Recent studies have reported that quantifying symphyseal and auricular surface curvature changes on 3D models acquired by laser scanners has a potential for age estimation. However, no tests have been carried out to evaluate the...
moreRecent studies have reported that quantifying symphyseal and auricular surface curvature changes on 3D models acquired by laser scanners has a potential for age estimation. However, no tests have been carried out to evaluate the repeatability of the results between different laser scanners. 3D models of the two pelvic joints were generated using three laser scanners (Custom, Faro, and Minolta). The surface curvature, the surface area, and the distance between co-registered meshes were investigated. Close results were found for surface areas (differences between 0.3% and 2.4%) and for distance deviations (average <20 lm, SD <200 lm). The curvature values were found to be systematically biased between different laser scanners, but still showing similar trends with increasing phases/scores. Applying a smoothing factor to the 3D models, it was possible to separate anatomy from the measurement error of each instrument, so that similar curvature values could be obtained (p < 0.05) independent of the specific laser scanner. Recent studies have shown the benefits of quantitative methods using 3D laser scanner models in addressing fundamental issues in physical and forensic anthropology. Sexual dimorphism and population and ancestry variation have been investigated quantifying surface areas or extracting curves (1–5). The morphological features of the symphyseal and the auricular surfaces used for age estimation have been examined looking at the surface curvature changes (6–8). Laser scanners have also been used to investigate cranial facial variation and for facial identification (9–13). In all these applications, the precision and the repeatability of the measurements among different instruments are essential for the reliability of each method; in fact, there are different models of laser scanners and differences between the software used for postprocessing the scans (i.e., aligning, merging, and fusion of the single scans to create a 3D model). It has been demonstrated that the surface areas were reproduced with high precision and the measurement errors of the extracted information varied from 0.2% to about 1% (5,14); it has been also shown that the location of points on the surface (landmarks) and the measurements of length could be accurately repeated on 3D models (11,12,15). When some of the parameters used for the scans made by the same laser scanner have been changed, measurement error was reported to increase slightly to 2% (14). However, all these studies have tested the repeatability of measurements made by the same instrument. Laser surface scanning (and CT-scanning) of bones is proposed as a method to make osteological data more widely and easily available, for example, the Smithsonian 3D collection (16), " Digitised Diseases " (17) and " From Cemetery to Clinic " (18). It is thus important to investigate whether 3D models acquired by different laser scanners may exhibit larger differences than those acquired by the same scanner. Irrespective of the performance of each laser scanner, the nodes of the resulting 3D model surfaces represent the true anatomical shape, plus some error due to measurement uncertainty. The formal expression of this uncertainty is difficult, as it requires full knowledge not only of the instrument and object properties, but also of operational details which may not always be available, such as calibration steps followed and the exact distance between object and scanner (19,20). A simple but general approach to assess measurement error that does not rely on any external knowledge beyond the scanned 3D mesh would therefore be particularly convenient. We assumed the measurement uncertainty manifests itself exclusively as independent random error in the position of mesh nodes, which can be effectively reduced by a smoothing operation over neighboring nodes. If this assumption is appropriate , the use of an adequate amount of smoothing would filter out the measurement error while preserving the overall anatomical shape of the bone surface and homogenize 3D models from different instruments. The aim of this study was to experimentally verify the validity of our assumption using 3D models of sym-physeal and auricular surfaces generated with three different