Applied Cosmetic Science and Technology Applied Cosmetic Science and Technology

Effectively managing stress is essential for enhancing one’s quality of life but requires a good understanding of one’s stress levels. Nonetheless, the diversity and dynamic nature of stress symptoms can complicate this endeavor. In this study, we explored the potential of leveraging facial information to develop technologies for stress assessment. A prospective multicenter study was conducted in Tokyo, Japan, from January 2018 to December 2021, enrolling 2343 participants aged between 20 and 60 years. Participants completed a facial skin-related questionnaire, and their facial images and videos were collected for analysis. Stress levels were measured using both objective outcomes, including autonomic, blood, and urine markers, as well as subjective outcomes, such as fatigue scales and quality of life questionnaires. Various machine learning techniques were employed to create separate evaluation models to predict the 5 categories of stress outcomes from the 3 sources of facial data. The criteria for model accuracy were set at >0.7. The models using facial image data emerged as the most accurate models for predicting various static stress states derived from questionnaires or from blood/urine biomarkers. Facial skin data from subjective questionnaires also accurately predicted static stress states. Facial video data accurately predicted dynamic stress states reflected by autonomic nervous system-based biomarkers such as the heart rate, coefficient of variation of R–R, and the ratio of the low- and high-frequency bands in heart rate variability. In this study, we developed several machine learning-based prediction models to assess static and dynamic stress levels using facial information, including images, videos, and questionnaires. The ease of capturing and analyzing facial data with readily available camera-equipped devices, such as smart devices and personal computers, makes this facial skin-based stress analysis promising for organizational health management and individual well-being. It enables early stress detection through self-assessment, exemplifying the application of cosmetics research knowledge to overall well-being.