fMRI brain scans without the new acceleration techniques (top row) and with increasing numbers of multiplexings and slice accelerations. The bottom row was obtained seven times faster than the top row, although all show similar resolution. Only 4 of the 60 slices of a full, 3D brain scan are shown. (David Feinberg/UC Berkeley) |
An international team of physicists and neuroscientists has
reported a breakthrough in magnetic resonance imaging that allows brain scans
more than seven times faster than currently possible.
In a paper in PLoS ONE, a Univ.
of California, Berkeley,
physicist and colleagues from the Univ.
of Minnesota and Oxford Univ.
describe two improvements that allow full three-dimensional brain scans in less
than half a second, instead of the typical 2 to 3 seconds.
“When we made the first images, it was unbelievable how
fast we were going,” said first author David Feinberg, a physicist and
adjunct professor in UC Berkeley’s Helen Wills Neuroscience Institute and
president of the company Advanced MRI Technologies in Sebastopol, Calif.
“It was like stepping out of a prop plane into a jet plane. It was that
magnitude of difference.”
For neuroscience, in particular, fast scans are critical for
capturing the dynamic activity in the brain.
“When a functional MRI study of the brain is performed,
about 30 to 60 images covering the entire 3D brain are repeated hundreds of
times like the frames of a movie but, with fMRI, a 3D movie,” Feinberg said.
“By multiplexing the image acquisition for higher speed, a higher frame
rate is achieved for more information in a shorter period of time.”
“The brain is a moving target, so the more refined you
can sample this activity, the better understanding we will have of the real
dynamics of what’s going on here,” added Dr. Marc Raichle, a professor of
radiology, neurology, neurobiology, biomedical engineering and psychology at
Washington Univ. in St. Louis who has followed Feinberg’s work.
Because the technique works on all modern MRI scanners, the
impact of the ultra-fast imaging technique will be immediate and widespread at
research institutions worldwide, Feinberg said. In addition to broadly
advancing the field of neural-imaging, the discovery will have an immediate
impact on the Human Connectome Project, funded last year by the National
Institutes of Health (NIH) to map the connections of the human brain through
functional MRI (fMRI) and structural MRI scans of 1,200 healthy adults.
“At the time we submitted our grant proposal for the
Human Connectome Project, we had aspirations of acquiring better quality data
from our study participants, so this discovery is a tremendous step in helping
us accomplish the goals of the project,” said Dr. David Van Essen, a
neurobiologist at Washington Univ. and co-leader of the project. “It’s
vital that we get the highest quality imaging data possible, so we can infer
accurately the brain’s circuitry—how connections are established, and how they
perform.”
The faster scans are made possible by combining two
technical improvements invented in the past decade that separately boosted
scanning speeds two to four times over what was already the fastest MRI
technique, echo planar imaging (EPI). Physical limitations of each method
prevented further speed improvements, “but together their image
accelerations are multiplied,” Feinberg said. The team can now obtain
brain scans substantially faster than the time reductions reported in their
paper and many times faster than the capabilities of today’s machines.
Probing the brain with radio waves
Magnetic resonance imaging works by using a magnetic field and radio
waves to probe the environment of hydrogen atoms in water molecules in the
body. Because hydrogen atoms in blood, for example, respond differently than
atoms in bone or tissue, computers can reconstruct the body’s interior
landscape without the use of penetrating X-rays.
The new technique accelerates diffusion MRI as well. The colored tracks show the direction of nerve fiber bundles, providing a 3D image of the axonal pathways in the white matter of a resting human brain. A normal structural cross sectional image of the brain bisects the diffusion 3D fibertrack image. The entire 3D image was scanned in 8.5 minutes instead of 30 minutes. (David Feinberg) |
Nearly 20 years ago, however, a new type of MRI called
functional MRI (fMRI) was developed to highlight areas of the brain using
oxygen, and thus presumably engaged in neuronal activity, such as thinking.
Using echo planar imaging (EPI), fMRI vividly distinguishes oxygenated blood
funneling into working areas of the brain from deoxygenated blood in less
active areas.
As with standard MRI, fMRI machines create magnetic fields
that vary slightly throughout the brain, providing a different magnetic
environment for hydrogen atoms in different areas. The differing magnetic field
strengths make the spin of each hydrogen atom precess at different rates, so
that when a pulse of radio waves is focused on the head, the atoms respond
differently depending on location and on their particular environment. Those
that absorb radio energy and then release the energy are detected by magnetic
coils surrounding the head, and these signals, or “echoes,” are used
to produce an image of the brain.
With EPI, a single pulse of radio waves is used to excite
the hydrogen atoms, but the magnetic fields are rapidly reversed several times
to elicit about 50 to 100 echoes before the atoms settle down. The multiple
echoes provide a high-resolution picture of the brain.
In 2002, Feinberg proposed using a sequence of two radio
pulses to obtain two times the information in the same amount of time. Dubbed
simultaneous image refocusing (SIR) EPI, it has proved useful in fMRI and for 3D
imaging of neuronal axonal fiber tracks, though the improvement in scanning
speed is limited because with a train of more than four times as many echoes,
the signal decays and the image resolution drops.
Another acceleration improvement, multiband excitation of
several slices using multiple coil detection, was proposed in the U.K. at about
the same time by David Larkmann for spinal imaging. The technique was recently
used for fMRI by Steen Moeller and colleagues at the Univ. of Minnesota.
This technique, too, had limitations, primarily because the multiple coils are
relatively widely spaced and cannot differentiate very closely spaced images.
In collaboration with Essa Yacoub, senior author on the
paper, and Kamil Ugurbil, director of the Univ. of Minnesota’s
Center for Magnetic Resonance Research and co-leader of the Human Connectome
Project, Feinberg combined these techniques to get significantly greater
acceleration than either technique alone while maintaining the same image
resolution.
“With the two methods multiplexed, 10, 12 or 16 images
the product of their two acceleration factors were read out in one echo train
instead of one image,” Feinberg said.
fMRI moving closer to speed of EEG
The ability to scan the brain in under 400 msec moves fMRI closer to
electroencephalography (EEG) for capturing very rapid sequences of events in
the brain.
“Other techniques which capture signals derived from
neuronal activity, EEG or MEG, have much higher temporal resolution; hundred
microsecond neuronal changes. But MRI has always been very slow, with 2 second
temporal resolution,” Feinberg said. “Now MRI is getting down to a
few hundred milliseconds to scan the entire brain, and we are beginning to see
neuronal network dynamics with the high spatial resolution of MRI.”
The development will impact general fMRI as well as
diffusion imaging of axonal fibers in the brain, both of which are needed to
achieve the main goal of the Human Connectome Project. Diffusion imaging
reveals the axonal fiber networks that are the main nerve connections between
areas of the brain, while fMRI shows which areas of the brain are functionally
connected, that is, which areas are active together or sequentially during
various activities.
“While it simply is not possible to show the billions
of synaptic connections in the live human brain, the hope is that understanding
patterns of how the normal brain is functionally interacting and structurally
connected will lead to insights about diseases that involve miswiring in the
brain,” Feinberg said.
“We suspect several neurologic and psychiatric
disorders, such as autism and schizophrenia, could be brain connectivity
disorders, but we don’t know what normal connectivity is,” Feinberg added.
“Although the fMRI and neuronal fiber images do not have the resolution of
an electron microscope, the MRI derived Connectome reveals the live human brain
and can be combined with genetic and environmental information to identify
individual differences in brain circuitry.”
Raichle, a collaborator in the NIH Human Connectome project,
is one of the pioneers of “resting state” MRI, in which brain scans
are taken of patients not involved in any specific task. He believes that the
ongoing spontaneous activity discovered during such scans will tell us about
how the brain remains flexible and maintains a degree of homeostatis so that
“you know who you are.”
“Being able to sample this ongoing activity at
increasing temporal fidelity and precision becomes really important for
understanding how the brain is doing this,” Raichle said. “David is
superclever at this kind of technical stuff, and I have been cheering him
along, saying that the faster we can go, the better we can understand the
brain’s spontaneous activity.”