Self-supervised Methods for Ugly Duckling Detection in Wide Field Images

Vullnet Useini, Nicolaus Andratschke, Stephanie Tanadini-Lang, Quentin Lohmeyer, Ralph P. Braun, Javier Barranco Garcia

Keywords: Self-supervised, melanoma, ugly duckling, outlier

TL;DR: Self-supervised algorithms for ugly duckling scoring

Abstract: Screening skin lesions is a very time-consuming process in which the dermatologist examines hundreds of lesions all over the patient's body in a limited period of time. The decision as to which lesions should be further examined is made based on the "ugly duckling" sign. The dermatologist compares all lesions on the same patient and identifies those that are different from the average-looking lesions. Deep learning algorithms have been shown to be efficient tools for detecting outliers in large image datasets. In this study, we propose a self-supervised approach for lesion clustering and outlier detection to identify and suggest lesions of interest for each individual patient.

Registration: I acknowledge that acceptance of this work at MIDL requires at least one of the authors to register and present the work during the conference.

Authorship: I confirm that I am the author of this work and that it has not been submitted to another publication before.

Paper Type: novel methodological ideas without extensive validation

Primary Subject Area: Application: Dermatology

Secondary Subject Area: Unsupervised Learning and Representation Learning

Confidentiality And Author Instructions: I read the call for papers and author instructions. I acknowledge that exceeding the page limit and/or altering the latex template can result in desk rejection.