Evaluation of facial attractiveness for purposes of plastic surgery using machine-learning methods and image analysis

Many current studies conclude that facial attractiveness perception is data-based and irrespective of the perceiver. However, analyses of facial geometric image data and its visual impact always exceeded power of classical statistical methods. In this study, we have applied machine-learning methods to identify geometric features of a face associated with an increase of facial attractiveness after undergoing rhinoplasty. Furthermore, we explored how accurate classification of faces into sets of facial emotions and their facial manifestations is, since categorization of human faces into emotions manifestation should take into consideration the fact that total face impression is also dependent on expressed facial emotion. Both profile and portrait facial image data were collected for each patient (n = 42), processed, landmarked and analysed using R language. Multivariate linear regression was performed to select predictors increasing facial attractiveness after undergoing rhinoplasty. The sets of used facial emotions originate from Ekman-Friesen FACS scale, but was improved substantially. Bayesian naive classifiers, decision trees (CART) and neural networks were learned to allow assigning a new face image data into one of facial emotions. Enlargements of both a nasolabial and nasofrontal angle within rhinoplasty were determined as significant predictors increasing facial attractiveness (p <; 0.05). Neural networks manifested the highest predictive accuracy of a new face classification into facial emotions. Geometrical shape of a mouth, then eyebrows and finally eyes affect in descending order final classified emotion, as was identified using decision trees. We performed machine-learning analyses to point out which facial geometric features, based on large data evidence, affect facial attractiveness the most, and therefore should preferentially be treated within plastic surgeries.