Machine Learning for Facial Recognition in Orthodontics

The assessment of a patient’s cranio-facial morphology using facial photographs and lateral cephalograms is indispensable for making an appropriate diagnosis and planning treatment in modern orthodontics. The observation of the cranio-facial morphology enables the detection of not only orthodontic or orthopedic problems (e.g., the size and positions of the maxilla and mandibles) but also genetic problems. Recent improvements in technologies with the development of artificial intelligence (AI) systems will facilitate access to human experts’ knowledge and experiences. For example, an AI system that can detect rare syndromes in the same way a dysmorphologist does may help orthodontists develop appropriate treatment plans for patients. In this chapter, we will introduce two AI systems for analyzing cranio-facial morphologies: a system that automatically provides clinical descriptions of facial images for orthodontic diagnostic purposes and a system that automatically identifies cephalometric landmarks using landmark-dependent multi-scale patches. Such AI systems enable orthodontists to understand patients’ cranio-facial morphological problems correctly and quickly and can even help inexperienced orthodontists detect genetic problems. These systems will help orthodontists maximize patient benefits while minimizing risks. In the final section, we will also discuss the challenges associated with deep learning in medical images.