Rosacea Erythema Severity Assessment Scores from a Multimodal AI Framework

1,001 patient records form the basis for an objective Clinician’s Erythema Assessment grading tool for rosacea. The dataset was created by Jun Zhang and published on figshare in 2026. It integrates handcrafted image-derived tabular features with deep learning representations from standardized VISIA® imaging.

Use Cases

• Train machine learning models for objective rosacea severity grading based on multimodal image-derived features.
• Benchmark AI model performance using expert consensus labels and metrics like Quadratic Weighted Kappa.
• Analyze model interpretability for clinical safety based on layer-wise relevance propagation results.
• Develop tools for clinical trials and teledermatology requiring consistent, digital erythema measurement.

Strengths

• Dataset is based on a cohort of 1,001 patients, providing a substantial sample size.
• Labels were established via unanimous consensus from three expert dermatologists, creating a robust reference standard.
• The described multimodal AI framework achieved a Macro-AUC of 0.902 and a Quadratic Weighted Kappa of 0.800, indicating strong performance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect temporal or source bias inherent to the specific retrospective cohort studied.

Provenance

Source

Jun Zhang via figshare.

Collection Method

Retrospective analysis of patient imaging data with expert consensus labeling.

Freshness

Last updated 2026-05-11 05:26:00; freshness should be verified.