Towards an inexpensive open-source desktop CT scanner | Hacker News Reader

Towards an inexpensive open-source desktop CT scanner

tricorderproject.org

128 pointsmatthewbadeau41 comments

A topic I can speak intelligently about! (I am a Radiologist.)

This article is a great discussion of modern imaging technology and I 100% agree that MRI is incredible. But CT has become much more impressive than the single-slice scanners of yore.

Watch this video of a modern scanner rotating at full speed:

http://www.youtube.com/watch?v=CWnjqeB7Mk8

Currently, the race to improvement is about speed and lower radiation dose. The newest scanners can take hundreds of 2D slices in a single rotation of less than a second and reconstruct the image with a huge variety of algorithms accounting for body part, patient size, etc. that all allow for lower dose and noise resulting in better diagnostic quality.

MRI is truly the cutting edge of human achievement. Watch this video to see the forces involved:

http://www.youtube.com/watch?v=6BBx8BwLhqg

And this for a detailed explanation of MRI principles:

http://www.youtube.com/watch?v=pGcZvSG805Y

Medicine uses 1.5T and 3T routinely for imaging, but higher field strengths are in use for research purposes. Innovation lately has been about developing new sequences (ways to differentiate types of tissue), and increase speed of aquisiton and comfort.

So, if we're late 20s, early 30s, what's the risk in taking a head CT?

Off topic but what would you recommend for brain imaging, a CT scan or MRI?

CT is a great screening exam because it is cheap, quick and more readily available. MRI is much more specific in diagnosing problems in the brain and has a lot more 'features' that can be added to aid in diagnosis. So generally, everyone gets a CT and then moves on to an MRI if there is an abnormality... or if they fail to find an abnormality when the clinical exam remains abnormal/severe. There are a lot of exceptions... like in little kids, you should always avoid radiation when possible and people will go straight to MRI.

Why is there such a big difference in patient price between a CT and MRI scan?

My doctor told me to have a CT scan done but I am worried about the radiation. I feel that if an MRI was cheaper, then I would have been prescribed an MRI.

The MRI scanner requires more precautions and safeguards (being built around a 1+ tesla magnet), and acquiring images takes more time, thus more human resources.

Edit: to illustrate ris' point (the sibling comment), check this XKCD chart, linked in the article: http://xkcd.com/radiation/

The equipment and maintenance is much more expensive for MRI. The liquid helium bill alone for an MRI machine is not something you want to pay for.

Ask them about the radiation dose in banana-equivalent, then be less worried.

You're confusing CT scans with conventional x-ray images. CT scans deliver the radiation equivalent of hundreds of conventional x-ray images. The radiation dose is high enough to epidemiologically significant (meaning higher cancer rates in people who have been scanned, especially children).

Believe me, I'm not confusing them - I just think it's significantly less risky than the parent does. And yes, there can be higher cancer rates in people who have a _lot_ of CT scans done e.g. because of a serious illness.

I'd rather get that back to me as a percentage compared to a Phone Booth filled with nuclear waste (but only 75% of the way up) while standing 13.43 meters away with the humidity levels at 95%.

To expand a bit on what Shinkei has said, and mostly this applies outside the brain as well:

CT is basically xrays in the round. You'll nominally use a radon transform (the actual approach is usually a back projection algorithm) to reproduce 3D position from a number of individual images. There are many good properties of this, it is relatively fast and cheap, it has very good spatial accuracy (e.g. a distance you measure on the image corresponds to a distance in the subject), but you have very little flexibility with tissue contrast. If an x-ray won't differentiate the features you are looking for, mostly you can only hope to add a contrast agent (typically expensive, fiddly, and more dose) where that is applicable.

By comparison, an MRI is more of a programmable imaging machine. You can get a lot of different contrasts by use of different pulse sequences, so you may be able to see features that are completely invisible in CT. On top of that, you can do interesting things with the time dimension and how things are changing while you image. This leads to approaches like functional MRI mapping which most people have heard of, but also diffusion and perfusion imaging. This can give you an idea of what is happening functionally, particularly in the brain. Downsides: time consuming (hence expensive), poor spatial accuracy, artifacts.

So for diagnostics or surgical planning, you get a lot more information from an MRI, but this also can take a lot more effort to interpret and understand.

Wow... that is some scary stuff. The spinning sensors make me really weary of putting my head / body in there.

Well, the momentum of the pieces means that if anything goes flying off, it will travel in a straight line i.e. outward, away from your head :)

The machine is entirely closed up when patients are in the mix--you don't see any spinning.

On a related note, there are a couple of good open source projects for reading the standard imaging files (DICOM). DICOM itself is an interesting mishmash of a standard, but is pretty powerful. Images are shuffled around using a PACS (Picture Archiving and Communication System) which lets the scanning machines send their images to a central location at the clinic, the clinic to send images to your radiologist, your radiologist to send them to the hospital, etc etc. Interesting stuff.

http://www.osirix-viewer.com/Downloads.html OsiriX is a DICOM viewer and PACS for Macs and iOS

http://www.dcm4che.org/ dcm4che is a collection of server based applications for DICOM and PACS. Sort of like OsiriX, but web based. It's Java Tomcat based and reasonably simple to setup.

also useful - 3D slicer: http://www.slicer.org/

This is a cross platform viewer that is fairly capable (although very limited in 3D) and has some dicom communication capability.

Shameless self promotion: MedXT (YC W13) is a cloud RIS/PACS with HTML5 viewer. We provide private DICOM endpoints with unlimited storage and archiving.

https://www.medxt.com

We're currently in private beta. If interested, I'd be happy to hook up HN friends. Contact info in profile.

I use pydicom[1] quite a lot. It is an open source, pure Python module for reading & writing DICOM files. Works very well by essentially transforming the whole DICOM file into a big dictionary.

[1]https://code.google.com/p/pydicom/

pydicom is nice but it lacks network functionality: find, send, store and so on. And it cannot decompress jpegs. Two large open source toolkits are dcmtk[1] and gdcm[2]. Both comes with commandline applications for viewing dicom files and querying pacs servers. gdcm is bundled with itk[3].

[1] http://dicom.offis.de/dcmtk.php.en

[2] http://sourceforge.net/projects/gdcm/

[3] http://www.itk.org/

It is a beautiful little machine, looking forward to seeing it in action.

A simpler approach to get a small CT scanner is to use a stationary source and detector. This requires a cone-beam source, which might cost a lot more, but it makes it a lot easier to experiment with phase-contrast and darkfield imaging, which can improve soft-tissue contrast.

For reconstruction from the projections the Cph CT Toolbox https://code.google.com/p/cphcttoolbox/ contains open source implementations of several reconstruction algorithms. It is written in python.

From my viewpoint, CS student focusing on medical imaging, no single imaging modality will rule them all. It is much more interesting to look at the combinations like PET/CT and PET/MR that gives us complimentary information.

(from following link...) Ah, an old CRT screen as a detector - hadn't thought of that.

This is a nice hack! It's unfortunate that medical CT scanning remains expensive due to reasons other than the cost of designing and manufacturing one of these devices. It's about image acquisition speed, safety, quality assurance, safety, rigorous testing, and of course safety.

While all those things (+ personnel costs) are of course worth paying for, I find it unbelievable that a simple scan could cost $5000 in the US! When I had to get a CT scan, it cost $2000, of which 90% was covered by the mandatory minimum health insurance. Nobody gets left out in the cold.

Just fyi, in Paris you can get an MRI (they call it an IRM)(or a CT Scan) at a private clinic with fairly-modern e.g: GE machines (don't know how old) for 180 euros (including an assessment by a radiologist). Ahead of an MRI, you additionally pay a pharmacy 20-50 euros for injections they'll administer to you during the MRI, depending on which type of scan you're getting (your weight), etc. You leave with a DVD that summarizes your results and a report. I imagine the fees are subsidized by the government, but I actually don't know if it is since it is a private clinic I'm talking about. I imagine the technicians, doctors, and assistants are likely paid less in France.

Average radiologist in the US makes roughly $315,000 (Medscape 2011); 1/3 earn over $400,000.

In France that would be closer to $100,000 (70k euros) typically.

The radiology fee is separate from what it costs to be scanned (usually). CTs are pretty cheap though, depending on what you get scanned it's usually $200-500. Not sure what happened with the OP, perhaps it was an ER or something where the prices have little to do with anything based in reality.

FYI, in the US, the thousands-of-dollars list price for an MRI turns in to a few hundred if Medicare or an insurance company are paying (and no, I'm not talking about what the patient gets charged out of pocket--the actual total amount charged is reduced that much).

It also turns into a few hundred if you tell them you can't pay thousands of dollars, but are willing to take any free slot.

MRI is a classic example of high fixed cost, negligible variable cost. Once they buy the machine, it pays to make sure it never runs idle.

Makes me glad I live in Canada. I've had several CT scans since 2007 as part of some post-cancer monitoring, and I haven't had to pay a cent.

I wondered about that. My wife has a rare brain disease and has had multiple MRI's but I've never had more than a $30 co-pay. I'm sure that if it really cost the insurance co $5k, then I'd be paying a percentage of that cost instead of a fixed amount.

Part of the reason scans cost so much is that medical imaging businesses are essentially a cartel in many states, thanks to so-called "CON" laws. In states where these laws are in effect, you can't just buy a CT scanner and set up your own medical-imaging business; you have to get a "Certificate of Need" first that certifies that there's unmet demand for imaging services in the community. To get this certificate, you have to go to a local committee, whose membership is primarily made up of the existing medical-imaging businesses. This effectively allows the existing businesses to form a cartel in restraint of trade to keep prices up, a cartel which, if it were not protected by specific state law, would almost certainly be in violation of the Sherman Anti-Trust Act.

This is only one example of how the medical industry has been ripping Americans off. And, before you say anything, PPACA does nothing to bring this system under control; in fact, it perpetuates and extends it.

Hah. I was recently considering the feasibility of making a homebrew CT scanner, only I was thinking of using one of the cheap chinese dental xray machines you can get on ebay for ~$200. (I think I would have probably vacated the room while it was in operation. I'm not saying I don't trust cheap chinese x-ray technology but - well , actually - yes I am...)

Would have required an awful lot of scanning of xray film and would probably only have been able to image something around the size of a walnut.

Here's another home made CT scanner project someone did, not as polished, but he does a great job explaining the principles involved.

http://www.youtube.com/watch?v=hF3V-GHiJ78

> Safety is very important to me. In medical diagnostic imaging it’s often important to have an image as soon as possible, but that’s not the case for scanning non-living objects purely for scientific or educational interest...[I] make use of a “check source”, which is an extremely low intensity source used to verify the functionality of a high-energy particle detector.

Thank you for doing this. It's easy to build something cool but extremely unsafe in your home, forgetting that medical devices are carefully controlled to ensure patient safety. All it takes is one irresponsible DIYer to cause problems for everyone.

Frog levitating in 10 Tesla

Off topic question: What kind of cardboard is that used in the prototype? I have been searching for such a material for some time.

It's not cardboard. It's laser-cut wood.

...and BTW, is anyone going to point out the obvious hazards of DIY x-ray dosing? Like, okay, sure, you can perform CT scans on inanimate objects in the privacy of your own home, without subjecting living creatures to x-rays, but still... open source x-ray machines?

Consider the infamous Therac-25:

https://en.wikipedia.org/wiki/Therac-25

There's a big difference between imaging and treatment radiation.

It's actually laser cut Medium Density Fiberboard (MDF). There are several types, and only some are good for laser cutting. The good types include various fiberboards of a tan color, and if you scratch the edge with your fingernail, you should be able to loosen (and observe) individual fibers. I have purchased pre-cut sheets of this material at Home Depot in the US (midwestern and western US). They also sell a type which has a flat wood pattern printed on one side, intended as "wall board". It is cheap - 12-18USD for a 4x8' sheet.

The type that is used in clipboard backers is NOT good for laser cutting. Although it is also called MDF, it is dark brown, sometimes semi-gloss on the surface, heavier, and more structural. Unfortunately, the glue used to bond and seal it soaks up laser cutting energy and makes it very difficult to cut without significant char.

If you're looking for small amounts of laserable material for testing, laserbits is not a bad source (http://www.laserbits.com/index.php?main_page=index&cPath=82_...). However, in general I would recommend finding it locally at a lumberyard or home improvement shop as the quantity and pricing are better, and they will typically also cut it for you for a low price. Another source for this kind of material (at least in the US) are university bookstores, who havea (expensive) supplies for architecture majors.

It looks like MDF.

http://en.wikipedia.org/wiki/Medium-density_fibreboard

An impressive project, truly some Hacker News!

Reply on news.ycombinator.com