Skip to content

Metadata and materials for "Lesion size matters: A validation dataset for melanoma detection"

License

Notifications You must be signed in to change notification settings

ISIC-Research/lesion-size-2020

Repository files navigation

Metadata for "Consecutive biopsies for melanoma across year 2020"

This is the metadata for the dataset of 1,295 consecutive lesions biopsied to rule out melanoma in 2020 at a tertiary cancer center in the United States, annotated with patient demographics, lesion diameter, human labelers, Melanocytic Pathology Assessment Tool and Hierarchy for Diagnosis (MPATH-Dx) class for degree of dysplasia or atypia, and reference standard histopathological diagnosis. We validated this dataset against an open-source binary classifier for melanoma that won the International Skin Imaging Collaboration (ISIC) grand challenge for melanoma detection in 2020 held on Kaggle. This experiment and results were described in a paper recently submitted to the NeurIPS 2022 Datasets and Benchmarks Track.

About

Metadata and materials for "Lesion size matters: A validation dataset for melanoma detection"

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published