It’s that time of year. The one when you’re desperately trying to think of present ideas for your nearest and dearest – maybe they’ve given you a few ideas, maybe they have hobby which requires a steady supply of supporting materials, or maybe you’re just doing booze this year. Just possibly you’ve bought yourself a 3D printer, and are desperately shoehorning printed presents into people’s stockings. So that’s how I came to be printing my brother’s skull.

The first step is image acquisition. Fortunately (for me), the target had previously been the victim of some radiographers, who had kindly taken the pictures for me. However, this left the problem of data exfiltration. Despite the assistance of an inside man (or woman), initial attempts were thwarted by password protection and threat of prosecution under the Computer Misuse Act (1990). Therefore, I persisted with more direct methods.

As an aside, for printing bone CT imaging is preferred. We need 3 dimensional images that provide a lot of contrast between soft tissues and bone, therefore the better soft tissue resolution of MRI is not needed.

Next comes segmentation, basically identifying and extracting the bits of the scan that we actually want to print. For this, I used the excellent piece of open source software 3D Slicer. It provides a simple interface to import and view DICOM images, the most common file type in this form of medical imaging.

Typically when people go for a scan, they’ll have multiple sets of imaging in one appointment. Each series can be optimised for looking at specific structures, or for looking at anatomy in particular planes. Because anatomy is tedious, we’ve simplified looking at the human body in 3 planes; axial, coronal, and sagittal. There is usually better resolution of images within the plane the series was taken, rather than in the axis of travel. Furthermore, the slice thickness – effectively distance between the slices of the images – can be thought of as the resolution of the image in that direction.

Basically, for the best images use the plane you’re most interested in for detail, with the series that has the smallest slice thickness, that’s correctly exposed/windowed for the tissue you’re trying to segment. There’s also the secondary concern that 3D printers are better at printing curves in the XY plane, rather than Z axis. Personally I went for an Axial 1mm slice thickness with bone windows.

I won’t go fully into the detail of the process of segmenting in 3D slicer, instead pointing to the documentation here. However, I had the most success with thresholding (to create a layer that was definitely bone), then using the ‘Grow from seeds’ option to widen the selection to cover the skull.

It’s worth remebering that we’re ultimately trying to print the thing, so I used a fair amount of smoothing, removed small floating pieces, and gave it a flat base and top. For skulls in particular, the sinuses are quite a challenge. Essentially they’re cavities with very thing walls of bones in quite a complex structure. I ended up simplifying the regions with a smoothing brush, while filling in the mastoid air cells entirely.

I don’t have too much more to add at this juncture, I used a fair amount of judicious trial and error, checking in the 3D view that I was making something that vaguely looked like a skull. There are AI-powered tools that apparently can segment these structures, but I didn’t have any luck with them. In any case, once you have something you’re happy with, it is a straight forward process to export it as a bog standard STL.

Next, import it into your slicer of choice. I used white PLA (of course), 15% gyroid infill, 0.2mm layer height, and organic supports, with a total print time of 31 hours. I have no idea if these are optimal, I’m new to 3D printing in general, but played it realtively safe as I didn’t want it to fail 20 hours into printing…

The initial few hours were spent in a state of nervousness, similar to checking on a new puppy to make sure it hadn’t urinated all over the floor or started to eat its own sick.

However, past the mid-face danger zone, I relaxed, and the Mk4S happily trooped on through the night.

With the printing finished, the only thing left was to clear away the supports, a process that involved tweezers, side clippers, and some suffering.

I guess posting the finished article counts as publishing identifying material, although I’m not sure how identifying it really is. Maybe those people who try to reconstruct how neanderthals look could give it a go. I’m happy with my results, and plan on presenting it with some snacks nestled securely within the vault.