20,971 RGB leg images are labeled by a certified phlebologist according to seven CEAP classes (C0–C6) of lower extremity venous insufficiency. The dataset was created by Marina Barulina and was last updated in April 2026.
Use Cases
- Train a convolutional neural network to classify CEAP class (C0-C6) from RGB leg images.
- Benchmark model performance on a dataset with a highly unbalanced class distribution across the seven CEAP labels.
- Develop a feature extractor for venous disease severity using labeled leg images as ground truth.
- Study the challenges of training classifiers on medical image data with significant label imbalance.
Strengths
- Contains 20,971 labeled RGB images.
- Labels are provided by a certified phlebologist, ensuring clinical validity.
- Images are categorized into seven specific CEAP clinical classes (C0–C6).
Limitations
- The class distribution is highly unbalanced, which can bias model training.
- The number of rows (images) is known, but the size, file formats, and specific image dimensions are unknown.
- No sample data or column information is available to inspect data structure.
Provenance
- Source
- Marina Barulina Dataverse
- Collection Method
- Images were labeled by a certified phlebologist according to CEAP classification.
- Time Range
- null
- Freshness
- null
- Geography
- null