Repeatability of Phoneme-Guided Versus Conventional Acquisition of Smiling and Rest-Related Facial States Using 3D Stereophotogrammetry: A Within-Participant Study

Repeatable acquisition of smiling and rest-related facial states is essential for longitudinal three-dimensional (3D) facial imaging, yet the effect of task instructions may depend on the state and measurement. This within-participant study compared conventional (CONV) and state-specific phoneme-guided (PHON) acquisition in 162 adults. For SMILE, CONV elicited a posed smile, whereas PHON used /tʃ/ articulation followed by maintenance of the resulting perioral configuration; for the REST-related condition, CONV elicited relaxed rest, whereas PHON used sustained /m/ to define a closed-lip reference configuration rather than unconstrained physiological rest. Each protocol–state condition was recorded twice. Within-session repeatability was characterized by absolute error between repeated acquisitions, technical error of measurement, intraclass correlation coefficient [ICC(2,1)], and Bland–Altman agreement. State-specific protocol differences were assessed using paired Wilcoxon signed-rank tests with Holm correction. During SMILE, PHON reduced median error for the 3D Sn–Gn distance (3.79 vs. 2.36 mm), interlabial distance (3.48 vs. 1.84 mm), and sagittal lip step (1.33 vs. 0.66 mm; adjusted p < 0.001), but increased commissure-distance error (1.50 vs. 1.89 mm; adjusted p = 0.007). During the REST-related condition, no prespecified linear outcome differed after correction. Phoneme guidance changed, rather than uniformly improved, repeatability, and its effect depended on facial state and outcome.