After collaborating with JP at the Field Museum on rendering CT scans a few years ago and seeing his article about this work in the spring MRCG newsletter, I was excited to see some images about this in person. JP has been working with CT scanners since 2006 starting out by taking advantage of the kindness of local hospitals and more recently renting a portable unit that came to museum on a truck.
As many of us know, CT scanners can look inside objects non-destructively and provide accurate images with 3D geometric accuracy. JP started the talk be reviewing some of the physics of getting a CT scan done, the benefits, and limitations. Here’s a run-down:
1. The scanner has a donut shaped gantry consisting of a steel ring containing the X-ray tube and curved detector on the opposite side, so your object has to fit within the imaging area inside the steel ring.
2. On each revolution you get lots of images scanned within 30 seconds to 5 min- this is very fast.
3. The biggest logistical challenge is moving objects to and from the hospital safely.
4. During the scanning you immediately get slices, which are cross-section images from three different directions. Volumetric rendering is done from the slices and there is free software for this.
5. Apparently it is relatively easy to do segmentation, segment out regions of interest, and extract wire frame models, just time consuming. From there you can get images of the surface and texture and can even print the models. It is relatively easy to go from slice to wireframe, but harder to achieve a manufacturing mesh to produce a 3D print, which can be expensive in comparison to traditional molding and casting.
6. PROs of scanning and printing: there is no contact with the object, complex geometry is not a problem, the scans and volumetric rendering are dimensionally accurate, you can print in lots of materials; prints can be scaled to make large things handleable or small things more robust for handling or increase visibility; subtractive manufacture, in which you can use a computerized milling machine to cut out a positive or negative, is also a possibility.
7. CONs of scanning and printing: printing is slow, the build volume is limited, a non-traditional skill set is required of conservators to produce the final product, and only a few materials age well. The best material is sintered nylon, extruded polyester may also be safe, but it doesn’t take paint well; it is hard to get the industry to think about permanence.
The object at the center of this project was a Magdalenian skull. The skeleton itself is of considerable importance, because it is the only magdalenian era skeleton of almost completion. A little history: it was excavated, quite professionally, in 1911 when they lowered the floor of the site. Unfortunately the burial was discovered when someone hit the skull with a pickax. Needless to say, the skull did not come out in one piece. In 1915 the full skeleton was removed in two blocks. My notes are a little fuzzy here, but basically at some point between the excavation the skull was restored and then went from being 2 pieces to 6 pieces, as it is documented in a 1932 publication by von Bonen. It appears that at that point the skull was also skin coated with plaster. Thankfully (?) those repairs have held up. Great, so why, did they need to scan and reconstruct the skull? Well according to Dr. Robert Martin, JP’s colleague at the Field Museum, the skull doesn’t look anatomically correct. Apparently during the time period when it was put together there was an interest in race and the skull fragments could have been lined up incorrectly accentuating cultural assumptions.
A previous x-ray showed that two fragments in the forehead are secured with a metal pin. In 2012, when the mobile CT scanner came to the museum, they were all geared up to start with the Magdalenian skull. Unfortunately there was not much difference in attenuation between bone and plaster making it tricky to define between the two materials in the scans. JP consulted a cranial reconstruction group and asked them to pretend this was a pediatric car crash victim with a cranial injury; they asked, why aren’t you using the mimics software package?
JP and his team also imaged the skull with a micro CT scan that has a 0.1 mm resolution versus the normal modern setting of 0.3 mm. They had previously identified 36 fragments of bone from the previous scan. It was hard to tell if some of those separations were just cracks or actual breaks between fragments. The hope was that the micro CT scanner could better define these areas. The micro CT scanner works opposite to the industrial/medical scanner. As you can see in the image to the left, the tube and detector are fixed, while the sample is rotated. Other differences are that it is slower, one scan takes 30-90 minutes and because of scanner geometry the skull had to be imaged in two scans . Because of this, JP used the previous scan to mill out a contoured support to hold the skull in the exact position. JP noted that digitally filling in the holes of the skull to create the support was the most time consuming part of that process and suggests using different radio-opaque marker dots to identify left and right for orientation during the later stitching process. With the new scans at least three separations were identified as cracks vs. breaks.
Now for the virtual reconstruction… the biggest obstacle in this stage was how to achieve something more anatomically correct using the virtual fragments when they have no boundaries. The fragments don’t push back in the computer- and the fragments can easily move into each other. With the software JP used mostly the translation and rotation functions and the free animation software Blender (which has a high learning curve and took several days to get accustomed to) to create hierarchical parent child relationships between the fragments as he joined them together. Just like putting a vessel together, right? In the virtual world at least there is no worry about lockout. They had a 3D printed of the final skull reconstruction and had an artist do facial reconstruction, which JP thinks always look related to Jean Luc Picard… So how successful was this? From a conservation perspective- awesome, it’s fully reversible! Scientifically though, it’s decent, well documented and scientifically justifiable- However, someone else could go through the same process and come up with a different reconstruction because of their reliance on left right symmetry for this reconstruction…
So what did I take away from this talk? This was a very cool project and if I have a question about CT scanning and 3D renderings, I will call JP! The scans can be extremely informational and there seems to be a lot of potential in their use for mount-making, crates, and storage, and possibly virtual reconstructions. Hopefully at some point in the future the software will become more intuitive and easier to use so that more of these types of projects can be done.