IMA++: ISIC Archive Multi-Annotator Dermoscopic Skin Lesion Segmentation Dataset
Kumar Abhishek, Jeremy Kawahara, Ghassan Hamarneh
2026-01-06
Summary
This paper introduces a new, large dataset called ISIC MultiAnnot++ designed to help researchers improve the accuracy of computer programs that identify and outline skin lesions in dermoscopic images.
What's the problem?
Developing reliable computer systems to analyze skin lesions is difficult because it requires a lot of accurately labeled images for training. Creating these labeled datasets is expensive and time-consuming, as it needs experts to carefully outline the lesions in each image. Currently, there weren't any large, publicly available datasets of skin lesion images with multiple experts independently labeling the same lesions, which makes it hard to build and compare different AI approaches.
What's the solution?
The researchers created ISIC MultiAnnot++ by collecting a large number of dermoscopic images of skin lesions and having multiple experts independently outline the lesions in many of those images. The final dataset includes almost 15,000 images with over 17,000 outlines, and importantly, it includes information about *who* created each outline and *what tools* they used. They also analyzed the dataset to understand its characteristics and created standard sets of images for testing and comparison.
Why it matters?
This dataset is important because it provides the research community with a valuable resource to develop and test more accurate and reliable AI systems for skin lesion analysis. The inclusion of annotator information opens up possibilities for studying how different experts approach the task and for building AI systems that can account for individual preferences, ultimately leading to better diagnostic tools.
Abstract
Multi-annotator medical image segmentation is an important research problem, but requires annotated datasets that are expensive to collect. Dermoscopic skin lesion imaging allows human experts and AI systems to observe morphological structures otherwise not discernable from regular clinical photographs. However, currently there are no large-scale publicly available multi-annotator skin lesion segmentation (SLS) datasets with annotator-labels for dermoscopic skin lesion imaging. We introduce ISIC MultiAnnot++, a large public multi-annotator skin lesion segmentation dataset for images from the ISIC Archive. The final dataset contains 17,684 segmentation masks spanning 14,967 dermoscopic images, where 2,394 dermoscopic images have 2-5 segmentations per image, making it the largest publicly available SLS dataset. Further, metadata about the segmentation, including the annotators' skill level and segmentation tool, is included, enabling research on topics such as annotator-specific preference modeling for segmentation and annotator metadata analysis. We provide an analysis on the characteristics of this dataset, curated data partitions, and consensus segmentation masks.