Skip to Main Content

Research often proceeds in a logical progression, new studies building upon a detailed understanding of the underlying processes revealed by earlier work. But a new brain imaging technique that can directly track the activity of neurons emerged from one of academics’ favorite questions: “What would happen if we tried doing it this way?”

The research, published Thursday in Science, describes a technique that promises to overcome limitations of current imaging methods such as functional MRI (fMRI), which has been a mainstay of neuroscience research for three decades. By more precisely pinpointing neuronal activity in space and time, the technique could let scientists learn more about neurological diseases like Parkinson’s and Alzheimer’s, the physiological processes underlying mental health disorders, and the workings of biological neural networks that inspired artificial neural nets.

If the new technique pans out, it could one day augment or replace widely used clinical tools including EEG or MEG, which measure electrical or magnetic signals from the brain but have a resolution on the centimeter level, and research tools like fMRI, which has a response time 10 times slower than neurons fire.


There are several hurdles to clear before this proof-of-concept study of a new phenomenon — called direct imaging of neuronal activity (or DIANA for short) — could be translated to clinical research, but it may not be that big a leap, researchers said. While the experiments were performed in mice, the newly discovered effect is only a few software clicks different than existing MRI techniques. In theory, it should be easy to apply to humans since it arises more from physical interactions than biological ones — or so the researchers think.

The South Korea scientists used the new DIANA MRI technique to image brains of sedated mice as their whisker pads were stimulated. Senior author Jang-Yeon Park, an associate professor of biomedical engineering at Sungkyunkwan University, and his graduate student Phan Tan Toi showed that the signals were closely correlated to neuron activity by matching them to direct electrophysiological measurements collected by collaborators Hyun Jae Jang and Jeehyun Kwag at Korea University, using the same protocols.


The DIANA method can be implemented with typical MRI machinery, but researchers used a new way of measuring the data that is fast enough to detect the tiny DIANA signal. Normal MRI scans acquire the entire image in one scan, but that’s too slow and would start recording after this signal had already dissipated, which is why it hasn’t been detected before. The new method splits up the process, acquiring a single trace of the data at a time.

While the researchers are pretty certain that the DIANA effect is real, they don’t know exactly where it’s coming from, though they suggest some theories in the paper. “I have to say, this is a hypothesis, but it’s a strong hypothesis, I think,” said Park.

During his graduate school years at the University of Minnesota, Park sat in a lot of department neuroscience seminars, even though he wasn’t a neuroscientist. Many of these seminars detailed what could be done with a functional MRI technique called blood-oxygen-level dependent contrast, also known as BOLD fMRI. This method uses differences in blood oxygenation as a proxy for where neurons are firing and has been used to map brain activity for the last 30 years.

There’s always been an undercurrent of acknowledgement that differences in vasculature, which is what BOLD fMRI measures, are perhaps not truly reflective of what’s actually happening in the brain. The timing of the blood oxygenation response is also known to not be fast enough to capture the true neuron response.

As an MRI physicist, “I was amazed by the neuroscience work but at the same time I also had the big question, ‘Can you really figure out the secret of the brain function using this BOLD fMRI with the temporal resolution [problems]?’” said Park. But his research focus at the time wasn’t related to neuroimaging, so he shrugged the question off.

Years later, Park read a 2014 Nature Methods study where the authors split the data acquisition, allowing them to take the data more frequently, and were able to increase the temporal resolution to 40 to 50 milliseconds. “Oh my God,” he said. “Using this way, I might implement what I have thought in my mind: a millisecond-order [measurement].”

By repeatedly stimulating a sedated mouse’s whisker pad and imaging one-millimeter slices of its brain at a time, the researchers were able to pack in more timepoints than if they had acquired the entire image all at once. This let them detect signals that would otherwise have faded by the time they had measured them.

Noam Shemesh, an associate principal investigator at Champalimaud Research in Lisbon, Portugal, who was not involved with Park’s research, saw an early version of the study on the preprint server bioRxiv last year and has been trying to replicate the work in his lab — not because he doubts the measurement, but because the larger the claims, the more you need verification, he said.

“Once we [the field] can replicate this, I’m going to be the biggest believer in this method that’s ever lived,” he said. “I think it’s just a necessary step because this is a study which I think has the potential to — I don’t want to sound corny or anything — but it’s a paradigm-shifting study. It really is.”

Both Shemesh and Padmavathi Sundaram, an instructor at the Martinos Center for Biomedical Imaging at Massachusetts General Hospital who also was not involved in the study, said that the very short, five-millisecond timescales Park and colleagues looked at were possible only because of the multiple-scan “trick” they employed.

“It’s almost like you’re listening to the radio and every time you’re listening to the same song over and over. But then you just change through the spectrum,” said Sundaram, and listen to each frequency separately.

Sundaram had some concerns about outside fluids flowing into the one-millimeter slice and altering the signal, but she said the authors did everything they could to prove it was real. “This might be the first believable in vivo neuronal current MRI images I’ve seen,” she said.

Shemesh pointed out that the technique Park and colleagues used isn’t new; Alan Koretsky, author of the 2014 paper that inspired the new study, had also used a similar method in a 2002 PNAS paper, but Shemesh said that previous researchers weren’t “audacious” enough to look for these signals at the very short timescales that Park investigated, mostly because no one was expecting anything to be there.

“Where most of MRI people would look at noise, [Park and colleagues] were looking to see if there would be some signal. And they apparently found one,” he said.

One obstacle to using the method in humans is that “you can’t anesthetize [humans] to image them,” according to Sundaram. But Shemesh is confident that the challenges aren’t anything the field hasn’t overcome before, noting that this MRI method isn’t the most sensitive to subjects moving while the imaging is taking place, though that’s certainly a concern.

Experts were optimistic that if replicated, this new MRI method could open the door for completely new neuroscience — clinical and preclinical applications as well as basic research.

“A brain full of electrodes inside of it might not behave as a normal brain in many different ways. This opens up a way to completely non-invasively just put the person in the scanner. You look at the signals and you potentially get some of the electrophysiological response more directly,” said Shemesh. “It’s a big deal.”

Create a display name to comment

This name will appear with your comment