Pitfalls at the root of facial assessment on photographs: a quantitative study of accuracy in positioning facial landmarks

In the last years, facial analysis has gained great interest also for forensic anthropology. The application of facial landmarks may bring about relevant advantages for the analysis of 2D images by measuring distances and extracting quantitative indices. However, this is a complex task which depends upon the variability in positioning facial landmarks. In addition, literature provides only general indications concerning the reliability in positioning facial landmarks on photographic material, and no study is available concerning the specific errors which may be encountered in such an operation. The aim of this study is to analyze the inter- and intra-observer error in defining facial landmarks on photographs by using a software specifically developed for this purpose. Twenty-four operators were requested to define 22 facial landmarks on frontal view photographs and 11 on lateral view images; in addition, three operators repeated the procedure on the same photographs 20 times (at distance of 24 h). In the frontal view, the landmarks with less dispersion were the pupil, cheilion, endocanthion, and stomion (sto), and the landmarks with the highest dispersion were gonion, zygion, frontotemporale, tragion, and selion (se). In the lateral view, the landmarks with the least dispersion were se, pronasale, subnasale, and sto, whereas landmarks with the highest dispersion were gnathion, pogonion, and tragion. Results confirm that few anatomical points can be defined with the highest accuracy and show the importance of the preliminary investigation of reliability in positioning facial landmarks.