ISIC-2018 Challenge Task 3 (ArtifactPlus)

The dataset consists of metadata for dermatological images (from the ISIC-2018 Challenge Task 3) with newly annotated artifacts such as the presence of hair, ruler marks, interface fluids, vignettes, and the 'other' category.
The dataset was constructed by carefully reviewing and labeling each image for these specific artifacts (one binary label per artifact), allowing for targeted analysis, training and combating data biases in multi-class skin lesion classification task.
Category, termed other, was designated to include various non-skin background elements (as in-ear or arm images), coloured markings, nostril holes, unusual vignettes, and other anomalous features that did not fit into the predefined four categories.
File List
├─── ISIC-2018 Challenge Task 3 (ArtifactPlus)           └─── HAM10000_metadata_with_artifacts_labels.csv           └─── ISIC2018_Task3_Test_GroundTruth_with_artifacts_labels.csv
- file HAM10000_metadata_with_artifacts_labels.csv contains 10 015 (original number of rows) * 5 (number of artifact categories labeled) new binary labels. 
- file ISIC2018_Task3_Test_GroundTruth_with_artifacts_labels contains 1 511 (original number of rows) * 5 (number of artifact categories labeled) new binary labels. 
In total, the ArtifactPlus dataset adds 57 630 new artifact label entries to the original datasets, enriching the data for more comprehensive model training and artifact analysis.

Citation
If you use this dataset in your research, please cite our accompanying paper:
J. Buler, R. Buler, K. Brzozowski, M. Ferlin, M. Bobowicz, and M. Grochowski, “A holistic approach to multi-modal skin lesion diagnosis supported by statistical and explainability-based investigation of artifacts,” Journal of Artificial Intelligence and Soft Computing Research, (in press), 2026.

Credits & Original Data Sources
HAM10000 Dataset: © by ViDIR Group, Department of Dermatology, Medical University of ViennaDOI:10.1038/sdata.2018.161
MSK Dataset: © Anonymousarxiv:1710.05006arxiv:1902.03368 Figures below present artifact counts per lesion class.