Accuracy of standard craniometric measurements using multiple data formats

Midfacial growth patterns in males from newborn to 5 years old based on computed tomography

“…Numerous studies concur in the conclusion that physical and CT‐based virtual measurements of dry skulls are in very good agreement, in particular for CT data with a higher spatial resolution and a smaller slice thickness (Hassan et al, ; Hildebolt, Vannier, & Knapp, ; Olmez et al, ; Richard, Parks, & Monson, ; Richtsmeier, Paik, Elfert, Cole, & Dahlman, ; Waitzman et al, ; see Kim et al, for a comprehensive review of earlier literature). The results of many studies seem to suggest that the main source of error is the intraobserver or interobserver inconsistency in identifying landmarks or measuring the virtual skull model rather than artifacts produced by the CT scanner recording (Barbeito‐Andres, Anzelmo, Ventrice, & Sardi, ; Fuyamada et al, ; Kim et al, ; Kragskov, Bosch, Gyldensted, & Sindet‐Pedersen, ; Olszewski, Tanesy, Cosnard, Zech, & Reychler, ; Papadopoulos et al, ; Richard et al, ; Van Cauter et al, ; Verhoff et al, ). Importantly, it has been shown that the difference between virtual and physical measurements is higher for the variables which show worse repeatability when taken on a dry skull as well (Decker, ; Stull, Tise, Ali, & Fowler, ).…”

Integração morfológica e modularidade em crânios das espécies do grupo Rhinella granulosa

“…For the landmarks expected to have higher error in the 3D images because of their positions in the skull or because they were located in thinner bones, only landmarks 4, 8 and 12 corresponded to the expectation. All of them are type I landmarks, agreeing with the findings of other authors working with 3D images of more error in placing landmarks at some particular bone sutures [18,19]. MED and HIGH presented very similar results, indicating that there is no difference in switching the resolutions to visualize the landmarks.…”

“…The magnitudes of differences that we found for DIG and the micro-CT resolutions are similar to differences reported by other authors when comparing measurements taken with CT data and digital calipers (Richard et al 2014;Fernandes et al 2014), although the last authors did find a systematic bias in the CT data (all distances under-estimated). Yet, when comparing our results with other authors that measured the same specimens with a 3D digitizer (Polhemus 3Space) and CT (Corner et al 1992;Stull et al 2014), our error betweenmethods is much smaller (around five to ten times smaller).…”

“…When looking at the between-methods mean percentage error, the error in relation to the distances mean lengths, we can see that the cases around 10 % correspond to three small distances: 5 (frontoparietal bone), 13 (nasal bone) and 25 (parasphenoid bone), and just one longer distance: 16 (squamosal bone), which is composed of two landmarks detected with high deviations for DIG (landmarks 8 and 12). The magnitudes of differences that we found for DIG and the micro-CT resolutions are similar to differences reported by other authors when comparing measurements taken with CT data and digital calipers [18,21], although the last authors did find a systematic bias in the CT data (all distances under-estimated). Yet, when comparing our results with other authors that measured the same specimens with a 3D digitizer (Polhemus 3Space) and CT [15,19], our error betweenmethods is much smaller (around five to ten times smaller).…”

Error measurement in craniometrics: The comparative performance of four popular assessment methods using 2000 simulated cranial length datasets (g-op)

“…In the anthropological literature, and more specifically in the literature concerning skulls as relevant to craniofacial identification, there has been a wide range of error statistics employed and no consensus as to which is ideal. Additionally, there is a large disparity in both the reporting and analysis of measurement error in craniometrics, with some studies reporting only one statistic and others reporting upwards of four [14,30,39]. This study aims to explore the utility of all the above-mentioned error metrics and determine which one performs best across multiple simulated error scenarios and thereby should be the statistic of choice for future reporting.…”