Hacker News Reader

Comment by jmhmd | Hacker News Reader

This is really cool technology, as a neuroradiologist anything that helps decrease time/cost and increase availability of imaging is a big win.

It should be clarified, though, that this machine uses a very weak magnetic field compared to traditional MRI, and while they are doing neat things to improve image quality, the resolution of their images is still far, far inferior to a standard 1.5 or 3 tesla magnet. This study does not compare to traditional MRI, just shows that it is feasible to deploy in a real clinical setting and that abnormalities can be detected. It should not be assumed that this cheap and portable technology can replace standard MRI for most indications.

That aside, I hope we get one at my hospital!

View on HN

You can see the lower image resolution in the JAMA Neurology paper.

It's not too shabby though, especially if you've looked at older MRI images enough.

I think part of their argument, at least implicitly, is that a lot of things don't necessarily need the resolution of standard MRI to be clinically useful, especially in places like ER settings.

My skepticism about papers like this with small Ns (maybe even some larger Ns) is how well they generalize to very unselected populations. It's sort of par for the course with early-stage medical products, even quality ones, but lots of times as the Ns increase and there's less control over the patient populations they're being used on, the patients become more heterogeneous, things don't work out quite as well.

Still good to see research in this area.

I'd have to look at the paper more closely but I wonder if this can be used for functional imaging as well.

Regarding functional imaging, I would be very surprised if they could get meaningful BOLD signal at that little field strength.

I may be MRI ignorant, but even if the Teslas are lower, if the radius is smaller (say, can only fit a head or arm or leg), does that make up for it?

I understand MRIs are largely a compromise. Organizations will buy only 1 or 2, and therefore buy one that will fit all but the most obese patients, to the imaging detriment of anyone/anything smaller. Or is that all wrong?

No. More teslas are needed for more signal to improve the signal to noise ratio. And there is a lot of noise with portable non-shielded MRI. The only way to compensate for this is by having bigger voxels (3-5mm per voxel rather than the standard 1-2mm), or repeat scans and perform averaging. The more drastic way is to do funky image reconstruction with compressed sensing or deep learning.

The wider bore does indeed come at a cost. You have more geometric distortion and a less homogenous magnetic field at the edges of the bore. Also, you will need stronger magnetic gradients to form image, causing more energy disposition (i.e. Heating the patient) and higher probability of peripheral nerve stimulation (PNS, involuntary muscle twitches due electric fields induced in the nerve caused by MR gradients).

To add to this comment, use of strong gradients also slows down scanning When you hit SAR limits. The SAR has to be brought down with pauses between scans, reducing gradient amplitudes, reduced resolution or shorter echo trains. It’s really painful.

The compromise for bore size is convenience in fitting things in it vs homogeneity of field, field strength, and cost. This is why standard bore 7T clinical imaging systems are pretty niche, but 9-12T mouse imaging systems aren't.

To get a good high resolution signal you need field homogeneity (and/or accurate mapping), high quality gradients, and good SNR. Field strength helps with the last part.

There are other trade offs too, for example imaging artifacts due to implants (or even being able to image them), susceptibility, etc. vary with main field strength.

>need field homogeneity (and/or accurate mapping), high quality gradients

It always confused me that MRI did millimeter scale imaging with radio waves that are meters long. I learned, and other's might be interested to learn, the trick is that it's the magnetic gradient that does the imaging. Each nucleus in the volume is tagged to it's place along the magnetic gradient axis by the local magnetic field strength shifting the emission frequency proportionally. That frequency shift is then inverted back to position for that axis.

Good point. It’s easy to forget this stuff isn’t obvious.

By comparison for head imaging, the Synaptive folks have pretty decent images at 0.5T, and can do DTI etc. Bulkier than this unit but still small, and much better IQ.

Do you think this would be useful for initial investigations for emergency suitations a car crash or trauma situations for example?

I did get bumped from my slot on a 3T MRI due to an emergency I a uk hospital recently due to some patients needing it more than me.

Long shot but do you think this would be sufficient to r/o tumor in someone with a first time seizure? What about assessing for hydrocephalus in peds with a shunt?

The images were a lot better than I was expecting from 0.064T

Reply on news.ycombinator.com