An automatic approach for classification and categorisation of lip morphological traits

Area of Science:

• Medical research
• Computer vision
• Biometrics

Background:

• Facial trait classification, particularly lip shape, is crucial for medical research, aiding in identifying genetic associations for conditions like cleft lip.
• Current methods rely on subjective, time-consuming manual assessments or 3D image analysis prone to operator error.
• There is a need for objective, automated methods for accurate facial trait classification.

Purpose of the Study:

• To develop and validate a novel, automated approach for the classification and categorization of lip area traits.
• To utilize advanced 3D geometric features for objective facial trait analysis.
• To compare different combinations of geodesic features for optimal classification performance.

Main Methods:

• The study proposes an automated method employing novel 3D geometric features derived from surface curvatures along geodesic paths between anthropometric landmarks.
• Various combinations of these geodesic features were analyzed and compared to determine the most effective classification parameters.
• A partial least squares method was used to visualize the impact of automatically identified categories on facial structures.

Main Results:

• The developed method was applied to classify six lip shape traits (philtrum, Cupid's bow, lip contours, lip-chin, and lower lip tone) in a large dataset of 4747 individuals.
• The automatic classification system demonstrated a high success rate, achieving up to 90% accuracy in correctly classifying lip traits.
• The visualization technique effectively illustrated the relationship between classified lip traits and overall facial morphology.

Conclusions:

• The proposed automated method offers a significant advancement over traditional subjective assessments for lip trait classification.
• This novel approach provides an accurate and efficient tool for medical research, particularly in understanding genetic influences on facial morphology.
• The high classification accuracy suggests strong potential for clinical applications in diagnosing and studying conditions related to facial development.