What secrets can tomography reveal about a Megalodon tooth?

The tooth itself is shown in Figure 1 – the photo was taken with a smartphone and the CT scan using a high resolution transmission tube system coupled with a 100 µm pixel pitch detector for catching a good amount of detail. The scan was made by TetraVision.

A couple of things are visible in the cross-sectional CT slices: lots of cracks are seen – this is to be expected for a fossilized sample that has undergone many extreme temperature changes over the millions of years of preservation. There is a visible enamel layer on the surface, but its brightness changes a lot across the sample and is in some places invisible. There are also some large channels that are interconnected and resemble a “root canal” network. Some of these channels are connected to the surface with holes seen on the surface on the front bottom of the tooth. In addition to this, some inclusions (bright particles) are found all over the tooth. These four components can all be visualized as shown in Figure 2 – the segmentation was achieved using Dragonfly’s Deep Learning. The deep learning process is achieved using the Segmentation Wizard tool – some input frames are provided and a 2.5D Unet model is trained based on this input. Some screenshots are shown in Figure 3. The typical work required to provide input in one or more input frames takes less than an hour, and the training process itself (with no operator dependance) can run for as long as 24 hours. In the example shown in Figure 3, the training was 10 hours in total. The speed of training depends on the computer specifications especially the machine RAM and GPU size, the size of the data itself, the choice of model type and amount of augmentation used. In this example, the dataset size was 8.5 Gb and the deep learning default parameters were used for a 2.5D Unet model on a machine with 64 Gb RAM and a Nvidia 3070 laptop GPU (8 Gb onboard memory).

• CT allows the non-destructive inspection of internal details of all kinds of objects, also Megalodon teeth
• Dragonfly allows imaging and inspection of cracks and internal channels, something not only useful for shark teeth but also for industrial CT applications
• The authenticity of such teeth can be verified by the natural internal features and evidence of canals, fossilization inclusions and cracks. This can be important if you are buying or selling Megalodon teeth. The same principle applies to other expensive and unique objects, that you might not want to cut open to evaluate its authenticity by viewing internal details.
• Dragonfly’s Deep Learning can segment complex datasets into separate components such as channels, cracks, layers and more. This was useful in the Megalodon tooth but it can also be applied to your manufactured parts to check for cracks, to check the location of internal channels or blockages of these channels, and much more.
• And finally: if you are an animal dentist, this can be useful to understand how to perform a root canal on a Megalodon shark, if you wanted to time-travel back a few million years and help out a Megalodon shark with some tooth-ache.