We've had preclinical 12T and way above machines for a while now (Bruker will sell you a ~21T, 1 GHz Larmour frequency machine, for instance). They are great for some forms of spectroscopy, but the issue for imaging is that t2* becomes very short, t1 becomes long, and it becomes impossible to have good b1 or b0 homogeneity. This means that the images are not necessarily filled with as much contrast as you'd like, tend to be a bit artefact-filled, and will keep physicists like me in pay for a long time.
Still very cool though, and well done to the CEA team.
Could you repeat this in english? It's good for imaging big features but not small features, or high frequency density changes, or takes too long, or what?
B0 is the magnetic field that separates the nuclear spin energy levels. It's the one that's really big in this machine. B1 is the radiofrequency pulse that excites the nuclear spins. Poor B0 and B1 homogeneity means you are taking the MRI with a slightly different machine at every point.
T1 is the time it takes for the energy that B1 adds to the nuclear spins to leak back out of the nuclei. It generally happens faster when the energy per nucleus, which is proportional to the degree to which B0 has split the energy levels, is low. If I had to guess, it's probably because there are more weak couplings by which a nucleus can get rid of a little energy than there are strong couplings that can support a big jump. One of the major things that limits the speed at which NMR can be done is the fact that you have to wait for this energy to leak out before you can put more in for another round of imaging. (Maybe there is a way around this but if so they never taught me about it.)
T2 is the time it takes for the excited spins to fall out of phase with each other. It happens faster when there is a greater diversity of electronic influences on each nuclear environment - a bunch of identical atoms in a perfect lattice would never go out of phase. If you "zoom in" on T2 and study it with incredible precision you can pick apart each distinct environment around each population of similar nuclear neighborhoods. A population of identical neighborhoods, in practice, means a chemical. Chemists use that a lot but it's not done in imaging.
T2-weighted imaging is definitely used in medical settings. Generally MR is good for soft tissue and nervous system diagnosis and T2 is good because it gives contrast between otherwise very similar intensity regions in the brain for example.
With stronger magnets the part of the signal that contains the actual information decays faster in this case. At the same time the total magnetization, most of which is useless to you, decays slower. And you need to wait until it decays to measure again.
This explanation is extremely simplified, and the details here depend on the situation. Relaxation is a critical part of MRI and NMR, and there are different mechanisms that contribute to it.
The sensitivity of magnetic resonance methods increases with the magnetic field strength to the power of 3/2. But other effects like the mentioned relaxation cause certain practical issues at higher fields. So you don't always get the full benefit of the higher magnetic field, depending on what exactly you are doing.
Yes, their claim is for the most powerful full-body scanner. AFAIK and for perspective, the highest man-made magnetic field is around 45 T, the LHC magnets are ~9 T, and those in ITER are (will be) 12 T as well.
What is it about a stronger magnetic field in practice that decreases b1, b0? Is it that we cant create a strong magnetic field with good homogeneity in a small space? Are sensors also a limiting factor?
First off -- sorry for a rather terse initial comment last night whilst I was trying to get to sleep! This explanation (a) won't be on my phone; and (b) will have a lot more detailed added.
B1 is the oscillating RF magnetic field that is responsible for both delivering power into the sample / patient, and once their nuclei have received that and undergone the magnetic resonance phenomenon, they emit RF at the same frequency. We / the MRI community divide this field into two: B1+ and B1-, for transmit and receive (and they also happen to be left and right-handed circularly polarised fields). It's worth pointing out that a physical device -- a lump of copper with capacitors in it -- called an RF coil transmits B1+ and receives B1-. Because we want B1+ to be homogeneous and uniform and B1- to be sensitive, in the last ~20 years, most clinical scanners now have a separate "body coil" for transmission and multiplee anatomically shaped ones ("breast/thoracic/spine/knee/ankle/wrist/whatever-coils") for reception. Multiple because you can noise-optimise the size of each little loop, and you can also do tricks ("parallel imaging" or "compressed sensing") to exploit their different spatial locations and speed up the scan.
Now, at low field, say at 1.5 T (a typical clinical MRI scanner's field), the Larmor frequency for protons that are most often of interest to the operator of the scanner is objectively low (about 60 MHz for 1.5T). By 'objective low', I mean "its wavelength is a fair bit longer than the instrument". Mathematically, if you took a course that involved Maxwell's equations, we're in the quasi-static regime: \nabla \times B = µ0 ( J + \epsilon_0 ∂E/∂t) and at low field ∂E/∂t ≈ 0 compared to J. So, Biot-Savart's law works for describing the coils – you put a current in them (and quite a large one -- 1-3 kV voltages are not unusual for transmit inside a body coil on a device that has a minimally low resistance) and get a magnetic field out. Everything is DC, doesn't interact, and that magnetic field is what permeates through the patient. The optional difference in electronics aside, there's no substantial difference between B1+ and B1- and all is hunky-dory and comparatively well understood.
Now, at high fields, a bunch of stuff happens. Firstly, the frequency goes up -- at ~12 T, it is ~500 MHz -- and with that, the wavelength goes down. At 12T (or 9.4T, or 7T, the other commonly-used high-field strengths where this is relevant), it's well below the size of a person, and rather than having a quasi-static problem, you now have a full-wave electrodynamic problem in which you can't just ignore the displacement current, ∂E/∂t ≠ 0. As well as making it harder to design RF coils, this actually means that you cannot ignore the fact that it's a wave -- to a first approximation you'll get nodes and anti-nodes in the tissue on a human length-scale. Not only that, but all of the really important electrodynamic parameters -- the µr's and \epsilon_r's that determine what speed the wave propagates at, and so on -- are (a) actually complex numbers; (b) different in different tissue types / diseases; and (c) not as well measured at high field as they are at low field. Another effect emerges from this at ultrahigh field: B1 twist. B1+ starts being ≠ B1-* (where the * denotes complex conjugation) and you get some rather "interesting" patterns appear. [1] From an electrical engineering point of you, you've got a near-field antenna design problem into a fundamentally unknown electrodynamic environment (unknown, because to some extent you are measuring it! Is that a cystic, fluid-filled region in the brain, or a dense, cellular tumour?)
So, at ultrahigh field, the only thing you can do is change your approach completely. Rather than having one homogeneous, optimised transmission device, you have multiple -- transmitting in parallel ('pTx') -- and change the phases and amplitudes between them to optimise the homogeneity in the response. This has the downside that you have to be careful not to accidentally start to heat your patient, and so throw in a lot of work, simulations and measurements about safety here. Rather than looking like a loop of wire, coils now tend to look like e.g. fractionated dipole antennas -- which makes them very sensitive to their electrodynamic tuning and exactly the value of µr and \epsilon_r (or equivalently the conductivity) of the sample. Fat is different here to muscle, and that difference is bigger at ultrahigh field...so you now see people selling "dielectric pads", foam-wrapped pads of a material with a favourable \epsilon_r to literally pad around the patient if you have a problem. These too have their own safety concerns, so you end up spending a week running a full-wave FEM simulation of Maxwell's equations prior to destructive heating testing before doing it on a human. The whole thing becomes a bit of a pain.
[1/2 -- I never knew HN had a comment character limit until now!]
I dont think you understand how valuable this information is for someone who is a layman. I really appreciate the response!
Oh shit so the math and detection gets stupid difficult at higher mag fields because of wavelength/scale and complex environments. I cant imagine how interesting those patterns for B1+ ≠ B1-* are from a diagnostic perspective /s. But as an academic pretty damn cool with a lot of opportunities.
Heating the patient sounds like a bad idea. So the higher mag field incurs a lot of problems due to physical properties and the sample media being the human body. It does look like we are innovating a lot to make this work. I'm thoroughly enjoying reading so far.
[2/2 -- see above about not knowing HN had a character limit!]
Now, for B0. The simple explanation is this: as alluded to above, we're all made of very different stuff. Bone, air, fat and muscle all have very different electrodynamic properties. At their interfaces, there's a "rule" that comes from Maxwell's equations† that means that you _must_ have a discontinuity in both the normal component of the electric (D) field in media and the tangential component of the magnetic field (H). This isn't negotiable, under the assumption that your patient does not have a net overall charge or current flowing through them. What this means, practically, is you get tiny bound surface charges or currents appearing at the interface of different materials. These then change the magnetic field homogeneity -- how B_0 varies over space -- because they create magnetic fields that add to it (the principle of superposition). Remember, all of MRI is based on having extremely homogeneous fields so that you can make a mapping between frequency and phase of your signal to the space of your sample. Inhomogeneities destroy that -- they can e.g. make two places look the same if they should be different, for example. B0 homogeneity is a big problem: we need the field to be uniform across the sample.
Practically, what that means is twofold. One, for big changes -- caused by air, or a dental filling typically -- you get large, far-field effects that distort the image macroscopically. The maxillary and ethymoid sinuses (in the front of your face) together with the hard palette make it harder to image certain regions of the brain -- the bits that are basically directly above the roof of your mouth. This is _entirely_ due to the difference in the dielectric properties of air and "meat". Lung imaging with MRI is very, very challenging (but possible!). What works well for the brain does _not_ work at all, really, on the long -- and especially not at ultrahigh field. Even something as "simple" as cardiac imaging becomes more difficult, because the heart contains oxygenated and deoxygenated blood in close proximity -- and they have different dielectric properties.
Secondly, all tissue on a microscopic level is inhomogeneous. Cells have bilayers and a complex microstructure. Organs have microstructures made up of different types of cells. The liver has a a beautiful microstructure, and a lot of iron in it. This causes a decrease in the effective static field inhomogeneity -- which manifests itself in terms of an NMR relaxation parameter called T2* -- meaning, in short words, that the MRI signal vanishes more quickly over time than it does at lower fields. That's why livers tend to look a bit darker than their surrounding organs on some types of MRI scans. If you play a single RF pulse and acquire signal on an NMR tube of pure water in a chemistry lab, you'll get a voltage you could measure with a hand-held AC voltmeter appearing for seconds after the initial excitation. Put a sphere of water in a 12T MRI scanner, and frankly you're doing well to have signal there after ~500 ms. In vivo, that's likely to be ~30-50 ms. That makes actually doing the experiments hard - your [gradient-echo] signal vanishes quickly.
The whole B0 homogeneity thing is a coupled problem, best solved by considering Laplace's equation for the scalar magnetic potential (\nabla^2 \psi = 0). This famously has a solution space that can be expressed as an infinite sum of spherical harmonic functions. So, MRI scanners (yes, even clinical ones!) have a series of loops of wire that generate B-fields that approximate the first few spherical harmonics -- in Cartesian coordinates, they're weird things like x^2-y^2 and look like the pictures of the orbitals for electrons inside a hydrogen atom, if you've ever seen that. Patient goes in scanner, B0 field inhomogeneity is measured, magic regression‡ works out the right coefficients to dial a current into each coil, and the field is magically more homogeneous.
Now, the trouble with this is that the spherical harmonics are a sum until infinity. And you can't build an infinite number of coils. The first few components get you most of the way there...but not all. And again, at higher field building effective shim coils, working out what your artefact is and where it comes from, and so on, all become more difficult.
</end lecture>
MRI is complicated -- it spans mathematics, physics, chemistry, engineering, biochemistry, and medicine. Frankly, that's why I like working in it as an academic area -- there's always something to learn. There are about 6000 others who do so technically, and we have a big, week-long worldwide conference once a year to both teach this stuff (where I have lectured) in quantitative detail, and show off our latest and greatest results (the usual big "evil" companies typically take academic work and put it in their products -- which I find rewarding, as it means your work can have a difference quickly and you don't have to deal with the FDA). I'm sure CEA / ISEULT will present there and I am genuinely excited that their mega-machine has finally been built. But there are a lot of challenges to overcome. Of course, there are also a huge number of opportunities -- it is (hopefully!) going to make it possible to image at higher spatial resolution, and also be much better for MR spectroscopy, and reveal information about low concentrations of biomolecules. But I guess the main point of my comment originally last night was trying to convey "the journey is really just beginning" and not "expect this in your local hospital in five years time".
† Specifically Gauss's law and Ampere's law at the boundaries for the D or H fields giving rise to the rule that there must be a discontinuity in D, with the difference normal components being equal to the bound surface charge density; and the difference in the tangential part of the H field being equal to the bound surface current. Putting these together with the constitutive relationships for D and H, and you get that the surface current is equal to B_{tangential, 1} / µ_1 - B_{tangential, 2} / µ_2 = J_s. In biology, µ is approximately fixed...so the surface current changes.
Your last paragraph has definitely increased my optimism. I wish i had more time to reply. I might have to wait for another weekend for this.
>there are also a huge number of opportunities -- it is (hopefully!) going to make it possible to image at higher spatial resolution, and also be much better for MR spectroscopy, and reveal information about low concentrations of biomolecules. But I guess the main point of my comment originally last night was trying to convey "the journey is really just beginning" and not "expect this in your local hospital in five years time".
What specifically in your estimation will aid in detecting low concentration biomolecules?
I always thought an MRI worked a bit like a microwave oven that it would jiggle water or the hydrogen molecules in a person. The jiggling made the molecules emit energy, radio emissions. But it's much more precise than that.
nibib.nih.gov has a nice explanation.
>MRIs employ powerful magnets which produce a strong magnetic field that forces protons in the body to align with that field. When a radiofrequency current is then pulsed through the patient, the protons are stimulated, and spin out of equilibrium, straining against the pull of the magnetic field. When the radiofrequency field is turned off, the MRI sensors are able to detect the energy released as the protons realign with the magnetic field. The time it takes for the protons to realign with the magnetic field, as well as the amount of energy released, changes depending on the environment and the chemical nature of the molecules. Physicians are able to tell the difference between various types of tissues based on these magnetic
What's strange is for clarity of images an old-fashioned x-ray/CAT scan is better it's not as fuzzy as 1T or 3T MRI. From what I can gather is the more powerful an MRI is the faster it can capture images and with more detail. The speed is why they are clearer you the patient has less time to squirm around. It's also mean less time for the patient to be in the machine and the machine is less constrictive (bigger hole).
X-ray / CT use ionizing radiation and basically trade off image quality with how much damage they are doing to you (I don't mean they are dangerous, but they deal with high power radiation). MRI uses harmless magnetic fields, the tradeoff being that the Zeeman interaction is orders of magnitude weaker than whatever happens in an xray and so the signal to noise ratio is much lower.
In an MRI, stronger field means more magnetism to work with an a higher db/dt so higher voltage induced in detection coils so you end up with a notional square law (actually 7/4 due to skin depth) scaling of sensitivity with field strength. But higher field introduces other problems - more RF power deposited into the patient, field variations over the spatial scale of the patient, static field homogeneity, others, so it's not automatically better. Also, making the machine bigger means more challenges creating a homogeneous field over the large region (for example, its easier to make a narrow bore nmr magnet than a big clinical MRI (but the nmr would want better homogeneity).
Speed really again trades off against quality. You end up with essentially a fixed amount of signal per time, which you can use for more resolution or more SNR
But they are. CT uses pretty high doses, and each CT scan increases lifetime risk of cancer by order of magnitude 0.1%. CT scans absolutely should not be done willy-nilly.
...increase by 0.1% on average, over the average a priori risk. If I didn't know I was a risk to begin with, I certainly won't care about a rounding error magnitude increase.
Pretty often for cerebral scans you don't have a whole lot of room to move due to being literally restrained by the focusing gear. Most people who have done MRIs regularly get pretty good at not blurring the images; the more still you remain the less chance of them having to do another round.
I stayed still for my first, too. It was really noisy. I dislike the part where it suddenly changes the "rhythm" but I prepared myself for that now. Going to get an MRI soon again.
You do get used to it, believe it or not. One thing that helped me understand what was going on was looking at the output files, each pulse of the magnet is a slice within the images, and each sound corresponds to different types of measurement. I tend to just count them to work out what type of imaging is currently going on, it's ultimately pointless but it's something to focus on given the lack of other stimulus. You'll know you're there when the MRI tech comments that it's obvious you've done a lot of them because every image was perfect.
Massachusetts has almost universal health care coverage due to it being a priority of the Commonwealth. We have the lowest uninsured population in the USA at ~97% coverage. (Lowest coverage state? Texas, ~82%.)
I used to go to a local medical center in Boston that was apparently mostly serving Medicaid recipients. It had an MRI area that was like an assembly line -- huge waiting room, dozens of people constantly shuffling in and out of MRI rooms to get scans for whatever reason. So, a faster MRI machine means more people can get more scans. Which might be the difference between life and death, especially when you're on Medicaid.
"better" depends on the problem domain. MRI is useful because it offers better contrast between different tissue types that look the same on CT. The modalities can also be combined.
Interesting that they mention the link to fusion magnets.
I'm currently in the market for a used MRI scanner to supply the external field to my prototype fusion reactor. One reason I'm looking at MRI is the extreme uniformity or the field. I'm likely to get a much better field than I'd otherwise get if I tried to build one myself. I only need a fairly old, ~.35T permanent magnet model to do what I need, but the uniformity is vital.
Is there any public information about what you are building? .35T is not that high and it may be easier and cheaper to build a permanent magnet array that meets your needs vs being to repurpose a scanner. I'd be curious to know what kind of field profile you want
Cool! I worked at the NeuroSpin lab in ten years ago, and I recall the 11.5T giving people a lot of grief. There was significant doubt about whether it would ever work. IIRC one of the biggest issues was that it would pull in microscopic metal dust, which would then line the imaging tunnel and distort the image. I wonder how they resolved this.
Very good question. I remember them mentioning this in an internal video, but I don’t think they detailed a solution. It’s a bit out of my field, to be honest.
I’m not at Neurospin, I’m a materials physicist on the other side of the fence, so to speak :). I just have a keen interest in what they’re doing there. My daughter also participated in a study on neuroplasticity and cerebral development in infants there and it’s been an interesting experience.
I doubt the push from the positive pressure would be enough to offset 11.5T worth of pull, but perhaps that was part of the solution. I really don’t know.
Where nuclear stands for nucleus, atom, not for radiation or radioactive decay... In German language is called "Kernspintomographie", "Kern"=nucleus or nuclear, and people happily use both the word and the device.
I don't think use of the word "radiation" is as necessary as you may be suggesting for the purpose of technical specificity. In fact it is a very broad word. Visible light is radiation.
The theme of the comment thread is "isn't it bad that the public's fear of radiation prevents us from calling them NMRIs," but I am arguing that it's not such a tragedy after all.
Is there a (theoretical) possibility of having portable MRI scanners someday? The size of a big suitcase or maybe even something like the tricorder in Star Trek.
The scanner needs a static field that's homogeneous over the region you're imaging. And magnetic fields can't be homogeneous outside the bore of a magnet, i.e
in free space (Ernshaw's theorem maybe). So any portable scanner has to either have a small, locally homogeneous spot (see Blümich et al from the late 90s or what Schlumberger has done with borehole logging) or be some kind of bigger closed thing like the Hyperfine systems that a sibling post mentioned.
It's a very interesting field from a technical perspective, but there are fundamental physical principles standing in the way of making a portable scanner like an ultrasound
There are some ways to correct for field inhomogeneity, but there are also fundamental limits like shorter signal lifetimes (or analogously bandwidth requirements) in inhomogeneous fields that you can't really get around.
It's always possible there will be some advance, but I'm not aware of compute power as any real bottleneck
While it's not the single barrier, but room temperature super conductors would go a ways to reduce the necessary bulk.
But the magnetic fields would still be huge, you wouldn't want to just turn one on in your living room before some serious magnetic shielding technology was also invented.
I attended a talk that was trying to develop something like that. Lots of challenges, but his goal was to make MRI cheap and portable. Sorry, I don't remember his name, but googling portable MRI or low-field MRI will give lots of results, e.g. this one: https://www.goodnewsnetwork.org/fda-approves-worlds-first-po...
Wikipedia says the previous world record[1] for an MRI magnet was 11.75 Teslas. This article says this one uses 11.7 Teslas. Anyone more knowledgeable care to comment on the difference? They say a record for this volume (large enough to pass a human) but surely the other one would be that large as well?
It’s very simple: it’s because the story and your link talk about the same magnet. Yours was (initially) written when it was in the design phase, and it has now produced its first images.
CEA is definitely not (yet) for clinical use, i.e. treating sick people routinely. It's for clinical research, i.e. scanning sick or healthy people, potentially (if and only if in the context of a registered clinical trial) changing their treatment as a corollary.
Different operating principle. CT scan uses X-rays beams from many angles, like taking X-ray photographs from different angles, and the images are combined mathematically into a 3D image. X-rays are a lot like regular visible light, they travel in straight beams and form clear images when focused, except unlike visible light, many substances are more transparent to X-rays.
MRI uses radio signals in the microwave or below frequency range combined with a high strength magnet with a shaped magnetic field to make a field strength gradient. The radio waves resonate with some atomic nuclei at a frequency which depends on the magnetic field strength, so the combination of magnetic field gradient and radio frequency creates a thin layer in space where resonance occurs, affecting the radio waves in that layer while they pass through the rest of space much less affected. The layer is moved around, radio receivers pick up the affected signal, and the 3D image is reconstructed from that signal.
Fruit and vegetables are pretty popular test objects: they sit nice and still, they have interesting structures (i.e., look cool), and they avoid ethics/safety issues involved in scanning humans or animals.
"Phantoms", test devices with specific and well-known properties, are used for more careful testing but produce works pretty well for a quick sanity check.
The pumpkin here is just a test. They are still tuning it and at some point want to look at brains with a 100 μm resolution. They just cannot put living people in it before it’s been tested and certified.
I’ll pass. Not gonna stick my head in anything more than 3T ever. Even at 7T you start getting disoriented which means the field has some effect on your head? Not sure why anyone would ignore that.
I had the pleasure of being a test subject for my local hospital’s new 7T last week. They strapped my head down to the table and I was only able to listen to the intermittent “DO-DO-DO-DO” of the machine, while a mirrored television played 15 minute aerial footage of Norway’s fjords on repeat for an hour. Reminded me a little of the brainwashing scene from A Clockwork Orange.
The radiology report said, “No significant incidental findings identified,” so that’s good at least.
Wow, that's really nice of you to help your hospital get the bugs worked out on the new machine. Did you experience any heating? If so was it in your dental fillings?
Still very cool though, and well done to the CEA team.