Machine learning methods for determining skin age: A systematic review.

Area of Science:

• Dermatology
• Computer Science
• Biotechnology

Background:

• Skin aging is a complex process influenced by intrinsic and extrinsic factors.
• Accurate chronological age estimation from skin appearance is a growing area of research.

Purpose of the Study:

• To systematically review machine learning applications in determining skin aging.
• To evaluate the accuracy, limitations, and research gaps in current literature.

Main Methods:

• Systematic literature search of OVID Embase, OVID Medline, IEEE Xplore, and ACM Digital Library.
• Screening of 1467 articles, with 27 included in the final review.
• Analysis of machine learning model performance and methodologies.

Main Results:

• Machine learning models achieved accuracies with a mean absolute error ranging from 2.30 to 8.16 years.
• Full facial image analysis was the most common approach; biomarker analysis (methylome, proteome) also utilized.
• Dynamic facial expressions improved age estimation accuracy (MAE 3.74); confocal microscopy showed up to 85% accuracy.
• High risk of bias was noted in many studies, often due to small sample sizes.

Conclusions:

• Machine learning shows promise for accurate skin age estimation.
• Future research requires more diverse datasets (ethnicity, variables) to enhance generalizability.
• Addressing limitations like small sample sizes is crucial for robust model development.