After receiving ethical approval (reference no: B32220 1316317) from the Ethical Review Board of the University Hospitals Leuven, Belgium, 50 healthy Caucasian subjects (25 males, 25 females) were recruited for the study. All participants provided written informed consent. The exclusion criteria were maxillofacial abnormalities and excess facial hair, which might disguise anatomical landmarks.18

Participants were seated in a natural head position with relaxed lips and a closed mouth. Soft tissue anatomical landmarks were then marked on their faces directly following inspection and palpation for linear measurements (Table 1, Fig. 1). Direct anthropometry acted as a gold standard for evaluating linear distances in relaxed facial expressions in comparison to indirect techniques. Pupillary distance was assessed directly without placing any marks.

Two sets of 2D and 3D pictures images were collected from all participants. The images were taken with participants in a relaxed position, with a closed mouth and a normal smile. Before the images were captured, subjects were positioned in a natural head posture.20 After capture, the images were checked, and image acquisition was repeated if they were blurred, too bright, not well oriented, or missing a part, or if the subject had his/her eyes closed. The indirect techniques utilized to capture images included 2D, SPG, and LS imaging systems.

In a clinical setup, 2D photographs were acquired (Fig. 2A) with a professional camera (Nikon D800; Nikon Corporation, Tokyo, Japan). The image capturing process was standardized by utilizing a reproducible set-up of a camera, tripod, and chair with a fixed distance to the camera lens. An L-shaped ruler was placed to the upper right of the individuals for calibration. For each subject, a total of 5 relaxed non-smiling pictures and 5 pictures with a normal smile (frontal, left oblique, right oblique, and lateral) were acquired. Thereafter, images were exported to Mirror® Software (Canfield Scientific, Inc, NJ, USA) in JPEG format for evaluation.

All subjected were scanned twice (relaxed position and normal smile) utilizing Planmeca ProFace™, which is a 3D facial scanning protocol integrated within the ProMax® 3D Mid (Planmeca OY, Helsinki, Finland) CBCT machine. Before the image was captured, each participant's head was fixed and the chin was placed on a chin-support to avoid distortion. Volunteers were asked to remove any objects (glasses, jewellery) that might potentially cause artefacts from light reflection during the scan of the head and neck region. All images were saved and exported in the OBJ file format to Proplan CMF® 2.1 (Materialise HQ, Leuven, Belgium) for assessment (Fig. 2B). Thereafter, a quantitative and qualitative assessment was conducted of all 2D and 3D images after the observers were trained and calibrated. All observations were performed under standardized conditions on the same computer.

Three-dimensional stereophotogrammetric images with a relaxed facial expression and normal smile were acquired using a portable 3D Vectra® H1 (Canfield Scientific Inc., Parsippany, NJ, USA) camera after calibration of the device. This 3D camera consisted of a stereo-optic system with ranging lights for easy patient positioning and had an on-board modular intelligent flash unit. It provided images with a 0.8-mm geometry resolution and had a 165 mm×270 mm×100 mm (x, y, z) capture volume and a capture time of 2.0 ms.21 Three pictures were taken for each participant, which were than stitched into a single 3D image utilizing VECTRA® Face Sculptor® software (Fig. 2C). For each capture, subjects were asked to remain stable. The first image was captured by holding the camera at the patient's chest level (30 cm below the mid-face and angled upward) and positioning it at a 45° angle from the front towards the right side of the face. The second image was acquired from the front, with the camera held at level of the patient's nose and positioned directly in front of the patient. The third image was captured for the left side by following the same protocol as used for the right side. After stitching, images were exported in the OBJ file format to Proplan CMF® 2.1 software (Materialise HQ, Leuven, Belgium) for further evaluation.

Two observers identified soft tissue landmarks and performed linear measurements on subjects' faces directly (as an objective gold standard) and both linear and angular measurements indirectly on the images captured using 2D and 3D imaging modalities. Direct anthropometry involved the measurement of linear distances on patients in a relaxed, closed-mouth position using a digital sliding calliper in a well-illuminated room. In contrast, both linear and angular measurements were performed after identifying soft tissue anatomical landmarks on 2D and 3D images (Table 1, Fig. 1). All images were landmarked and assessed twice by each observer at an interval of 4 weeks to determine inter-observer and intra-observer reliability.

For the qualitative analysis, 10 dental observers and 1 expert (subjective gold standard) experienced in facial imaging completed a questionnaire (Table 2) related to facial characteristics on 2D and 3D images twice at an interval of 4 weeks for calculating inter-observer and intra-observer reliability. Respondents used a 5-point Likert scale to assess their confidence level for each given answer (1: not confident at all, 2: somewhat confident, 3: neutral, 4: confident 5: very confident). Figure 3 illustrates the flowchart for both the quantitative and qualitative assessments.

Data analysis was performed utilizing S-Plus 8.0 for Linux (Tibco, Palo Alto, CA, USA). Inter-observer and intra-observer agreement for quantitative data was estimated using intraclass correlation coefficients (ICCs), where <0.40, poor; 0.40–0.59, fair; 0.60–0.74, good; and 0.75–1.00, excellent.22 A bootstrapped sample was acquired, and the overall standard error was calculated for the quantitative analysis of linear measurements with direct anthropometry as the gold standard, and for analysing angular and linear differences between all indirect imaging modalities.

For the qualitative analysis, the Fleiss kappa index was applied to assess agreement between and within observers (<0.00, poor agreement; 0.00–0.20, slight agreement; 0.21–0.40, fair agreement; 0.41–0.60, moderate agreement; 0.61–0.80, substantial agreement; and 0.81–1.00, almost perfect agreement).23 Confidence scores were compared between imaging techniques using a linear mixed model with the observer as random factor. Furthermore, the observers' scoring was compared to that of an expert and the significance of differences was calculated. A P value of <0.05 was considered to indicate statistical significance for both the quantitative and qualitative data.