New MRI Technique Tracks Brain Activity at Millisecond Timescales

ABOVE: Images taken at 5 millisecond intervals (running in rows from top left to bottom right) show the propagation of signals in an anesthetized mouse's brain—in this case, in response to a light being shone in the mouse's eye. P.T. TOI ET AL., SCIENCE, 2022.

A new approach to magnetic resonance imaging could allow neuroscientists to noninvasively track the propagation of brain signals on millisecond timescales, according to a study published yesterday (October 13) in Science.

The technique, which its creators call “direct imaging of neuronal activity” (DIANA), uses existing magnetic resonance imaging (MRI) technology to take series of quickfire, partial images, and then combines those images to create a high-resolution picture of which bits of the brain are active when.

DIANA has so far only been tested in anesthetized mice, and the mechanisms underlying it aren’t entirely clear, notes Matthew Self, a neuroscientist at the Netherlands Institute for Neuroscience who wasn’t involved in the work. But provided it can be replicated in other labs, the method could represent a “major advance” in brain imaging, he says.

“This would be the first technique which would be able to noninvasively measure neural activity with both a high spatial and temporal resolution,” explains Self. “I’m definitely very keen to try it.”

Researchers who spoke to The Scientist say they are already enthusiastic about the new technique’s potential.

MRI technology uses magnetic fields and radio waves to produce detailed images of tissue. Its use relies on the fact that different materials have distinct magnetic properties, allowing a scanner to distinguish between different tissues or to monitor changes in tissue over time.

Researchers have long used one version of this technology, known as blood oxygen level–dependent functional MRI (BOLD fMRI), to study how the human brain works. This method detects changes in blood flow to particular regions of the brain as a proxy for neuronal activity.

BOLD fMRI can pinpoint activity to a millimeter or less of brain tissue. But the technique’s temporal resolution is less impressive. Changes in blood flow occur over seconds—far slower than the millisecond timescale of neuronal signals. Images from fMRI will often show a whole neural pathway active all at once, when really, there’s a neural signal propagating from one part of the pathway to the next.

See “Which Neurons Go to Sleep First in Humans? fMRI Can Tell”

Other noninvasive techniques that directly measure electrical activity, such as electroencephalography (EEG) and magnetoencephalography (MEG), are much better at pinpointing the timing of neuronal firing, but much worse when it comes to spatial resolution.

In the new study, Jang-Yeon Park, a biomedical engineer at Sungkyunkwan University in South Korea, and his colleagues came up with a novel way to tackle the problem. Rather than taking full images of a particular cross-section of the brain every few seconds, as in conventional fMRI, he and his colleagues set their MRI equipment so that it would gather sequences of much smaller, partial images at very short intervals—just a few milliseconds apart. They’d then be able to stitch together these partial images to get a full view of that brain cross-section at each timepoint.

To see if they could identify any signal of brain activity with this approach, the researchers popped anesthetized mice into the MRI scanner, then lightly zapped the animals’ whisker pads with an electric current. They found that the images produced by their technique were registering some sort of signal in the somatosensory cortex—the bit of the mouse brain that senses whisker stimulation—in the 25 milliseconds or so after the zap.

Exploring this further, they discovered that the “DIANA signal” actually moved around over time. It appeared in a brain region called the thalamus about 10 milliseconds after the whisker zap, moved to one section of the somatosensory cortex at around the 25 millisecond second mark, and then sprung up in another part of the somatosensory cortex a few milliseconds later.

A bigger puzzle is what DIANA is detecting, exactly.

By taking measurements of the same brain area with invasive techniques such as electrophysiology and optogenetics, the team showed that their DIANA signal was in fact tracing the propagation of neuronal activity in response to whisker stimulation.

Peter Bandettini, a neuroscientist and physicist at the National Institute of Mental Health (NIMH) who was not involved in the study, calls the team’s work “incredibly convincing.” Several teams have attempted to boost MRI’s temporal resolution before, he adds, but few have gone to such lengths to bolster their claims. The paper included a “tour de force of experimentation” to show that the technique was indeed tracking the propagation of neuronal signals.

Park tells The Scientist that he isn’t sure why researchers haven’t reported this effect before, given that it doesn’t require particularly special equipment, but says it’s likely that people just didn’t think to create images in this way. Bandettini notes that hacking an MRI machine to take rapid partial images like DIANA requires a significant amount of expertise—and a belief that it might turn up something interesting.

A bigger puzzle is what DIANA is detecting, exactly. Park and colleagues show in their study that BOLD effects are unlikely to be responsible, and suggest instead that their method is registering changes in the membrane potential of firing neurons, perhaps via fluctuations in the amount of water on the membrane’s surface or via cell swelling. That’s a possibility, says Self, but overall, “the mechanism isn’t super clear. . . . I think that needs to be demonstrated in future studies.”

In its current form, DIANA has a few limitations, notes Park, who is named as co-inventor on a patent related to the method. Due to the way it stiches together full-brain snapshots by combining partial images taken at different times, the technique is likely to be susceptible to so-called motion artifacts—disruptions caused by the animal moving its head between takes. That could present some challenges in translating DIANA to awake animals or people.

The signal that DIANA is picking up is also relatively weak—around an order of magnitude smaller than that in BOLD fMRI, notes Bandettini. Groups would need relatively sophisticated MRI equipment to be able to mimic the team’s approach, as well as an experimental protocol that included repeat tasks or stimulations to allow for averaging results from multiple scans, he says. “You need a lot of repetitions of the same thing, and very, very precise resolution.”

However, researchers who spoke to The Scientist say they are already enthusiastic about the new technique’s potential. Bandettini points to another of the team’s experiments that suggests DIANA may be able to distinguish between excitatory and inhibitory neuronal signals—something that is challenging even with invasive techniques such as electrophysiology. “That’s super exciting. That would open up an entire area of understanding how the brain interacts.”

Self, who studies visual processing, says that he knows of several groups already trying to put DIANA to work in people. Although the technique still can’t offer the single-cell resolution achievable with some invasive technologies, there are wide-reaching implications if it works in other labs, he says. “In principle, it could be taken into humans, it could be taken perhaps even into patient studies—it could open up a whole world of research into understanding the brain in health and disease.”