Evaluation of facial skin age based on biophysical properties in vivo.

OBJECTIVE

The evaluation of skin age, reflecting overall facial characteristics, has not been established. Previous studies focused on visual assessment or individual-specific feature such as wrinkles or skin color. We studied the evaluation model of skin age index (SAI) including the overall aging features including wrinkles, skin color, pigmentation, elasticity, and hydration.

METHODS

Total 300 healthy women aged between 20 and 69 years included in this study. Pearson correlation analysis performed to identify the key factors among the biophysical properties with aging and developed the prediction model of SAI. Statistical regression analysis and machine learning technique applied to build the prediction model using the coefficient of determination (R ) and root mean square error (RMSE). Validation study of the SAI model performed on 24 women for 6 weeks application with anti-aging product.

RESULTS

Prediction model of SAI consisted of skin elasticity, wrinkles, skin color (brightness, Pigmented spot, and Uv spot), and hydration, which are major features for aging. The cforest model to assess a SAI using machine learning identified the highest R and lowest RMSE compared to other models, such as svmRadial, gaussprRadial, blackboost, rpart, and statistical regression formula. The cforest prediction model confirmed a significant decrease of predicted SAI after 6 weeks of application of anti-aging product.

CONCLUSION

We developed a prediction model to evaluate a SAI using machine learning, and led to accurate predicted age for overall clinical aging. This model can a good standard index for evaluating facial skin aging and anti-aging products.