Inter-Observer Processing and Measurement Error Are Low for 2D Dental Measurements on Shared microCT Scans

Sharing micro-computed tomographic (μCT) scans of teeth increases data accessibility and reduces the need for repeated scans of any given specimen. However, the use of the same TIFF stacks or DICOMs by multiple individuals has the potential to introduce new sources of error. Here, we explore whether use of the same μCT scans by different persons produces comparable results.

Worn (N = 11) and unworn (N = 4) Cercocebus atys upper molars (UM1 N = 8, UM2 N = 7) were μCT scanned using a Bruker Skyscan 1172 High Resolution Ex Vivo Scanner at a resolution of 22 μm. Two individuals (K.N.G. and M.C.O.) created a 2D mesial slice for each TIFF stack (tooth). Worn teeth were reconstructed by K.N.G. and M.C.O. Three researchers (M.C.O., K.N.G., and J.R.) measured tooth shape, linear enamel thickness, average enamel thickness, and relative enamel thickness (AET and RET). Inter-observer percent error was calculated for each measurement. Univariate ANOVAs were calculated to evaluate variance due to slice maker, reconstructor, tooth, and measurer when percent error averaged > 5%.

For unworn teeth, error was generally low and largely due to the person doing the measurement. For worn teeth, wear reconstructor was a statistically significant source of variation for AET and RET.

We found that (1) inter-observer error was generally low, (2) linear measurements are prone to error, (3) worn teeth did not present an additional source of error as compared to unworn teeth, and (4) different people can use the same μCT scans to reliably reconstruct, slice, and measure teeth.