Beyond Fitzpatrick: automated artificial intelligence-based skin tone analysis in dermatological patients.

Area of Science:

• Dermatology
• Computer Vision
• Medical Imaging

Background:

• Human skin tone is influenced by genetic, environmental, and cultural factors, impacting dermatological disease presentation.
• The Fitzpatrick scale, commonly used in dermatology, has limitations in capturing the full spectrum of skin tone diversity.
• Accurate skin tone classification is crucial for dermatology, clinical research, and personalized medicine.

Purpose of the Study:

• To develop and validate an automated algorithm for objective skin tone assessment.
• To calculate the Individual Typology Angle (ITA) from color values for skin tone classification.
• To map ITA values to established skin tone scales, including Fitzpatrick and Monk.

Main Methods:

• Utilized DensePose and OpenFace for extracting color values from images.
• Developed an algorithm to compute the Individual Typology Angle (ITA) from CIELAB color data.
• Validated the algorithm on 3D body scans and AI-generated images, comparing results against Monk and Fitzpatrick classifications.

Main Results:

• The algorithm demonstrated high agreement with Monk skin tone classifications.
• While showing promise, the algorithm's alignment with Fitzpatrick types was less consistent.
• The algorithm proved reliable in classifying skin tone to the Monk scale, even with class imbalance.

Conclusions:

• The developed algorithm offers a reliable method for automated skin tone assessment, particularly for the Monk scale.
• This technology has significant potential for applications in teledermatology, clinical research, and personalized medicine.
• Further external validation is recommended to confirm the algorithm's generalizability and robustness.