Classification of traditional Chinese medicine constitution based on facial features in color images

Abstract

Objective: To explore the possible correlation between traditional Chinese medicine (TCM) constitution and facial features in color images and to improve the accuracy of automated constitution classification. Methods: Color images were taken of 5150 individuals of different professions. Automated face detection and key point positioning were performed on the collected images, which were then transformed into a standard size. The relationship between facial features and TCM constitution based on the red, green, blue (RGB) pixel and the local binary pattern (LBP) texture features was explored. Results: The overall accuracy rate and robustness of TCM constitution classification based on RGB features were low. Classification results of the phlegm-dampness, damp-heat, blood stasis, and balance constitutions achieved high accuracy rates. Classification accuracy rate using the LBP texture feature was higher than that of the RGB feature, with the best accuracy observed for the balance constitution. Conclusion: Application of computer image acquisition and processing of facial features may serve as an adjunct to the TCM diagnostic method of inspection.

Keywords

• TCM constitution
• Automatic TCM diagnosis
• Local binary pattern