The present systematic review evaluated the accuracy and reproducibility of facial soft tissue landmark assessment across different three-dimensional face scanning modalities used in dentofacial practice. The review was conducted in accordance with PRISMA guidelines and registered in PROSPERO (CRD42025628750). Electronic searches of PubMed, Scopus, Web of Science, Cochrane Library, Semantic Scholar, and Google Scholar identified studies assessing facial soft tissue landmarks using 3D facial scanning systems and reporting accuracy and/or reproducibility outcomes. Eighteen cross-sectional studies were included in the qualitative synthesis, comprising 498 participants. The included modalities spanned photogrammetry, laser scanning, stereophotogrammetry, structured-light scanning, and smartphone-based depth/photogrammetry workflows. Comparator methods included direct anthropometry, manual measurements, electromagnetic digitizers, geometric references, 2D photographs, stereophotogrammetry reference systems, and cone beam computed tomography-derived soft tissue models. Accuracy was evaluated in 13 studies, while reproducibility or reliability was reported in all 18 studies. Most studies reported clinically acceptable landmark accuracy and repeatability under standardized conditions, with stronger and more consistent performance generally seen with stereophotogrammetry and structured-light systems. Smartphone-based systems also showed promising and often clinically acceptable results, although performance was more variable across facial regions, particularly in lateral landmarks. Reproducibility was typically better for well-defined midline landmarks than for less distinct lateral soft tissue points. Owing to substantial heterogeneity in scanner types, landmark protocols, comparator standards, and outcome metrics, meta-analysis was not performed and a narrative synthesis was undertaken. Overall, 3D facial scanning appears suitable for soft tissue landmark assessment, but careful device selection and standardized protocols remain essential.
Naikwadi et al. (Fri,) studied this question.