Segmentation Style Discovery: Application to Skin Lesion Images

This is the code repository for our paper published at Medical Image Computing and Computer-Assisted Intervention (MICCAI) ISIC Skin Image Analysis Workshop (ISIC) 2024:

Segmentation Style Discovery: Application to Skin Lesion Images
Kumar Abhishek¹, Jeremy Kawahara², Ghassan Hamarneh^1
1 Medical Image Analysis Lab, School of Computing Science, Simon Fraser University, Canada
² AIP Labs, Budapest, Hungary

Abstract

Variability in medical image segmentation, arising from annotator preferences, expertise, and their choice of tools, has been well documented. While the majority of multi-annotator segmentation approaches focus on modeling annotator-specific preferences, they require annotator-segmentation correspondence. In this work, we introduce the problem of segmentation style discovery, and propose StyleSeg, a segmentation method that learns plausible, diverse, and semantically consistent segmentation styles from a corpus of image-mask pairs without any knowledge of annotator correspondence. StyleSeg consistently outperforms competing methods on four publicly available skin lesion segmentation (SLS) datasets. We also curate ISIC-MultiAnnot, the largest multi-annotator SLS dataset with annotator correspondence, and our results show a strong alignment, using our newly proposed measure AS², between the predicted styles and annotator preferences.

Code

--coming soon--

ISIC-MultiAnnot Dataset

--coming soon--

Citation

If you find our work useful or if you use one or more of the StyleSeg method, the ISIC-MultiAnnot dataset, and the AS² measure in your work, please cite our paper:

Kumar Abhishek, Jeremy Kawahara, Ghassan Hamarneh, "Segmentation Style Discovery: Application to Skin Lesion Images", Medical Image Computing and Computer-Assisted Intervention (MICCAI) ISIC Skin Image Analysis Workshop (ISIC), 2024.

The corresponding bibtex entry is:

@InProceedings{abhishek2024segmentation,
author = {Abhishek, Kumar and Kawahara, Jeremy and Hamarneh, Ghassan},
title = {Segmentation Style Discovery: Application to Skin Lesion Images},
booktitle = {Proceedings of the International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI) ISIC Skin Image Analysis Workshop},
month = {October},
year = {2024}
}