Automating reflectance confocal microscopy image analysis for dermatological research: a review

Area of Science:

• Dermatology
• Biomedical Imaging
• Artificial Intelligence

Background:

• Reflectance confocal microscopy (RCM) provides noninvasive, in vivo cellular-level imaging of skin.
• It is valuable for studying skin conditions where biopsies are impractical or traumatic.

Purpose of the Study:

• To review RCM applications in skin research.
• To overview machine learning methods for extracting information from RCM images.
• To assess RCM's potential for studying skin structure, barrier function, inflammation, and lesions.

Main Methods:

• Literature review using PubMed and reference lists.
• Analysis of RCM applications in dermatological research.
• Review of machine learning algorithms for RCM image analysis.

Main Results:

• RCM yields biologically relevant information on skin structure and changes.
• Machine learning algorithms have been developed for tasks like epidermal layer classification and lesion demarcation.
• Challenges include low image quality (noise, poor contrast) and limitations of supervised learning due to manual labeling.

Conclusions:

• RCM holds significant potential for studying skin structures and diseases.
• Artificial intelligence can enhance RCM image analysis for more precise and reproducible studies.
• Further research is needed to establish gold standards and address image quality, data limitations, and variability.