Manuscript received February 5, 2026; revised April 13, 2026; accepted May 6, 2026; published September 10, 2026.

Abstract—A cleft lip is a common congenital malformation that often has a negative impact on patients. Surgery can improve patients’ quality of life, but the criteria for facial symmetry are unclear, so quantitative facial symmetry analysis is expected to make the surgery more effective. In this study, we propose a new evaluation method based on segmentation and registration using 4D data. In the segmentation step, a deep learning model divides the 3D faces into different regions, such as the upper lip and nose. In the registration process, the Coherent Point Drift (CPD) algorithm is applied to the regions of interest extracted based on the segmentation results, which simplifies the registration process. Using the registration results, the corresponding points are searched for between intra- and inter-faces. Finally, symmetry analysis is performed. Our method quantifies facial asymmetry from two perspectives: shape and movement differences. Both difference indicators are easily calculated using the point correspondences. The experimental results showed that our proposed method provided a reasonable analysis. We also discussed the relationship between segmentation accuracy and analysis quality. The results demonstrated the need to improve the segmentation to obtain better analysis results.

Keywords—cleft lip, facial symmetry analysis, 3D point cloud, 4D data, segmentation, registration, deep learning, coherent point drift

Cite: Narumi Kihara, Namiko Kimura-Nomoto, Takako Okawachi, Guangxu Li, Norifumi Nakamura, and Tohru Kamiya, "A Novel Approach for Facial Symmetry Analysis Based on Semantic Registration," Journal of Image and Graphics, Vol. 14, No. 5, pp. 756-762, 2026.

Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).