'What and how should we share?' An inter-method inter-observer comparison of measurement error with landmark-based craniometric datasets

摘要

The present study evaluates the precision and accuracy of photogrammetric 3D modeling of human crania in landmark acquisition and explores the limitations of combining datasets acquired by different observers and different measurement methods. Our working sample comprises 50 adult human crania, which were modeled with 3D photogrammetry. 3D coordinates of 56 landmarks were collected from the 3D models with Meshlab software and an existing corresponding dataset digitized with Microscribe-3DX has been utilized. Measurement error for landmark configurations and Inter Landmarks Distances (ILDs) for each type of landmarks has been assessed through least root mean squared deviation and mean absolute error respectively. Inter-observer error has been assessed on a sub-sample of 20 crania, which was also used for caliper measured ILDs. Between-methods Technical Error Measurement (TEM) based on ILDs has been calculated for evaluating the interchangeability for different datasets. Photogrammetric 3D models and Microscribe-3DX share identical rated accuracy regarding craniometric applications and both methods show increased accuracy in locating type I landmarks as opposed to types II and III. However, photogrammetric 3D models perform better in terms of inter-observer error suggesting higher reliability of measurements. Furthermore, ILDs are less prone to measurement error than landmark configurations. Finally, ILDs exhibit similar relative TEM of about 1.5% between Microscribe, caliper and 3D model based measurement methods. Combining datasets of landmark coordinates acquired from photogrammetric 3D models does not compromise the statistical integrity in terms of measurement error, which also applies to pooling ILD datasets from multiple methods. Nevertheless, compiling 3D datasets from multiple methods for 3DGM analysis should be done cautiously.