Crowd annotations of visual features for 100 images from the ISIC 2017 lesion classification challenge

This item is shared privately

dataset

modified on 2018-10-31, 12:26

This dataset consists of additional images for 100 images from the ISIC 2017 challenge training data.

We collected the annotations during a first year undergraduate project course on medical image analysis (course code 8QA01) at the Department of Biomedical Engineering, Eindhoven University of Technology. In groups of five or six, the students learnt to automatically measure image features, such as "asymmetry", in images of skin lesions from the ISIC 2017 challenge.

Each group was encouraged to decide which features they wanted to measure, invent their own way of grading the images, and assess each feature visually by at least three people. The students were not blinded to the melanoma/non-melanoma labels in the data, since the data is openly available online.

The shared file contains the annotations of group 7. They annotated 5 types of visual features by 6 annotators each. For example, the column Asymmetry_7_1 is the assessment of Asymmetry by the 1st annotator, and so forth.

The file also contains several variables from the ISIC challenge: ID of image, melanoma and keratosis labels, and sex and age of the patient, if available.

Keywords

crowdsourcing skin lesion melanoma classification problems Medical Imaging Artificial Intelligence and Image Processing Computer Vision

Licence

CC BY 4.0

Crowd annotations of visual features for 100 images from the ISIC 2017 lesion classification challenge

Categories

Keywords

Licence