MRI to 3D Print Gets Much Faster

A surprising use of 3D printing has been in creating life-like models of human body parts using MRI or CT scans. Surgeons and other medical professionals can use models to plan procedures or assist in research. However, there has been a problem. The body is a messy complex thing and there is a lot of data that comes out of a typical scan. Historically, someone had to manually identify structures on each slice — a very time-consuming process — or set a threshold value and hope for the best. A recent paper by a number of researchers around the globe shows how dithering scans can vastly improve results while also allowing for much faster processing times.

As an example, a traditional workflow to create a 3D printed foot model from scan data took over 30 hours to complete including a great deal of manual intervention. The new method produced a great model in less than an hour.

One thing the researchers note is that the technique should be easy to adopt since it uses all open source software and existing image processing algorithms. There are some limitations, though. There are several things that limit the resolution and can introduce inaccuracies. For example, MRI intensity versus actual tissue appearance is highly variable based on the scanning machine’s settings and operator.

The researchers also note that advances in scanning technology will make even better 3D printed models possible. Naturally though, these prints aren’t coming off a $150 hobby-grade printer. The Connex500 printer used costs a cool quarter of a million dollars. It can print up to 14 different materials in the same job and has a reasonably large build volume (500x400x200 mm). That price, however, doesn’t include the water station to wash away support material, so budget accordingly.

We couldn’t help but wonder if you will one day have a bad part of your body scanned, printed, and then you’ll get the new part to replace the old. It seems like if you have a model of a body part, it would be just a little math to print a perfect cast, brace, or splint, too. But, then again, we aren’t doctors.

Photo Credit: Steven Keating and Ahmed Hosny/Wyss Institute at Harvard University

19 thoughts on “MRI to 3D Print Gets Much Faster

    1. The material is around 900USD/3.6kg and this print would probably be around 1.5kg (including the waste and support material). The water station was below 10k as far as I remember. The print time would probably be around 20hours. Source: Had an Objet Eden 260VS at my former workplace.

        1. It depends… On complex and fragile parts cleaning is very time consuming which increases the cost heavily. I can’t tell how much an hour would cost since it depends on how good you maintain the machine. If you don’t do maintenance a head can clog easily. A replacement head costs around 1250USD. (There are 8 heads with 8 nozzles each in one machine)

  1. I had some questions after reading this article, and I hope others will benefit from this post.
    For example, to print the foot as described above:

    1) Threshold and subtract all CT/Ultrasound/MRI data so that it removes air (easy). Now you’re left with the “data to print” (the foot).
    2) Map remaining data (organic material density) so that it spans 0-100% (not too difficult)
    3) use dithering to create infill data of your 2-head 3D printer (this is the novel bit).
    4) Setup your had a printer with flexible transparent material and white PLA.
    5) Print

    2) Mapping
    I assume the mapping should be linear, but it may not be.
    For instance, if organic material of (2) is represented as raw MRI/CT/ultrasound data with raw data values of, say, 233 to 478, the algorithm could map from 0 to 100%.
    The data set could determine how successful this step is in getting realistic looking prints.

    3) Dithering
    The Floyd-Rosenberg dithering will then convert the 0-100% data to print pixel of transparent material, OR print pixel of non-transparent material. You can get this plugin for programs such as

    4) flexible transparent material:

    5) If the resolution is high enough, this would then create the right print. So in theory you could do this at home, if you can get your hands on the data set.

      1. If the data uses the DICOM file format, in theory it shouldn’t matter where the data comes from.
        But the image source (CT/MRI) may have inherent differences that result in different mappings when they are used to represent for instance bone density. That is because CT and MRI measure different things when you want to see “the bone”.

    1. Thanks to Al Williams and to mime for followup :-)

      I then interpret from the comparative success as above, it should be possible and at amateur level to print a rubber glove which appears as a human hand flexible enough to pull over prosthetic prototypes either transparent or not. Interested to see if anyone has such experience and direct links to forums looking into this ?

      So far I can imagine a few interesting permutations which used to be on my back burner now foremost (eek) and could be the basis for an interesting and diverse project. I like to think this could be a great way for experimenters looking into robotic hands to add that extra touch ;-)

      1. It’s not entirely clear what you want.
        If you want to print only the skin as a glove of transparent material, then no, you can’t do this using the above. That is because you cannot separate the transparent layer from the rigid layer, because they are merged together.

          1. Perhaps what he could try is to keep transparent as transparent (the skin), but whenever there is solid (the bone), map that as dissolvable support as you suggest.

  2. So the basic current work flow is to take the MRI or CT data, work with radiologist to attain desired anatomy and define noise and other unwanted details with a technician to clean up later. The DICOM files are transferred to Materialize Mimics software which helps create the 3D file from all the slices and usually using a mix of Photoshop and Magics as well create the well defined 3d bodies to create an “Assembly” of Stls that will be loaded together in Objet studio or GragCAD print. The current flagship to do this process is far and away the Stratasys J750, which is their resin jetting unit that uses large inkjet heads to concurrently mix 6 materials with a soluble gel support in CMYKW and a clear or flexible for the 6th. This allows for clear outer tissue, vascular paths, bone, organs nerves etc all build into the same model/print. PS this is all cool and not cheap. Resin runs $330+ per liquid kg and that J750 is about $360,000 All in.

  3. “It seems like if you have a model of a body part, it would be just a little math to print a perfect cast, brace, or splint, too. But, then again, we aren’t doctors.”

    See: Invisalign

  4. And I wonder why there are so many mis or really mal-diagnosis? HHhmmm… just like an attorney racket with an “educated guess” or “professional opinion” (yeah, of course you want to make money off slow murdering who ever you can and do whatever you want unless there is threat of liability or retaliation… wonder why amnesic drugs and narcotics are pushed so much) when there are six sigma validated systems proven to function in extreme environments people can’t even survive.

    Someone needs to take the U.S. AMA , Bar Association and whatever U.N. treasonous war criminal traitors to military courts martial trial for practicing worse than Nazi Nursing and Medicine acting all innocent like they’re not committing death sentence able offenses. Including in remote sensing stations trying to remote communicate, brain wash, sonic assault, mind control and mass murder making everyone believe they’re telling the truth, including torturing people to believe. And I wonder why crime even exists? What a pan troglodyte industry trying to look smart pathetic archaic needing to go extinct like the evolutionary theory they preach and predators they are.

    Awesome 3D prints… about time. No wonder Open Water Company is struggling with ethical issues. Pathetic U.N. inbreds form countries where you can marry your first cousin.

  5. Reading the paper – it looks like this is only really usable in resin or jet 3d printers.
    Surely they convert regular models to raster layers when they slice already – rather than tool paths we’re used to in FDM.
    Could be wrong but I can see how one could use this to create a manifold .stl.
    Super cool though.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.