Comparing 2D and 3D representations for face-based genetic syndrome diagnosis

Area of Science:

• Medical genetics
• Computer vision
• Biomedical imaging

Background:

• Accurate clinical diagnosis of human genetic syndromes is often difficult.
• Facial phenotype analysis is crucial for diagnosing hundreds of genetic syndromes.
• Computer-assisted facial phenotyping offers a promising diagnostic aid.

Purpose of the Study:

• To compare the diagnostic performance of 2D versus 3D facial representations in computer-assisted syndrome diagnosis.
• To analyze subject-matched 2D and 3D facial data for comprehensive performance comparison.

Main Methods:

• Development of matched 2D and 3D facial representations from 1907 subject faces.
• Analysis included data from 43 distinct genetic syndromes.
• Training and evaluation of syndrome classification models using both 2D and 3D facial data.

Main Results:

• Three-dimensional (3D) surface-based syndrome classification models demonstrated significantly superior performance compared to 2D image-based models.
• Performance comparison was conducted on identically matched subject faces.
• The study represents the most comprehensive subject-matched analysis to date in this field.

Conclusions:

• Three-dimensional facial representations provide more accurate classification for genetic syndromes than 2D images.
• Clinical adoption of 3D facial scanning technology is recommended.
• Continued collection of 3D facial scan data for syndromic conditions may substantially improve diagnosis.