New X-ray technique images soft-tissue tumors clearer than MRI | Hacker News Reader

New X-ray technique images soft-tissue tumors clearer than MRI

newatlas.com

147 pointshhs77 comments

From what I remember, MRI doesn't exist without some fairly sophisticated signal processing. The teaser seems to be suggesting they're doing signal processing on Xrays instead of just exposing photography plates (or do they even do that anymore? Are they digital like cameras?)

Many many years ago I recall a project that looked at the tissue all around a tumor, and mapped the refraction from many different angles. They took that information and worked backward to find the exact angles to send radiotherapy beams so that they would cross inside the tumor, reducing side effects. I wonder if any of the principles are the same here.

They're pretty much all digital: no consumables and no giant filing system are big wins institutionally, and lower exposure is good for the patient too.

Ten years ago, I brought a film over to a hospital, and someone had to pull a dusty, disused lightbox out of the closet so we could look at it. I can only imagine it's become even more electronic since then.

> and lower exposure.

Often, but not always and it also depends on the technology. Computed radiography got all the benefits of going digital, but allowed you to keep your X-ray equipment. Doses mostly went up. True digital requires new X-ray rooms and mostly reduced dose. However for this to occur it depends on the staff, as higher dose images look really nice (lower noise). There are also aspects of film radiography that have been very hard to achieve with digital. One that comes to mind is the examination of joints in rheumatology cases. The resolution of film screen systems is very hard to beat and small erosions are best seen with increased resolution.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3045190/#!po=1....

It's all over the map. Personally, I have a very frustrating anecdote in that I got extensive x-rays on my toes by a podiatrist. He used a very sophisticated platform (at least from my POV) to show my layer by layer the damage to my toe joints and referred me to a surgeon. Multiple consults later, the best this surgeon could do (who also works with pro-sports teams at the highest level) was use microsoft paint to look at the x-ray images from the podiatrist to assay the damage. The Microsoft paint version was essentially a single flattened image which lost a huge amount of information. I suggested taking his own images thinking he would have a better way of looking at that data. Nope. Microsoft paint. He didn't seem to even have any idea how much information he was not getting despite multiple attempts to explain it to him. I am in an adjacent field so I know the lingo, but whatever surgeons just want to cut.

Signal processing on x-rays in 3d is a CT scan, as I understand it this is fundamentally different technique that doesn't involve the differential in opacity to x-rays, but seems like it's based on how a wavefront gets perturbed on its way through tissues.

A ct scan is essentially shining a light on something from a bunch on angles, observing the shadows, and trying to reconstruct a 3d object from your observations. Most ct scanners boil down to spinning an X-ray machine around your patient at ludicrous speeds. A ct scanner without cover is a sight to behold[0].

[0] https://www.youtube.com/watch?v=pLajmU4TQuI

I think they are, as a student I used to work on the FPGAs for pixel sensors for high energy particles, one of the potential uses was apparently xrays.

Not sure exactly what signal processing you're referring to - MRI would not work (would take prohibitively long) without compressed sensing. But I would not consider that fairly sophisticated signal processing, it's in the classical limit and uses a simple prior that k-space is sparse.

You may be the first person I've ever seen use "simple" and "k-space" in such close proximity.

... as opposed to a much more complicated prior, such as a natural tissue model...

Of course, but a) joke and b) if you haven't been exposed to it before--and in many cases, even if you have--it can seem pretty daunting.

Thanks for mentioning that, I didn't know about k-space. I worked with 2D Fourier transforms a lot in MATLAB about 15 years ago, but if anyone has info or examples of a 3D FT and how to interpret them, I'd love to see them.

MRI has been in use for much longer than compressed sensing has been understood, hasn't it?

In a research environment I encountered it about 3 years ago, and in a clinical one this year. Neither place were cutting edge with it, but the broader point is as you suggest - compressed sense/sensing (different vendor names, same thing) is pretty new.

To summarize the research, the idea is to use X-rays like ultrasound: visualizing the elastic properties of the target instead of the density, but at higher resolution than ultrasound. They were able to create a 2-D image of particles in a gel in the lab.

Keep in mind that this is physic research, not medical research. They show that this imaging is possible. There are a bunch of difficult steps to making this medically useful. First, it is currently 2-D, not 3-D. It also needs to move from creating an image in the lab to a repeatable process, and then turned into a medical equipment product.

TL;DR: This is a lab demonstration of a new imaging technique in a physics journal. The title is entirely speculative. Abstract and paywalled paper here: https://iopscience.iop.org/article/10.35848/1882-0786/ab7e06...

Just throwing out there: US is 2D, X-ray is 2D, MRI is 2D slices, and CT is also a series of 2D slices but you can reconstruct reasonable guesses at 3D structures with it.

This being 2D doesn’t strike me as a huge drawback

A collection of 2D slices /is/ a 3D image. Modern CT and MRI are 3D modalities.

Reading one 2D slice at a time and calling it 3D is... not wrong. I don’t think it’s right, either. But we are now splitting hairs and not adding value.

Correct me if I'm wrong, but aren't X-rays generally much more dangerous than an MRI? I would think that things like 7 tesla MRI technology would be the future of medical imaging, not X-rays, if only for the fact that X-rays are cancer-causing.

Depends on what you want to image (and why).

Bones don't produce a strong MR signal, and usually show up as black. While people occasionally try to interpret these (e.g., https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3500787/) I'm not confident you'd be able to find a hairline fracture on most scans.

The resolution of a modern CT scanner is also crazy good. Ours goes to 90 µm--you can see individual threads on bone screws--and the scan just takes a few minutes. MRI is usually closer to 500µm-1+mm isotropic voxels and the scans are rather long since you need to average many volumes to get rid of noise.

Finally, the risks are totally different. An X-ray or CT scan definitely exposes you to some radiation, but I don't think there are any absolute contraindications. An MRI is safer in that there's no radiation, but can be incredibly dangerous if some kinds of metals are involved; I would be terrified to scan a kid that ate an unspecified toy, for example. Other metals--especially the ones used in medical implants are safer but create giant artifacts around them that may defeat the point of doing a scan.

For the foreseeable future, I think there will be a place for both kinds of imaging (and others!).

> Bones don't produce a strong MR signal, and usually show up as black. While people occasionally try to interpret these (e.g., https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3500787/) I'm not confident you'd be able to find a hairline fracture on most scans.

MRI is superior to x-ray for fracture detection. It’s just vastly too expensive and slow. Bone shows up great, dense cortical bone has little signal but most bone isn’t cortical. When bone breaks it produces oedema which is perfect for MRI. When it’s hard to work out clinically and/or the x-day is inconclusive, patients end up in MR.

Bone tumours and other pathological conditions are also imaged with MRI very effectively.

I’m an MR radiographer.

https://www.researchgate.net/publication/12690778_Diagnosis_...

MR is not always superior to x-ray for fracture detection, in fact, in the acute phase MR is often normal as marrow edema has not yet developed. Resolution is far worse than CT or plain film xray, making subtle non-displaced fractures impossible to see. MR can be good for occult fractures that are not visible on xray, but usually better after at least a few days so that marrow edema will have developed.

Thanks for this. In keeping with what you say, the type of patient we see is nearly always a week post injury with pain but a negative x-ray. They want to work or travel or something and are after a definitive answer. Another key benefit to MRI, I suspect, is that if it isn't a fracture a CT isn't that helpful, while any soft tissue injury is likely well demonstrated on MR.

Interesting...

I try very hard to avoid everything below the neck and our tech was pretty dismissive when I wanted to use it as an impromptu X-ray (but we scan almost exclusively heads).

Is it really better than CT? I can see how anything 3D would beat out a planar X-ray.

The expensive part is solely on health institutions. They could still be super profitable at a much lower price.

Thanks for the informative answer. What other kinds of scans are there? I know of PET scans, how are those used? Do people combine different kinds of scans into one visualization?

Emissions based scans like PET/SPECT use positrons/gamma rays released by an infused tracer for imaging things like hungry tumors (labeled glucose) but can also be used to image other molecules or receptors. Very low resolution. Usually overlayed on CT scans.

Ultrasound has become the best point of care imaging for a lot of things.

tl;dr? Skip to the photoacoustic stuff at the end, which is amazing.

I feel like I didn't do MRI justice there, as it's really a family of techniques. You can tune the pulse sequences to image a lot of different things: a T1 scan of the brain, for example, emphasizes white vs. grey matter; T2* highlights fluids. fMRI uses slight changes in the properties of oxygenated vs. deoxygenated blood to infer brain activity, rather that structure. You can even find the distribution of other chemicals via Magnetic Resonance Spectroscopy. People use this to see if levels of neurotransmitters vary in the brain. However, you need some water/protons to image, which is why this works poorly for bone.

PET uses radioactive tracers that either bind to or replace other substances in the body. For example, in a cancer studies you're given a sugar analog called Fludeoxyglucose, or FDG. Different tissues take up glucose at different rates, and metastatic tumors are particularly hungry for it, so a lot of FDG ends up near the tumor. As it decays into oxygen and sugar, it emits positrons, the antiparticle of electrons. When the positron interacts with an electron in body's own matter, it gives off two photons that fly in opposite directions. By measuring coincidences across a set of detectors, you can figure out where the positron came from and the tracer and the tumor must be near by. FDG is the most common, but you can radiolabel lots of other things too, including drugs.

SPECT is similar, in that it uses a radioactive tracer. The SPECT tracers usually emit gamma radiation directly though, and the detector is a bit different--it doesn't rely on coincidences, which makes it cheaper but a bit lower resolution. An even simpler option is a gamma camera, where you get a single 2D image (like an X-ray vs a CT scan); not sure if this is still used much.

Ultrasound you undoubtedly know from people having babies. It gets used to look at other soft tissue too, especially knees. Ultrasound of the heart are often called echocardiograms.

Photoacoustic techniques are absolutely wild. You shine a light (usually a laser) into the subject. The wavelength of this light is carefully chosen so that it a) passes through stuff that's in the way and b) is absorbed by whatever you want to image. When a substance absorbs light, it heats up and when things heats up, they expand. Rapidly expanding things make pressure waves (i.e., sounds). By detecting those sounds from multiple locations, you can figure out where things of that color are. The color of blood varies slightly depending on how oxygenated it is, so you could potentially use this to figure out what parts of the brain are using more oxygen--and thus more active, just like fMRI. (This sounds terrifying, but the temperature changes are minuscule).

These all have different strengths and weaknesses, so combining them is pretty common.

If you were planning a brain surgery, you might want an MRI+CT, which would let you map points on the skull (or things inserted into the brain) onto a more accurate map of the brain. You might even add fMRI to know what these regions do (though this is fairly rare still).

Ditto for MRI+PET or CT+PET, where you might want to see where a tumor is relative to some bone or organ.

More likely to see DTI than fMRI today in practice for this sort of work.

CT + MRI is pretty standard, often with both a T1 and T2 type contrast of some sort, e.g. SPGR.

DTI identifies white matter pathways, but it can't (alone) tell you much about function, can it?

True, at least not directly (you can segment tractography to infer this to some degree).

But fMRI is rarely used in practice for this, and may not be specific enough for many needs.

Physical mapping is still very common for function.

> You can even find the distribution of other chemicals via Magnetic Resonance Spectroscopy.

Isn't magnetic resonance spectroscopy just another phrase for nuclear magnetic resonance or does the term mean something else in healthcare? The NMRs I've seen have cryogenic chambers for holding molecular samples and I just don't see that working very well with a live humans.

Or is it literally just taking the NMR spectrum for a neurotransmitter and targeting its resonance? (that's what MRIs do with water molecules right?)

> As [FDG] decays into oxygen and sugar, ...

For the confused:

FDG is glucose with a F for O substitution. Fluorine radioisotope in it decays into oxygen, converting FDG to sugar.

Others have answered part of this well so I won't repeat, but there are some other factors that aren't obvious so I'll mention them.

The idea that radiation == bad, magnetism == good is overly simplistic. It's not true that an MRI can't damage you, so much as that the design is careful to avoid the damage. Principal mechanisms here are SAR (specific absorbtion) and PNS (peripheral nerve stimulation). Basically you are dumping a bunch of energy into the body; if it gets to high you start to heat tissue too much, and/or induced currents can cause nerves to fire off aggressively.

As a patient you don't need to worry much about this - when you design an MRI you have safety systems in place to limit things like this to safe levels. But your point about high field MRI being "the future" also runs into trouble with the pulse sequences having tighter tolerances, for lack of a better term, than at lower fields, before things like SAR become an issue. So you gain some resolution but some things are harder to do. Also, from a clinical view, higher cost higher resolution scans aren't always what you want. High speed, low cost, lower resolution would be great for a lot of situations.

Another issue that is sometimes a problem is that MRI images are not stereotactically accurate (cf CT).

Also the somewhat confusingly used "contrast" in this context. It is used both to refer to the natural differentiation of tissues by a modality (e.g. bone is bright white and easy to discern in x-ray, but some soft tissues just "look the same") but also for the introduction of a "contrast agents". The latter case is when you introduce a chemical that will highlight certain anatomical structures. There are different ways to do this but for the most part a) the contrast agents are not particularly safe and b) there are different ones for both imaging modality and for type of contrast. So there are tradeoffs there also. See for a common example, angiography.

Finally, siting high field magnets is no joke, which keeps the costs high for capex, and the operational costs are high too.

IIRC, for fields >5T the movement of your blood is enough to induce currents you can detect as a weird tingling in your tongue.

Probably still safe, but it’s not as clear as I had previously assumed.

I'm not sure what you mean by movement of blood induces currents. The magnetic gradients are what induce current in the body, but usually more peripherally where the gradients are steepest and in big nerve bundles like sciatic.

When I would put humans in our 7T machine they'd sometimes feel their leg kicking during the acquisition. Because it's actively shielded with a pretty steep magnetic field from 0 to 7T you have to advance them into the scanner very slowly otherwise they can become very uncomfortable and nauseous.

>> I'm not sure what you mean by movement of blood induces currents.

Blood is a conductive fluid. When you move a conductor through a magnetic field there is an electric current induced.

what about rotating a ring on its axis of symmetry? Blood is in a loop, though not perfectly smooth and unchanging flow, as valves and stuff are squeezing at different time.

You sure it's the blood? I would have guessed it affects nerves directly.

The more I introspect, the less confidence I have in my own knowledge in this regard. At best, I forgot the important lesson of Gell-Mann Amnesia.

A chest x-ray is about 5x less radiation exposure then a transatlantic flight.[1]

1. https://www.gov.uk/government/publications/ionising-radiatio...

Let’s assume this new X-ray technique is not actually an ordinary X-ray but really is a modification to a CT scan, since otherwise there’s no reason for them to compare it to an MRI.

A head CT scan delivers 40 mGy of radiation to an adult brain. The dose of background radiation the average human body gets per year in the US is 3.6 mGy. The average adult brain is 1400 cubic centimeters. The average adult body is 62000 cubic centimeters. 1400/62000 = 0.02258. So the brain is actually only exposed to 2.258% of 3.6 mGy of background radiation per year, which is 0.081288 mGy per year. 40/0.081288 = 492. This means a head CT scan is equivalent to 492 years worth of background radiation blasted into the brain in the span of 10 minutes.

A transatlantic flight exposes your entire body to 0.1 mGy of which your brain only receives 0.002258 mGy. 0.002258/0.081288 = 0.028, and 0.028*365 = 10.22. This means a transatlantic flight exposes your brain to 10 days worth of background radiation over the span of 20 hours.

> A chest x-ray is about 5x less radiation exposure then a transatlantic flight.[1]

The more important measure is CT, not flat x-ray.

I think that is an interesting comparison but I'm not sure it's really applicable. A brain MRI can take up to 45 minutes or so. To be honest, I'm doubtful you could image someone's entire brain in detail via X-ray (as the article seems to imply) by using the sort of "one-and-done" X-ray normally used to look for run-of-the-mill bronchial infections, etc.

Right, the 3D equivalent is a CT scan which is a much higher radiation dose than a regular x-ray. On average every 1000 CT scans done causes one new case of cancer.

Source?

A brain CT doesn't take 45 minutes though--and a brain MRI doesn't use x-rays.

Always relevant http://xkcd.com/radiation/

MRIs are expensive. X-Rays are dangerous, but if you’re about to give someone radiotherapy I would think negligible. Especially if it means you can better target the treatment dose.

Also getting secondary cancers years down the line isn’t an issue for palliative type treatments.

So plenty of uses.

Long term, sure, but right now there's a limited number of very expensive MRI machines, X-rays are much more ubiquitous. Even if this involved a new machine, I imagine it would be much cheaper than an MRI.

In the US the cost of an MRI varies wildly and doesn't much correlate with the cost of providing them.

The machines are not cheap, but a busy site can churn through scans, so it can end up working out to something fairly reasonable and not at all the most expensive thing going on in US healthcare.

Liquid helium to cool superconducting magnets in MRI machines is also at a premium. This may use a much less expensive machine.

Those of us with certain implants can't have MRI without either our chests being torn apart or burned by red-hot metal from the inside. X-Rays are much safer in my case.

Today I've learned that elastography is a field of medical imaging that focuses on the stiffness or softness of tissues.

I might want to learn even more about it because I always have problems with my tissues and doctors fail to help me.

I'm hoping this is a step closer to being able to image connective tissues too. But being able to diagnose muscle tears and such would also be pretty helpful.

Hard to stay healthy when your range of motion is impaired.

MRI and US already do. It’s just not often helpful, because damage doesn’t need to be grossly visible and because we can’t do much for it regardless.

As someone that was shot in the neck and I have metal that cannot go through an MRI machine... will this be my savior?

As someone with metal implants, my experience is that metal causes extreme scattering in CT imagery which makes the modality often useless if you need to get anywhere near the site even with state-of-the-art computational filtering techniques. So I guess it depends on how much metal is still in there and where you need to look.

We are all stuck at home and therefore you cannot just drop this comment without elaborating.

> you cannot just drop this comment without elaborating

That's false.

What is needed? I get shot, I have metal, I have not decided to operate to remove it. Life goes on

I have always wondered how invasive the TSA scanners are and whether one can have an opt-in medical reports that flags potential serious issues from those scans directly sent to you (if something really stands out).

As far as I know, all of those scanners as deployed are millimeter-wave, not x-ray. They're not designed to penetrate the skin.

oh, so no useful medical info

Not sure if I would choose to get irritated with X-rays if I think I have a cancerous tumor. CT scans are already scary IMHO if you already have cancer. MRI may be safer option.

Apparently CT scan use a lot less radiation than they used to and are therefore much safer now.

Since the X-rays are being refracted rather than absorbed, will this technique make them much safer as well?

This title is a really bad representation of what the actual technology is. Instead, something closer to the actual research is: "X-ray elastography by visualizing propagating shear waves" which actually makes sense. Elastography in X-Ray is what is being described and not an imaging technique that is "clearer" than MRI...

How many % of the people not part of the field would understand what ‘shear waves’ or elastography is?

I'm not suggesting that the title be replaced with that, the current title is a blatant misrepresentation of the actual work.

It may not be clear, but "X-ray elastography by visualizing propagating shear waves" is the actual title of the paper. pavanagrawal123 shouldn't be getting downvotes.

How are these vibrations introduced into the body? And how do they affect the patient?

With MRI Elastography it’s a thing that looks like a ping pong bat, with a hose attached to a subwoofer. I’m sure their are better ways to describe it! It tickles if the patient is ticklish.

Depending on the technique, it could be something similar to an ultrasound vibrator (like those cheap ones they sell on aliexpress for cleaning your clothes) or to a phone vibrator.

Shouldn't, generally, be damaging to the patient.

I can't wait for how much they charge in the US for it. Hundreds of dollars? Thousands?!

MRI is actually not "clear" at all most of the time. People should be more aware of this IMO.

Both my wife and I were mis-diagnosed due to this, both with conditions for which it is impossible to spontaneously resolve. Mine required surgery with a 6 month typical recovery (which I declined), and would reduce my shoulder mobility permanently. My wife was told she would be in a wheelchair for the rest of her life in a few years due to an incurable spinal cord issue. We took the news pretty hard. I put my career on hold for a while, and we tried (and failed) to figure out how to live with this diagnosis, reading probably the entirety of scientific literature that exists on the topic and finding no reassurance. Six months later we got another MRI for her, which showed no "abnormality" and nothing to be concerned about.

And my "issue" resolved on its own in a few months as well, although I did not get another MRI - shit's expensive. The funny thing is, if the diagnosis was correct (SLAP tear in the shoulder, if you must know), such things do not resolve on their own.

So if someone tells you "you're gonna die" after looking at a blurry as fuck MRI, go get another MRI at a minimum, and seek a second or even third opinion before you take any of it seriously.

Reply on news.ycombinator.com