Reflectance Confocal Microscopy Skin Strata Depth Stacks

308 depth stacks of human skin from 54 volunteers in two age groups (20–30 and 50–70 years) were used to develop an automated segmentation model. The dataset supports classification of en-face sections into four anatomical strata: stratum corneum, viable epidermis, dermal-epidermal junction, and papillary dermis. The developed logistic regression classifier achieved 85.6% test accuracy.

Use Cases

• Train a classifier to predict the anatomical strata (e.g., stratum corneum, papillary dermis) from features extracted from RCM depth stacks.
• Analyze the relationship between volunteer age group and the predicted interface depths for strata like the dermal-epidermal junction.
• Validate segmentation algorithms by comparing predicted interface depths (e.g., viable epidermis/dermal-epidermal junction) against the reported mean absolute errors of 3.1 μm to 6.0 μm.
• Use the 308 depth stacks as a benchmark for developing new automated feature extraction methods for RCM imagery.

Strengths

• Dataset contains 308 depth stacks, providing a substantial basis for model development.
• Includes data from 54 volunteers across two distinct age cohorts (20–30 and 50–70 years), enabling age-related analysis.
• Model performance is quantified with an 85.6% test accuracy and specific mean absolute error metrics for three interface depths.

Limitations

• Sample size of 54 volunteers may limit the generalizability of findings across broader demographics.
• Data is restricted to two specific age groups, lacking representation from other age ranges.
• The dataset's utility for direct machine learning is dependent on the availability of the extracted 'bag of features', not the raw image stacks.

Provenance

Source

Dryad digital repository.

Collection Method

Reflectance confocal microscopy (RCM) imaging of human skin volunteers.

Freshness

Last updated in June 2020.