accurate interpretation of fMRI data. Neural inputs turn out to matter more than neural outputs. Unfortunately, it doesn’t clear up the whole mystery surrounding imaging data. For instance, if the cingulate cortex lights up during optimistic thoughts, we still don’t know the source of those synaptic inputs. Most brain regions integrate inputs from disparate brain regions, and it looks like the fMRI signal won’t be able to distinguish among these.

Nonetheless, these studies are a step in the right direction: away from neo-phrenology and toward a functional understanding of fMRI data.


Viswanathan, A. and Freeman, R.D. Neurometabolic coupling in cerebral cortex reflects synaptic more than spiking activity. Nat. Neurosci. 10, 1308-12 (2007)

One response to “Making Sense of fMRI

  1. While this paper refines our understanding of the coupling between the BOLD response and neural activity, any discussion of this relationship should really begin with this Nature article, Neurophysiological investigation of the basis of the fMRI signal.

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 )

Twitter picture

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

Facebook photo

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

Google+ photo

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

Connecting to %s