Ask HN: How can I 3D print my skeleton for research?

Not OP, but perhaps minor skeletal deformities. Or serious injuries. For example:

> By using a combination of 3D printing and adult stem cells, Dr. John Szivek hopes this will help injured combat veterans, cancer patients, and victims of serious injuries.

> "We developed the idea that maybe we could put in a scaffold that regenerates bone very quickly by adding stem cells from the patient, their own stem cells and by adding some calcium particles and using a pattern that would cause the bone to grow really quickly," he says.

> The 3D plastic bone shaped frames called scaffolds are locked in place with two screws and a rod that can all be easily removed when bone forms over the scaffold.

> "After about six months your body tends to give up producing new bone and starts producing scar tissue so if we can cause healing to occur really quickly we can save the patient from having that problem," Szivek explains.

https://www.wthr.com/article/rebuilding-bones-3d-printing

This is really cool. I did not know it was possible to print metallic structures like this. Thanks for sharing!

> I don't think many doctors recommend this for the general public

> A CT scan comes with a strikingly high radiation dose

Another challenge is that they probably wouldn't give you a baseline scan from the bottom of your toes to the tip of your head. Abdomen/thorax and/or head/neck, probably. So you can't get the whole skeleton. A planar "scout" scan of two dimensions is normal examination procedure prior to the diagnostic scan. Those can often run a long portion of the body. You might be able to project something from that.

Aside: The Sievert (SI unit referenced in the XKCD comic) is a really interesting (IMO) unit. Instead of measuring the radiation in normal "amount of radiation" style, it measures the "equivalent dose" intended to help understand the ultimate biological impact of the radiation. The radiation emitted might be measured in Joules and the radiation absorbed might be measured in Gray (Joules/kg), but Sievert can be weighted by the tissue that absorbed the radiation. Your gonads or your head would have a greater weight than your ankle (or so I figure). This, I would guess, gives radiologists a much better way to consider the net impact of a diagnostic procedure in order to weigh against the overall diagnosis & prognosis.

Also, another aside: dose has become a more competitive feature for CT manufacturers in the last decade or so and several offer features to measurably reduce the dose.

I would first start our by using CT since soft tissue subtraction is a very easy task. This is because soft tissue has a certain density compared to bone (houndsfield unit [1].) You would maybe need to smooth for a good 3D print, but it's probably an easy starting place. Hell, a quick numpy script could easily do this. I would also poke at radiological scanning softwear (I use Radiant).

As far as data sources, I think you can probably find some open source CT scans to do this [2, 3]. There are PLENTY of chest and abdominal open source CT scans that exist. You may need to "mix and match" some body parts, but it wouldn't be too hard as long as you're using same gendered scans with similar height and anatomical features. Sort of the way they 'interpolate' for dinosaur bones.

Is this for a research project?

Good luck!

[1]. https://en.wikipedia.org/wiki/Hounsfield_scale [2]. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3043884/ [3]. https://www.radrounds.com/profiles/blogs/list-of-open-access...

So CT would be the best way to get a clearer picture of your own skeleton as high-density structures like bones stand out much better (Hounsfield units and all that). Then you'd have to segment the skeletal system and do a surface-extraction. I have a decent writeup on how to do that with Python and VTK under [1].

MRI might cut it, and unlike CT its not as radioactive, but typically in MR the contrast between soft tissue and bones isn't as pronounced so segmentation becomes trickier. I wrote a post on segmenting my own brain from a head MR I had done and as you can see from the pics the bones don't stand out [2]. Honestly, I dunno how easy it'll be to convince a doctor to write you a referral for a full-body CT scan given the potential health hazards.

Then comes the tricky part, no imaging modality or segmentation is perfect and you'll end up with a noisy dataset that won't play well in 3D printing. Prior to the surface extraction you'd have to post-process your segmentation to clean little islands of tissue, possibly even manually, and close holes in the segmented structures. A lot of that can be done automatically but you'll likely have to do some by hand. Medical image processing frameworks like ITK (and SimpleITK in Python) are your friends.

What I've found works very well (large part of my PhD was on medical imaging) is multi-modal imaging where you perform a segmentation on two datasets, e.g., MR and CT, registered to one another so you can best disambiguate between soft tissue and bone [3].

You'd then have to extract the surfaces into something like STL and post-process it in some software like MeshLab [4] to create a smooth mesh which you'll extract in a clean STL. That you can print but that's when you'll have to deal with the quirks of 3D printing. Depending on the filament you use overhangs are gonna become and issue (something about molten plastic and gravity I presume) so you may have to split your model into separate pieces that can be 3D printed (I think the skull will be particularly tricky) and then eg glue them together.

Before you go down that rabbit hole I propose you try and print a 'clean' skull that has already been segmented. I found [5] to be a fantastic resource of anatomical part models! Have a look at [6] on how to configure the interface to get the skull bones.

Good luck!

[1] https://pyscience.wordpress.com/2014/09/11/surface-extractio... [2] https://pyscience.wordpress.com/2014/10/19/image-segmentatio... [3] https://pyscience.wordpress.com/2014/11/02/multi-modal-image... [4] http://www.meshlab.net/ [5] http://lifesciencedb.jp/bp3d/ [6] https://imgur.com/a/uGqDA5Y