AI-based 3D facial symmetry analysis before and after OGS

Automatic Assessment of 3-Dimensional Facial Soft Tissue Symmetry Before and After Orthognathic Surgery Using a Machine Learning Model: A Preliminary Experience

Authors: Lun-Jou Lo, Chao-Tung Yang, Cheng-Ting Ho, Chun-Hao Liao, Hsiu-Hsia Lin

---

Summary

This study from Chang Gung Memorial Hospital (Taiwan) presents the first application of a machine learning model for the automatic quantitative assessment of 3D facial soft tissue symmetry before and after orthognathic surgery (OGS).
A transfer learning (TL) convolutional neural network (CNN) was trained using 3D contour line features extracted from facial surface scans obtained with the 3dMD system.
The goal was to create an objective and reproducible evaluation tool for assessing postoperative outcomes, reducing subjective bias in facial symmetry analysis.

---

Results

• Preoperative symmetry score: Mean = 2.74 (range 1.59–3.88)
• Postoperative symmetry score: Mean = 3.52 (range 1.92–4.97)
• Improvement: 21% increase in symmetry (p = 0.000).
• The Xception model with contour features provided the most accurate classification of facial symmetry levels.
• Developed a Web-based system enabling clinicians to query patient IDs and visualize symmetry scores interactively.
• The tool improved communication between clinicians and patients, providing objective visual metrics for aesthetic outcome evaluation.

---

Keywords

3D Facial Symmetry, Orthognathic Surgery, Machine Learning, Deep Learning, Transfer Learning, Convolutional Neural Network, Facial Contour Mapping, 3D Photography, Digital Dentistry, Aesthetic Evaluation, Craniofacial AI.

---

Extracted Data

| Field | Information | |--------|--------------| | Year | 2021 | | Modality | 3D Facial Photography (3dMD) | | Dataset | 158 patients (72M / 86F; 19–28 years) | | Network Architecture | Xception CNN (Transfer Learning) | | Metrics | Mean improvement 21%; p = 0.000 | | AP – Strategy | Retrospective cohort with pre/post comparison | | AP – Professional Qty | 5 authors (Plastic Surgery, Orthodontics, Computer Science) | | AP – Supervisor Presence | Yes (Prof. Hsiu-Hsia Lin) | | AP – Experience Level | Experienced craniofacial surgeons and ML engineers | | AP – Expertise Area | Orthognathic Surgery, Craniofacial Imaging, AI Development | | AP – Tool or System | Web-based Facial Symmetry Assessment Platform | | Task | 3D Symmetry Classification (Pre- and Post-Operative Comparison) | | Project Objective | To apply a machine learning model to quantify 3D facial symmetry improvements after OGS. | | Clinical Relevance | Enables objective, reproducible aesthetic outcome evaluation, improving clinical communication and planning. | | DOI | https://doi.org/10.1097/SAP.0000000000002687 |

---

Clinical Relevance

• Clinical importance: Provides an objective, quantitative approach to evaluate 3D soft tissue symmetry before and after orthognathic surgery.
• Innovation: First use of machine learning with 3D contour line features for automatic facial symmetry evaluation.
• Practical impact: Enhances clinical decision-making, postoperative evaluation, and communication with patients through Web-based visualization.

---

:::YouTube Video https://youtu.be/EwEzUbNMT28 :::