Trueness and precision of facial scan and virtual patient representation workflow

To investigate the feasibility and accuracy (trueness and precision) of facial scanning and virtual patient representation (VPR).

One participant was recruited and informed consent was obtained. VPR was performed 30 times with a custom fabricated intraoral scan body (ISB). Thirteen adhesive markers were added to the face as an extraoral scan body (ESB). Two facial scans were obtained for each VPR using an infra-red laser accessory sensor (Structure sensor; Occipital Inc) mounted on a computer tablet (iPad Pro; Apple Inc), including one with seated ISB and one without ISB. Two maxillary intraoral scans were obtained using an intraoral scanner (Omnicam; Dentsply Sirona) with and without the ISB. All files were imported to a dental software program (exocad; exocad GmbH) and VPR was obtained by aligning the facial and IOS scans using the ISB and ESB as common elements for alignment. Five fiducial face landmarks, four intraoral dental landmarks, and six perioral landmarks were selected for measurements. A total of 32 linear measurements, including 14 face-face (for facial scan accuracy) and 18 face-dental (for VPR accuracy) representing total face, lower face, and perioral regions, were performed directly using a digital caliper (FINO Digital Caliper; FINO GmbH) and virtually on the VPR. Trueness was evaluated by mean absolute difference (MAD) between the virtual and direct measurements, and the standard deviation represented Precision. Statistical analyses were performed with a statistical software package (IBM SPSS Statistics v25; IBM Corp), with α = 0.05. Data were analyzed for normality with Shapiro-Wilk test, and 1-sample t- (or Wilcoxon signed rank test), technical error of measurement (TEM), and relative error magnitude (REM).

The facial scan had 2.04, 1.66, 0.8 trueness, and 1.05, 0.92, 0.91 precision for total, lower face, and perioral regions. VPR had higher MAD (lower trueness) than facial scan, including 3.32, 2.40, 1.21 trueness and 2.2, 1.47, 1.2 precision for total, lower face, and perioral regions. Both TEM and REM were lowest for the perioral region and increased with increasing measurement distance.

Error in face scanning increased with increased distance and intricate details. VPR accuracy was lower than face scan accuracy because of added errors in the alignment process. The investigated VPR workflow might be feasible for treatment planning and smile design. However, it would be unreliable for more demanding prostheses manufacturing purposes.