Facial aging analysis across four major Chinese cities using deep learning techniques

Summary: Facial aging patterns exhibit significant inter-individual and regional variability, influenced by both intrinsic and environmental factors. Traditional methods for assessing facial aging often rely on subjective human evaluations, which are resource-intensive and prone to variability. In this study, we employed deep learning techniques to quantitatively analyze facial aging in 12,000 women aged 18-60 from four major Chinese cities—Beijing, Shanghai, Guangzhou, and Xi'an. Using an optimized U-Net architecture with multi-scale feature fusion, we quantified key aging features. Additionally, environmental factors such as climate and pollution were analyzed using Partial Least Squares Path Modeling (PLS-PM) to assess their impact on aging patterns. Results revealed distinct regional variations in facial aging patterns: Guangzhou exhibited milder wrinkle-related issues but more pronounced pigmentation concerns, while Xi'an displayed prominent dark circles and frown lines. Beijing was characterized by noticeable wrinkle-related aging, whereas Shanghai showed minimal overall facial aging issues but higher pore-related concerns. Environmental analysis identified air pollutants as a significant factor in Beijing (VIP=1.16), while ultraviolet radiation had a notable impact in Xi'an (VIP=1.12). Individual factors, such as marital and parental status, also correlated with aging severity, particularly in Beijing and Xi'an (VIP=0.84-1.48). Aging was categorized into three dimensions—structural change("Structure"), hyperpigmentation ("Color"), and sensitivity ("Sensitivity")—with pigmentation exerting the strongest overall influence on perceived aging (I=0.42). The PLS-PM model demonstrated a strong fit (R²=0.81), highlighting robust relationships among these dimensions. This study not only advances the understanding of facial aging in Chinese women but also provides a foundation for targeted interventions, combining deep learning-based quantification with environmental and individual factor analysis. Fan Hu¹, Qianqian Wang¹, Ye Zhong¹ 1. Huashan Hospital Fudan University, Shanghai, Shanghai, China. Clinical Research: Interventional Research