Skin Cancer Clinical Features & Staging

This upload contains the full technical documentation and a free evaluation sample (799 lesions on 786 patients, 3,087 observation rows, an 8 % stratified draw from the full cohort) — not the full dataset. Full dataset (10,000 synthetic skin lesions, 8,221 patients, 37,497 observation rows, 4 Parquet tables): https://sentineldata.com.ua/dataset/skin-cancer-clinical-features-staging Sample contents (SAMPLE.7z, verified by unpacking and counting): File Rows × Cols Content skin_patients_sample.parquet 786 × 9 Demographics, Fitzpatrick phototype, UV exposure index, personal and family history, immunosuppression, nevus count skin_lesions_sample.parquet 799 × 17 Latent ground truth: diagnosis, melanoma subtype, Breslow thickness, ulceration, mitotic index, T-category, longitudinal flags skin_observations_sample.parquet 3,087 × 19 One row per lesion × rater × visit (3 observer arms; 1,029 rows per arm; 2,397 visit-1 and 690 visit-2 rows) skin_lesion_arms_sample.parquet 799 × 9 Cross-arm comparison per lesion SCHEMA_REPORT.txt — Column-by-column schema report of the sample (dtype, nullability, null counts, enums, constraints) Sample passport (PDF) — Provenance and stratification of the sample The sample is a stratified draw, not the first N rows: 8 % random draw within each diagnosis stratum (random_state = 42) — 706 nevus, 36 BCC, 35 melanoma, 22 SCC — with 230 lesions in the ... Full description with all tables is on the Wiki page of this project. Full dataset and purchase: https://sentineldata.com.ua/dataset/skin-cancer-clinical-features-staging . Zenodo copy: https://doi.org/10.5281/zenodo.23456486