• MicroCT scans
  April 13, 2011
• Preview Our Data!
  August 12, 2010
• Interactive watershed code!
  August 3, 2010

• View all entries

• Subscribe

MicroCT scans
April 13, 2011

One of our goals for the Parametric Skeleton project is to register together bone surfaces extracted using different digitization modalities. To this end we have acquired a set of Micro CT scans of the bones of a hand and a foot.

Why micro CT? A few reasons. We always strive to push the resolution and accuracy of our models to their highest. Additionally, we were interested to see if the improved resolution of these scans would also translate to improved segmentation of the bony tissue using standard thresholding algorithms that we've discussed in a previous blog post.

So we hopped over to the STTARR facilities and requested some time on their micro CT scanner. They currently have a Locus Ultra microCT scanner from General Electric. This scanner has a similar frame to the regular clinical CT scanner, but its bore and scan bed are much much smaller. The field of view was set so that a few bones could be scanned at a time. In total, we obtained seven scans: four for the foot bones (tarsals, metatarsals, and phalanges) and three for the hand bones (carpals, metacarpals, and phalanges). Here are some example scans for the foot phalanges:

The top row shows the phalanges of the right big toe, while the bottom row shows the phalanges of the second toe, both on the right foot. The two images on the left show the original microCT image, while the two images on the right show the results of applying the basic thresholding algorithm. The scan volume has a resolution of 154 microns per pixel and the same for the slice separation. In contrast, most of the images we have posted on this blog come from volumes with around 750 microns per pixel and a 1mm separation distance between slices. Note that the bones were placed by hand on the scanner bed and are not in anatomical position.

At this resolution, it is evident that these smaller bones exhibit the structural architecture of long bones, with a thicker cortical shaft and spongy ends. Unfortunately, even at this small scale, the bone density is not enough to prevent spurious holes using a thresholding segmentation, as shown by the red arrows. This further supports our multi-modal approach of combining different digitization techniques to generate a more complete geometric model of the human skeleton.

The 3D images for the phalanges, as well as a subset of the tarsal bones (talus, calcaneus, cuboid and navicular), are shown above using volume rendering. Note the smooth articular surface of the talus (leftmost bone, right image) where the tibia sits, supporting the entire body. For comparison, the point set and reconstruction for the talus bone are shown below.

Previous