Skin_Cancer_PAD-UFES-20: 2,298 Skin Lesion Images with 26 Clinical Features

2,298 images of six types of skin lesions, including three cancers, collected from 1,373 patients. The dataset includes 26 metadata features per lesion, such as patient demographics, habits, and lesion characteristics. It was published in July 2020 by researchers from the Federal University of Espirito Santo, Brazil.

Use Cases

• Classifying skin lesion types based on visual image data.
• Predicting diagnostic outcomes using combined image and clinical metadata features like age, skin type, and lesion diameter.
• Studying correlations between patient lifestyle factors (smoking, alcohol, pesticide exposure) and lesion types.
• Developing models for biopsy-proven versus clinically diagnosed lesion verification.

Strengths

• Includes 2,298 images across six clinically relevant lesion categories.
• Provides 26 metadata features per sample, including patient history, lesion measurements, and biopsy confirmation status.
• Approximately 58% of samples are biopsy-proven, adding diagnostic reliability.

Limitations

• Image sizes vary due to collection from different smartphone devices, which may require preprocessing for uniform model input.
• The description metadata is limited; actual data quality and completeness of all 26 columns require manual inspection after download.

Provenance

Source

Pacheco, A. G., Lima, G. R., Salomao, A. S., Krohling, B., Biral, I. P., de Angelo, G. G., ... & de Barros, L. F. (Federal University of Espirito Santo, Brazil).

Collection Method

Images collected using different smartphone devices; diagnoses are biopsy-proven or based on dermatologist consensus.

Time Range

Time of patient examination is not specified.

Freshness

Published July 2020; last update date is unknown.

Geography

Likely Brazil, based on the author's institution, but not explicitly stated.