Researchers tweak fMRIs to map the brain's wiring schematic

For many years, tracking the activity of nerve cells involved inserting electrodes directly into neural tissue and painstakingly reading the tiny changes in voltage associated with their activity. In recent years, however, a technique called functional MRI (fMRI) has proved popular with researchers and the public alike. Much of that popularity is because fMRI promises the ability to track activity across the entire brain, identifying the regions that light up as people recognize objects or contemplate financial decisions. In the popular imagination, fMRI was often viewed as one tiny step away from mind reading. Despite its successes and the public's apparent trust, however, there was one small problem: nobody has been entirely certain what the MRI machines were actually reading.
Thanks to a paper released over the weekend by Nature, that has changed. Researchers were able to limit the firing of nerve cells to a specific individual type, and show that these triggered normal-looking fMRI signals in rats. Not only does this place fMRI on a firmer empirical footing, the technique allowed the researchers to track networks of connected nerves within the brain.
Functional MRI is based on a variety of assumptions, most of which have empirical support. Neural activity requires rapid changes in the voltage at the surface of nerve cells, enabled by both the release and uptake of ions from the surroundings. The process of getting ready to fire off another pulse requires that the cells draw in ions against the local concentration gradient, which takes a fair bit of energy. Put this all together, and there's an inescapable conclusion: when an area of the brain is very active, it needs an extra dose of the body's standard energy carrier, glucose.
To keep busy nerves supplied, the body needs to increase the blood flowing to the area. This change in blood flow is what fMRI is intended to pick up—the technique was formally called BOLD, since it involves the search for blood oxygenation level-dependent signals.
All of that is pretty straightforward. The problem is that researchers haven't been quite certain as to whether any specific type of neural activity is more or less likely to produce a BOLD signal. "Candidate circuit elements for triggering various kinds of BOLD signals include excitatory neurons, mixed neuronal populations, astroglia, and axonal tracts or fibres of passage," the authors write. "Importantly, it is not clear which kinds of activity are capable of triggering BOLD responses." A few attempts had been made to look into this, using electrodes that stimulated small regions of the brain, but these necessarily activated every nerve cell in the area, so they were a rather crude test.

Engineering active nerves

The new work gets around this problem by using a virus that has been engineered to express a bacterial protein called channelrhodopsin. (We've previously discussed the use of channelrhodopsin to engineer fake memories.) When exposed to light, channelrhodopsin allows ions to pass through the cell membrane, which mimics the normal firing of nerve impulses. The virus was engineered to ensure that only a specific type of nerve cell expressed it—excitatory nerves that rely on calcium ions, for example. By combining that restriction with a localized injection of the virus and local exposure to light, the researchers obtained very fine control over the activation of nerve cells.
The biggest hurdle to the work seems to have involved setting up the apparatus that exposes the nerve cells to light inside a rat-sized MRI machine, and ensuring that none of its components would respond to the large magnetic fields involved.
With everything in place, the researchers confirmed that firing an impulse in excitatory neurons produced a signal that matched nicely with the ones observed during regular experiments. Putting the channelrhodopsin into inhibitory neurons produced a small BOLD signal in the area where the light triggered an impulse, but it was surrounded by a halo of depressed activity, consistent with the neurons' inhibitory role.
But the BOLD signals weren't limited to the area where the light triggered activity. With a slight delay, signals started showing up in other areas of the brain, with the precise locations changing based on where exactly the activity was triggered. The authors indicate that these additional signals provide an indication of the brain's wiring—the nerves at the site of the initial activity were simply doing what they normally did, and communicating with other areas of the brain. With enough time, they suggest, their technique could be used to map functional connections throughout the brain.
The new work goes a long way towards providing a solid empirical footing for fMRI, and further use of it will improve matters, as the patterns of activity induced by different types of nerve cells are mapped. And that will provide a better understanding of what researchers have been looking at in the multitude of fMRI studies that have already been published.
What's perhaps most striking about this is how willing the public has been to accept fMRI despite what might generally be viewed as a glaring hole in the science behind it. There are consulting services that will fine tune political and product advertisements based on fMRI results, and lawyers have recently tried to introduce fMRI data in a civil suit (that request was denied). It would seem to provide further evidence that the scientific uncertainties the public chooses to tolerate are chosen very subjectively.