Landmark precision and reliability and accuracy of linear distances estimated by using 3D computed micro-tomography...

The wider availability of non-destructive and high-resolution methods, such as micro-computed tomography (micro-CT), has prompted its use in anatomical and morphometric studies. Yet, because of the actual scanning procedure and the processing of CT data by software that renders 3D surfaces or volumes, systematic errors might be introduced in placing landmarks as well as in estimating linear distances. Here we assess landmark precision and measurement reliability and accuracy of using micro-CT images of toad skulls and the TINA Manual Landmarking Tool software to place 20 landmarks and extract 24 linear distances. Landmark precision and linear distances calculated from 3D images were compared to the same landmarks and distances obtained with a 3D digitizer in the same skulls. We also compared landmarks and linear distances in 3D images of the same individuals scanned with distinct filters, since we detected variation in bone thickness or density among the individuals used.

We show that landmark precision is higher for micro-CT than for the 3D digitizer. Distance reliability was very high within-methods, but decreased in 20 % when 3D digitizer and micro-CT data were joined together. Still, we did not find any systematic bias in estimating linear distances with the micro-CT data and the between-methods errors were similar for all distances (around 0.25 mm). Absolute errors correspond to about 6.5 % of the distance's means for micro-CT resolutions and 3D digitizer comparisons, and to 3 % for the filter type analysis.

We conclude that using micro-CT data for morphometric analysis results in acceptable landmark precision and similar estimates of most linear distances compared to 3D digitizer, although some distances are more prone to discrepancies between-methods. Yet, caution in relation to the scale of the measurements needs to be taken, since the proportional between-method error is higher for smaller distances. Scanning with distinct filters does not introduce a high level of error and is recommended when individuals differ in bone density.