A study to determine the validity and accuracy of a low cost, widely accessible 3d facial scanner compared to an established sterephotogrammetry based system

Thomas, Kelly (2024). A study to determine the validity and accuracy of a low cost, widely accessible 3d facial scanner compared to an established sterephotogrammetry based system. University of Birmingham. M.Res.

Preview

Thomas2024MScByRes_Redacted.pdf
Text - Redacted Version
Available under License All rights reserved.
Download (2MB) | Preview

Abstract

ABSTRACT
AIMS
The aims of the in vitro study were to determine the validity and accuracy of two Bellus3D devices - Bellus3D Face Camera Pro (Android device) and Bellus3D and iPhone 12 (Apple device), and two commercial 3D facial imaging systems - 3dMD and Di4D SNAP. The aims of the in vivo study were to determine the validity and accuracy of Bellus3D Face Camera Pro (Android device) compared to Di4D SNAP whilst
capturing the face at rest and at maximum smile.

MATERIALS AND METHODS

For the in vitro study a plastic mannequin head was pre-marked with 35 facial landmarks and the 3D coordinates of each of the landmarks was found using a coordinate measuring machine (CMM). This produced the gold standard 3D configuration. The head was also scanned using four three-dimensional imaging systems i.e., Bellus3D Face Camera Pro (Android), Bellus3D Application (app) and iPhone 12, 3dMD and Di4D SNAP. For each device the image was digitised on-screen. The 3D landmarks configuration for each device was compared to the gold standard
CMM data following partial Procrustes superimposition. The Euclidean distance between the landmark pairs was measured as well as mean absolute difference in the x, y and z-directions. For the in vivo study 21 landmarks were placed directly on the faces of 30 volunteers
meeting the inclusion criteria. Participants rehearsed rest position and maximum smile. For each participant images were captured at rest and maximum smile, using the Bellus3D Face Camera Pro and Di4D system. The images were digitised and both the 3D landmark configurations at rest were aligned using a partial Procrustes superimposition, this was repeated for both 3D landmark configurations at maximum
smile. The Euclidean distance and the mean absolute difference in the x, y and z direction between each landmark pair identified, a threshold of 2.0mm difference was set as being clinically significant different. Systematic and random error was assessed. Repeatability of rest position was also assessed.

RESULTS
For the in vitro study a two-sample Students t-test showed the mean difference in Euclidean distance (0.9 ± 1.4mm) between the Bellus3D devices and the commercial systems was statistically significant (p = 0.001) with a 95% confidence interval for the
difference of -1.4mm to -0.4mm). For the in vivo study, images taken by Bellus3D Face Camera Pro and Di4D at rest,
showed that 18 of 21 extracted landmark pairs had a mean Euclidean distance of less than 2.0mm. Three landmark pairs; right and left gonion and menton were outside of the clinical level of acceptability (>2.0mm). The smallest differences were seen in
landmarks located close to the midline of the face. For maximum smile, an additional 4 landmark pairs showed a mean Euclidean distance of greater than 2.0mm. These were right and left exocanthion and right and left cheilion.

CONCLUSIONS
The difference in accuracy and precision between the systems is most evident in the lateral and inferior landmarks because of the centralised camera location of the Bellus3D Face Camera Pro system and the need for subject rotation. Further reduction in accuracy was observed in maximum smile as the longer scanning time of the Bellus3D system captured involuntary changes in micro expression. Bellus3D Face
Camera Pro is suitable for clinical application if limited to areas adjacent to the midline in rest position. Bellus3D Face Camera Pro is not suitable for clinical use where investigation of lateral and inferior areas of the face is required, nor for transverse smile investigation.

Type of Work:

Thesis (Masters by Research > M.Res.)

Award Type:

Masters by Research > M.Res.

Supervisor(s):