Functional magnetic resonance imaging (fMRI) has the enormous potential to pinpoint neuronal damage in almost any part of the brain and could play an important part in the fight against Alzheimer disease (AD). But the technique remains slightly controversial because it doesn’t directly measure neuronal activity. Instead, it relies upon small changes in blood oxygen, triggered by respiratory surges in nearby neurons (see ARF related news story). Now, thanks in part to the 1967 Sergio Leone Western, The Good the Bad and the Ugly, critics of blood oxygenated level-dependent, or BOLD MRI, may have been robbed of some ammunition. Researchers in Israel and the US have found that neuronal firing elicited in people watching the movie can accurately predict the MRI signal that will be obtained from others watching the same clip. Also, researchers in Germany report that BOLD MRI signals are driven by synchronized neuronal firing. The findings may be particularly interesting to AD researchers because these results suggest that the technique is well-suited to measuring higher cognitive and executive function.
One of the problems with BOLD MRI is that it may be measuring synaptic activity that does not result in any action potentials. In other words, it may be measuring every time the gun is drawn, rather than every time it is fired. This would obviously make interpretation of BOLD MRI signals a little tricky. But lest you outlaw the technique completely, take a gander at the two papers in tomorrow’s Science. They report that hemodynamic, or BOLD MRI signals, correlate tightly with neuronal firing.
In the first, Rafael Malach and colleagues at the Weizmann Institute of Science, Rehovot, and the Sackler School of Medicine, Tel Aviv, together with collaborators at New York University and the University of California, Los Angeles, compared fMRI signals with neuronal action potentials from the auditory cortex in humans.
First author Roy Mukamel and colleagues elicited the help of two epilepsy patients who were being monitored with intracranial electrodes. This allowed the authors to directly record spike, or action potentials, from 53 neurons in the Heschl’s gyrus of the auditory cortex as the two patients watched a 9-minute clip from the Western. The spike potentials recorded were then used to build a “spike predictor,” which converts the firing spikes into a predicted fMRI signal. By comparing this predicted BOLD MRI to actual scans taken from 11 healthy volunteers as they watched the same movie clip, Mukamel and colleagues found that the predicted and actual MRI scans had almost a one-to-one correlation.
The correlation between the predicted fMRI from the first epilepsy patient and the averaged fMRI scans from the 11 volunteers was about 0.75 for the entire duration of the recording. But there was a 3-minute section in the middle of the clip that seemed to particularly grip all the volunteers, because during these scenes the individual fMRI scans were highly matched. Here, the correlation between the predicted and averaged fMRI scans was 0.9, and the value was highly significant, statistically. The correlation between the predicted fMRI signals obtained from the second epilepsy patient were poorer, however, only 0.56 for the entire 9-minute scan and 0.75 for the 3-minute section in the middle. But the predicted fMRI scans for the two patients did not entirely match either, probably because the electrodes were not placed in exactly the same locations, which may explain the weaker correlation obtained from the second patient.
The authors also found that there was significant correlation between the BOLD fMRI signal and local frequency potentials (LFPs), which are thought to be related to synaptic activity in the dendrites rather than the axons, in other words, signals coming into the neurons rather than those going out. The correlation was highly positive, reaching as high as 0.9 (at frequencies between 30 and 300 Hz). “Our results indicate a high linear correlation between spiking activity, high-frequency LFP, and fMRI BOLD signal measured in human auditory cortex during natural stimulation,” write the authors. While their method cannot identify whether the spike potential or the LFPs were the driving force for the BOLD signals, nevertheless, they conclude, based on the high correlation, that “BOLD fMRI signals can be trusted as a faithful measure of the average firing rate of the underlying [neuronal] population.”
In the second paper, Ralf Galuske and colleagues at the Max Planck Institute for Brain Research, Frankfurt, and the Technical University of Darmstadt did manage to pinpoint the driving force. First author Jörn Niessing and colleagues found that there is a direct relationship between the average spike rate and the BOLD fMRI signal in the visual cortex of the cat. However, under constant stimulus, the MRI signal fluctuated, and though only loosely correlating with action potentials, these fluctuations were tightly correlated to LFPs in the 50-90 Hz range, commonly called gamma oscillations. The data would suggest that BOLD signals best reflect input via dendrites rather than output via action potentials, bringing us back to one of the criticisms of the technique.
But not so fast there, pardner. If you are a critic, you might have to holster that six-shooter. Niessing and colleagues also report that it is the precise synchronicity of neuronal firing that determines the scale of the BOLD fMRI response. Because the MRI signal most tightly correlates with LFPs in the gamma range, which only occur when neuronal firing is most synchronized, the MRI signal does indeed reflect output as well as input.
But in an interesting twist, the authors also speculate that what drives the BOLD MRI signal is inhibitory interneurons, which only represent about 20 percent of cortical neurons. This is because there is evidence to suggest that gamma frequency oscillations, which are driving the MRI signal, are due to synchronized discharges from this particular subpopulation (see Traub et al., 1996 ).
So what does this mean for the study of AD and other neurodegenerative diseases? Well, oscillations in the gamma frequency range are associated with many cognitive and executive functions, such as short-term memory, sensory-motor coordination, and movement preparation (see review by Engel et al., 2001). So as Niessing and colleagues write, “hemodynamic responses may thus be ideally suited to visualize neural processes associated with higher cognitive and executive functions.” Notch one up for the fMRI proponents.—Tom Fagan
- Traub RD, Whittington MA, Stanford IM, Jefferys JG. A mechanism for generation of long-range synchronous fast oscillations in the cortex. Nature. 1996 Oct 17;383(6601):621-4. PubMed.
- Engel AK, Fries P, Singer W. Dynamic predictions: oscillations and synchrony in top-down processing. Nat Rev Neurosci. 2001 Oct;2(10):704-16. PubMed.
No Available Further Reading
- Mukamel R, Gelbard H, Arieli A, Hasson U, Fried I, Malach R. Coupling between neuronal firing, field potentials, and FMRI in human auditory cortex. Science. 2005 Aug 5;309(5736):951-4. PubMed.
- Niessing J, Ebisch B, Schmidt KE, Niessing M, Singer W, Galuske RA. Hemodynamic signals correlate tightly with synchronized gamma oscillations. Science. 2005 Aug 5;309(5736):948-51. PubMed.