Medical television shows sometimes depict thoughts skipping across the brain as action potentials that ignite like exploding stars. While it looks dramatic and impressive, today’s brain imaging technologies can’t visualize brain activity so sensitively. A new magnetic resonance imaging (MRI) technique called DIANA – direct imaging of neuronal activity – may get us closer, though.
An alternative to BOLD fMRI
A brain signal begins with an action potential caused by rapid changes in voltage across cellular membranes. Researchers involved in this proof-of-concept study, reported in Science, say that DIANA might measure this neuronal activity by capturing the intracellular voltage of a group of neurons.
The technique could fill a gap in brain imaging. Functional MRI (fMRI), for example, allows us to view neuronal activity in the brain non-invasively by measuring a surrogate – fluctuations in the blood oxygenation level-dependent (BOLD) signal that result from changes in neuronal activity. But the temporal specificity of fMRI is too slow (on the order of 100 ms) to follow neuron activation during cognitive processes. Alternatives such as electroencephalography (EEG) and magnetoencephalography (MEG) have higher temporal resolution, but spatial resolution that’s limited to the centimetre range.
DIANA can achieve both temporal and spatial resolutions on the order of cognitive processes (5 ms and 0.22 mm, respectively) by isolating groups of neuronal signals with a distinct data acquisition scheme, explains senior author Jang-Yeon Park.
Park, a professor of biomedical engineering at Sungkyunkwan University in Korea, was inspired by a study in Nature Methods that described how data can be split to acquire images in pieces for BOLD fMRI. After mulling over the theory, Park says he realized that the method could be adopted to ultrahigh-temporal resolution MRI.
Acquiring images in pieces
Data from MR signals are stored in a temporary image space called k-space. Information about image contrast is held at the centre of k-space, which contains low-spatial frequency information, while image details are held at the high-spatial frequency edges of k-space. Over the course of an MRI scan, k-space is filled. At the end of a scan, k-space is full, and data are reconstructed to produce an image.
The most common way to fill k-space is to acquire data line-by-line and collect a series of complete images to follow a signal through time and space. DIANA, on the other hand, collects a series of partial images using a 2D fast line-scan approach. Here, single lines of k-space are acquired repeatedly between the intervals of a repeated stimulus, and different k-space lines are acquired in different stimulus periods. This way, each stimulus period adds one line of k-space to all the time-series images within the period.
DIANA requires no contrast agents or new equipment. Neural activation can be imaged on an ultrahigh-field scanner using a conventional 2D gradient-echo imaging sequence with a short echo time and short repetition time in a line-scan acquisition scheme.
To produce repeatable neural action potentials, the researchers repeatedly flicked the whiskers of anesthetized mice, a technique common to brain imaging studies. They observed that in response to the stimulus, neurons in the mouse somatosensory cortex were activated after deep regions, such as the thalamus, were activated. Imaging this level of neuronal activity, the researchers say, could help us understand communication between areas of the brain in the future.
“If it works in the human body, it can be a game-changer for neuroscience, because if we detect neuroactivation directly at high temporal, high spatial resolution, then I think we can really start looking at the brain network as a neural network in time and space,” says Park. “Using BOLD fMRI…it is very difficult to look at the dynamic neural network and it is also difficult to explore the hierarchical functional connectivity in the neural network.”
T2 to the future
The biophysical source of the DIANA signal isn’t clear, but the researchers think they have a strong hypothesis supported by additional experiments and simulations – that changes in neuronal membrane potential are reflected in a positive correlation with the transverse relaxation time (T2) of the MRI signal, which determines how quickly MRI signal disappears.
Such neural activation signals haven’t been observed in BOLD fMRI experiments because they capture haemodynamic responses on the order of several seconds rather than milliseconds, Park explains.
He and his group still take inspiration from fMRI experiments, however, as they look to future work. BOLD fMRI measures signal in the absence of stimulus using resting state techniques, and sensitivity is increased by processing data using a neuronal response function. The team hope to develop analogous techniques for DIANA.
Ultrasound technique captures micron-scale images of brain activity
Park says that his research group is using DIANA to study visual neural networks in mice and how the neural network changes in animal models of neurodegenerative disease. They are also translating the technique to use in human studies, where motion artefacts and trial-by-trial variability may be prevalent and activation patterns may span multiple time scales.
“Neuroscientists and people working on neuroimaging studies, they want to understand the brain and how it works,” Park says. “I think if we understand the real dynamic neural network, we can really start understanding how the brain works, as well.”