Two of the hottest areas in magnetic resonance imaging (MRI) are at the extremes.
At one end of the spectrum, research teams are working on huge, ultra-high-resolution MRI machines capable of performing precision brain imaging well beyond that available with the traditional 3 Tesla (3T) scanners typically used in large hospitals. These promise to provide new and wide-ranging detail that will advance neuroscience research, and potentially lead to clinical applications.
At the other end, scientists and engineers are developing portable, ultra-low power and low-cost MRI machines at around the 25–70 milliTesla (mT) range. These machines are designed to make the scanning technology more accessible by not only bringing MRI machines to the patient instead of the other way around, but also by introducing the far-less-expensive machines into resource-limited settings that are in dire need of MRI to aid in diagnostics, patient monitoring, and perhaps neurosurgical navigation.
Going bigger
Over the past year, ultra-high-power MRI machines have churned out stunning images that are already revealing new insights into how the brain works. One is a next-generation 7T (NexGen 7T) MRI scanner, the culmination of a multi-year, international collaboration directed by MRI physicist and neuroradiologist David Feinberg (Figure 1), Ph.D., M.D., professor of neuroscience at the University of California, Berkeley, and president of the technology company Advanced MRI Technologies of Sebastopol, California. Late last year, the collaborative group published a description of its new scanner along with the world’s highest detailed functional MRI (fMRI) images of the visual cortex [1], which is central to interpreting information relayed from the retinas. “We showed that we can see activity at different depths in the cortex and in the superficial cortical layers, including the direction of information being transmitted from one area of the brain to another (Figure 2),” Feinberg said. “These are interactions we couldn’t see before.”
Figure 1. David Feinberg, Ph.D., professor of neuroscience at the University of California, Berkeley, and president of the technology company Advanced MRI Technologies of Sebastopol, California, cuts a ribbon to inaugurate the NexGen 7T brain scanner, which can produce ultra-high-resolution fMRI scans. Feinberg was director of the multidisciplinary and international project. At his left is John Ngai, director of the NIH BRAIN Initiative, which provided the bulk of funding for the project, and Andreas Schneck from Siemens Healthineers, which was a key part of the international team. (Photo courtesy of Brandon Sánchez Mejia/UC Berkeley.)
Findings like this are vital to advancing neuroscience, Feinberg said. “By looking at the brain at this finer scale and doing experiments on a level more than 150 times smaller than conventional fMRI, we are anticipating a better understanding of how the brain is organized. This represents a whole new direction in neuroscience.”
Figure 2. These images contrast a human brain scan taken with the NexGen 7T MRI scanner at higher resolution (left), a standard 7T scanner (middle), and a standard 3T hospital scanner (right). With higher resolution, neuroscientists can more precisely localize signals (orange) in the brain to understand normal brain circuitry and the changes associated with brain disorders. (Photo courtesy of An (Joseph) Vu, UC San Francisco; and David Feinberg and Alex Beckett, UC Berkeley and Advanced MRI Technologies.)
One way to build a high-power MRI machine is to focus on an ever-larger main magnet. The main magnet creates a stable field around the patient and provides the baseline signal to generate an image. The French Alternative Energies and Atomic Energy Commission (CEA), for instance, has an 11.7T magnet at the core of its Iseult MRI scanner, which is housed at the CEA Paris-Saclay center [2]. In April 2024, CEA released strikingly sharp brain images of human subjects that were taken in just 4 minutes with the 11.7T scanner, noting that such a high-resolution image would have required an hours-long session in a typical 1.5–3T hospital MRI machine.
Feinberg’s concept was to use an ultra-high field 7T magnet, but incorporate major performance improvements to other subsystems of a head-only scanner so that it could precisely follow brain activity over time (Figures 3 and 4). Specifically, he planned to generate fMRI scans that could track neuronal interactions within different levels of the brain’s very thin outer layer, or cortex, where the neurons are located. Although fMRI cannot pick up neuronal signals directly, it is able to discern the small changes in blood oxygenation and blood volume that occur when neurons are activated. A standard 7T MRI provides average blood flow for all the veins and the capillaries throughout the depth of the cortex, but what they needed was higher spatial resolution, so they could see neuronal activity both temporally and spatially at different depths in the cortex, which is only 2–4 mm thick.
Figure 3. This cross-sectional diagram of the NexGen 7T scanner shows the new head-only gradient coil (green) and receiver-transmit coil (white) resting on a movable bed (brown) and connected to an electronic interface (blue) containing nearly a thousand wires (blue) that extend out of the magnet. (Photo courtesy of Siemens MRI Division, Erlangen Germany, Bernhard Gruber, and David Feinberg, UC Berkeley.)
Figure 4. This NexGen 7T fMRI scan can pinpoint active groups of neurons (blue arrows) in even the thinnest human brain cortex, which is between 1.5 and 2 millimeters thick. (Photo courtesy of David Feinberg and Alex Beckett, UC Berkeley and Advanced MRI Technologies; Renzo Huber, Maastricht University.)
With initial funding from the National Institutes of Health (NIH) Brain Research Through Advancing Innovative Neurotechnologies (BRAIN) Initiative, he pulled together a multidisciplinary team from the MRI industry, academia, and medical institutions to customize the design of a 7T scanner for higher resolution brain scans. They set their sights on several subsystems, notably the magnetic field gradient coil and the radiofrequency (RF) receiver array coils.
Higher resolution is limited by both the maximum gradient amplitude and how quickly the gradients switch, or pulse, on and off, Feinberg described, so with Siemens Healthineers MRI Division of Erlangen, Germany, they designed a coil to do the microsecond pulsing required to encode the signal spatially and pinpoint different depths in the cortex. He explained, “When you switch the gradients on and off, by Faraday’s Law, a current is induced in other pieces of metal as well as any kind of structure that has wiring in it, so we had to also deal with unpredictable interactions with other pieces of hardware.”
While that work was in progress, they concurrently upgraded the receiver array coils to obtain a stronger MRI signal and image faster. That meant adding small receiver loops, or channels, to the array. “A standard 7T scanner has 32 channels, but we ended up with 96 and 128 channels, something that had never been done with a 7T scanner,” Feinberg said. To fit that many additional channels, they then had to increase the inner diameter of the magnetic field gradient coil.
“It took a tremendous amount of problem-solving, including both unpredicted problems as well as synergetic gains, and the scanner is making the highest spatial resolution functional MRI images, and major improvements in structural imaging and diffusion imaging,” Feinberg said. As an example of the improvement in resolution, Feinberg described the NexGen 7T scanner as having an approximately 12 times higher isotropic, three-dimensional resolution than a standard 7T scanner in zoomed views. “That can be applied to study brain disorders,” he remarked.
Overall, the project took 6 years and $22 million (USD) in funding from the BRAIN Initiative, UC Berkeley and the Weill Neurohub, a West Coast research network centered on developing therapies for disorders and diseases of the brain and nervous system. Neuroscience research groups at UC Berkeley are already employing the scanner to answer research questions, and outside research groups will soon begin visiting campus to take advantage of the scanner’s capabilities.
In addition, the multinational corporation Siemens Healthineers, and MR CoilTech Ltd., of Glasgow, Scotland, will soon be disseminating additional NexGen 7T scanners to neuroscience imaging centers in coming year. “That was a goal: to disseminate the NexGen scanner to research centers around the world, so as many expert neuroscientists as possible can do as many experiments as possible,” Feinberg said.
He added, “I am very interested in seeing how research groups will use the NexGen 7T scanner to get a greater understanding of global interactions in the brain, test theories of how the brain is organized, and ultimately use that information for medical diagnosis and treatment. That’s what’s really exciting, and I’m looking forward to taking this scanner to that level right away.”
Going smaller
Ultra-low-power MRI scanners may not have the image quality higher-power scanners do, but they do have advantages, including far lower costs and portability.
On the cost side, a typical 1.5-3T clinical scanner in use in hospitals today can run into the millions of dollars (USD), partly because it needs a dedicated site, which includes a shielded scanning room and a control room, plus chiller systems to keep the machine at the proper and consistent temperature, and a massive power source to run it. That price tag is often out of reach for rural hospitals in industrialized countries, and medical facilities in resource-limited countries, said electrical engineer Stephen Ogier, Ph.D., a researcher in the U.S. National Institute of Standards and Technology (NIST) Magnetic Imaging Group (Figure 5). The group is currently putting its focus on ultra-low-field scanners and their possibilities for research and clinical applications.
Figure 5. Researchers in the U.S. National Institute of Standards and Technology (NIST) Magnetic Imaging Group include electrical engineers Kalina Jordanova, Ph.D., (left) and Stephen Ogier, Ph.D. Using a 64 mT scanner provided by Hyperfine, they are imaging white matter, gray matter, and CSF to determine their unique ultra-low-field MRI signals. (Photo courtesy of R. Jacobson/NIST.)
Access for patients is also an issue with larger scanners, said Michael Poole, Ph.D., vice president of engineering at Hyperfine Inc. of Guilford, Connecticut (Figure 6). Hyperfine has developed and received the U.S. Food and Drug Administration (FDA) approval for its portable 64mT MR system for brain imaging [3], which is called Swoop (Figure 7). “With a conventional MRI system, the patient is taken away from their site of care and care team, and transported to the MRI suite. This can be dangerous for critically ill patients and limiting for others who may have to travel to a specialized center to get an MRI.” As an example of an access issue, he noted that the tissue differentiation of MRI is significantly better at detecting ischemic stroke than is computed tomography (CT), but “most ERs only have ready access to CT for their suspected stroke patients, so if the CT is negative, the patient may end up waiting hours to get the MRI they need.”
Figure 6. Michael Poole, Ph.D., vice president of engineering at Hyperfine Inc., which has developed and received U.S. FDA approval for Swoop, its portable 64mT MR system for brain imaging. (Photo courtesy of Hyperfine Inc.)
Ultra-low-field MRI scanners solve those types of problems: They run a fraction of the cost of high-field scanners; some of the machines are portable and can perform scans in a patient’s room; and they can have such low power requirements that they need only plug into a standard electrical outlet, Ogier said.
Figure 7. Swoop is a portable unit that uses just 900 watts of power and plugs into a standard electric outlet. With artificial intelligence (AI)-powered, image-enhancing software and interference-canceling technology for noise-free images, it can produce clinically useful images of the brain in 15 minutes. (Photo courtesy of Hyperfine Inc.)
Nonetheless, ultra-low-power MRI shares the same underlying physics as its high-power counterparts, Poole said. “First, the magnetic field of the permanent magnet in the Swoop system aligns the hydrogen nuclei (protons) inside the water molecules in the patient’s head. Then the system uses radio waves to disturb the protons, causing them to wobble in synchrony. Once the radio waves are turned off, receiver coils detect the continued wobble, and the application of additional magnetic fields encode the position of the protons’ signals, which can be used to create a picture of the structures and tissue inside the body,” he explained.
The big difference is that high-power scanners have much larger magnets that produce more signal. “The amount of signal we get in an MRI scan is roughly proportional to the how strong the magnet is. If we go from 3T to 64mT, for example, we’re going down by a factor of around 40 or 50,” Ogier said. “The lower the signal, the longer the scan time needed to get a detailed image, so to keep the low-field scans to a reasonable length, we’re limited to a lower resolution image.”
To compensate for the decrease in magnet size and signal strength, Hyperfine has made numerous improvements to its Swoop system [4] and now gets clinically useful images of the brain with just a 15-minute scan. “There are innovations throughout the device from the ultra-compact magnet design to the simplicity of the user interface, ultra-efficient and robust electronics, motion-compensated imaging, and deep-learning image reconstruction and filtering, all in a safe and portable MRI,” Poole said. One of the biggest challenges was eliminating the need for a radio-frequency-shielded room around the scanner, so it could be a portable machine. “Without it, the electromagnetic interference that is all around us is detected on the very sensitive receivers in MRI systems and completely swamps the tiny MRI signals that make up the images,” he said. “The game changer was the development of interference-canceling technology that recovers noise-free images from almost impossibly noisy environments.”
Overall, he said, “We have made tremendous strides in image quality since the system was initially launched. The system can give actionable images at the point of care more immediately than can conventional MRI, because it’s always available.” The company’s overall vision is to “bring the power of MRI closer to patients who wouldn’t traditionally have access to MRI by breaking out of the four walls of the hospital and into clinics, offices, and low-resourced settings,” he said. “We’re pairing this vision of expanded use of MRI with continued technical advancements for improved image quality, speed, and usability.”
NIST is also looking into the possibilities of low-field MRI, and is particularly interested in the technology from the quantitative perspective, which will have both basic-science and clinical applications, Ogier said. “Compared to high-field MRI images, the contrast in low-field MRI images is different, so we have been looking into quantifying the properties of healthy human brain tissue at 64 mT,” he explained. The difference in contrast is related to the MRI relaxation parameters, or the way water protons spin when hit by a RF pulse and then return to normal. Those parameters, particularly ones known as T1 and T2, cause the image-contrast differences between higher-powered and lower-powered MRI scanners.
Working with Hyperfine researchers, Osier, NIST engineers Kalina Jordanova, Ph.D., and Katy Keenan, Ph.D., and others at NIST imaged white matter, gray matter, and cerebrospinal fluid (CSF) to determine their unique ultra-low-field MRI signals. Armed with that information, they also began developing techniques to identify “partial-volume voxels,” or those that are mixes of CSF and either white or gray matter [5]. Such data could then be used to tune the MRI for easy identification of white and gray matter, as well as CSF, Ogier said. The researchers are now also considering expanding the approach to identify unhealthy tissue, such as tumors, with an ultra-low-power MRI. In another project, researchers from NIST (Figure 8), Hyperfine, and the University of Florence in Italy have developed contrast agents based on iron oxide nanoparticles that are well-suited to ultra-low-field MRI [6].
Figure 8. NIST researcher Sam Oberdick investigated contrast agents well-suited to ultra-low-field MRI machines. In particular, his group tested iron oxide nanoparticles (shown in liquid solution). (Photo courtesy of R. Wilson/NIST.)
Additionally, NIST researchers have joined a collaboration led by Andrew Webb and Tom O’Reilly of Leiden University Medical Center in Leiden, The Netherlands, and Joshua Harper of Universidad Paraguayo Alemana (German Paraguayan University) in San Lorenzo, Paraguay, to develop a completely new ultra-low-field MRI research scanner. Rather translating techniques and approaches from ultra-high to ultra-low field, or making small adjustments to an already-existing low-field scanner, Ogier said the group is looking at ultra-low field with fresh eyes, considering unorthodox designs in terms of receiver coils and other components, and designing systems that get the most out of the reduced signal and are potentially less expensive.
“Since NIST is a metrology institute, we saw that by building a new scanner, we would have more control over its development, which would give us more raw data to access, and that meant we could be more confident in the measurements we’re making,” he said. Ogier also sees the potential of lower-cost scanners to enable more longitudinal studies. “If you have a less expensive MRI, you can do more imaging, so you could image the same person every few months and see what numbers are changing, which makes it a cool opportunity to establish baseline values and help research groups that are doing studies of pathologies.”
He added, “there’s a lot to be done. This is an exciting time in MRI technology.”
References
- D. A. Feinberg et al., “Next-generation MRI scanner designed for ultra-high-resolution human brain imaging at 7 Tesla,” Nature Methods, vol. 20, no. 12, pp. 2048–2057, Nov. 2023, doi: 10.1038/s41592-023-02068-7.
- French Alternative Energies and Atomic Energy Commission (CEA). (Apr. 2, 2024). A World Premiere: The Living Brain Imaged With Unrivaled Clarity Thanks to the World’s Most Powerful MRI Machine. Accessed: Apr. 4, 2024. [Online]. Available: https://www.cea.fr/english/Pages/News/world-premiere-living-brain-imaged-with-unrivaled-clarity-thanks-to-world-most-powerful-MRI-machine.aspx
- Hyperfine, Inc. (Oct. 9, 2023). Hyperfine, Inc., Receives FDA Clearance for Updated AI-Powered Software With Improved Image Quality for All Swoop System Sequences. Accessed: Apr. 18, 2024. [Online]. Available: https://investors.hyperfine.io/node/8436/pdf
- Hyperfine, Inc. Swoop Portable MR Imaging System: Details and Specifications. Accessed: Apr. 20, 2024. [Online]. Available: https://hyperfine.io/swoop/details-and-specifications
- K. V. Jordanova et al., “In vivo quantitative MRI: T1 and T1 measurements of the human brain at 0.064 T,” Magn. Reson. Mater. Phys., Biol. Med., vol. 36, no. 3, pp. 487–498, May 2023, doi: 10.1007/s10334-023-01095-x.
- S. D. Oberdick et al., “Iron oxide nanoparticles as positive T1 contrast agents for low-field magnetic resonance imaging at 64 mT,” Sci. Rep., vol. 13, no. 1, p. 11520, Jul. 2023, doi: 10.1038/s41598-023-38222-6.