Hyperspectral Imaging Combined With Machine Learning Methods to Quantify the Facial Skin Melanin and Erythema.

Area of Science:

• Dermatology and biomedical optics.

Background:

• Melanin and erythema are key physiological skin responses to environmental factors, crucial for diagnosing skin conditions.
• Objective skin color assessment is vital for dermatological diagnosis and aesthetic evaluation, but traditional methods have limitations.

Purpose of the Study:

• To investigate the roles of melanin and hemoglobin in skin-light interactions.
• To develop an objective skin color assessment method using spectral reflectance and pigment values.
• To visualize skin pigment distribution using hyperspectral imaging.

Main Methods:

• Combined spectral reflectance data with single-point pigment values from Mexameter MX18.
• Employed the competitive adaptive reweighted sampling algorithm to select feature wavelengths.
• Evaluated seven machine learning methods, including stacked generalization, for model development.
• Utilized hyperspectral imaging technology for pigment distribution visualization.

Main Results:

• Selected feature wavelengths aligned with Mexameter MX18 bands, optimizing data and model accuracy.
• The stacked generalization model achieved high accuracy for melanin index (R²v = 0.8634) and erythema index (R²v = 0.7505).
• Hyperspectral imaging successfully visualized skin pigment distribution.

Conclusions:

• Hyperspectral imaging combined with machine learning offers a rapid, non-invasive tool for objective skin color assessment.
• This approach provides valuable data for dermatological diagnosis and aesthetic evaluation by visualizing pigment distribution.
• The study demonstrates the efficacy of advanced analytical techniques in understanding skin physiology and improving clinical assessments.