[Preliminary study on the method of automatically determining facial landmarks based on three-dimensional face...

Area of Science:

• Medical Imaging
• Computer-Aided Surgery
• Biometrics

Background:

• Accurate identification of three-dimensional (3D) facial anatomical landmarks is crucial for various applications, including prosthodontics and oral surgery.
• Traditional manual landmark determination can be time-consuming and subjective.

Purpose of the Study:

• To develop and evaluate an efficient, automatic method for determining 32 facial anatomical landmarks in 3D facial data.
• To compare the accuracy of the automatic method against manual expert determination.

Main Methods:

• Collected 3D facial data from 30 male patients with dental defects.
• Utilized Procrustes analysis for data normalization and alignment, creating a 3D average face model.
• Developed a 3D face template and used MeshLab and MeshMonk algorithms for automatic landmark localization.
• Compared automatic landmark coordinates with manually identified coordinates by an expert.

Main Results:

• The automatic method demonstrated a landmark localization error of (1.65±1.19) mm in patients without mandibular deviation.
• Errors were (1.19±0.45) mm for midline landmarks and (1.85±1.33) mm for bilateral landmarks in non-deviated cases.
• In patients with mild mandibular deviation, the overall error increased to (2.55±2.22) mm.

Conclusions:

• The proposed automatic method for facial landmark determination is feasible and efficient.
• Midline facial landmarks were determined more accurately than bilateral landmarks.
• The method's effectiveness is higher in patients without mandibular deviation compared to those with mild deviation.