Contrast based lesion segmentation on DermIS and DermQuest datasets.

In this companion paper, we present the results obtained by our contrast based approach for lesion segmentation on two datasets used by amelard et al. (Amelard, J. Glaister, A. Wong and D. A. Clausi, High-Level Intuitive Features (HLIFs) for Intuitive Skin Lesion Description, IEEE Transactions on Biomedical Engineering, 62(3) (2015), 820-831). The first is a subset of a DermIS database (Dermatology Information System, available online: http://www.dermis.net) that contains 43 macroscopic photographs with lesion diagnosed as melanoma and 26 diagnosed as non melanoma. The second is a subset of the DermQuest database (Available online: http://www.dermquest.com) that contains 76 images of melanoma lesions and 61 images of non melanoma lesions. The images are subject to various artifacts like drastic shadow effect and the variation of illumination. We also provide the ground truth segmentation to evaluate visually the accuracy of our results.

在本配套论文中，我们展示了基于对比度的病变分割方法在Amelard等人所使用的两个数据集上的实验结果（Amelard, J. Glaister, A. Wong及D. A. Clausi. 用于直观皮肤病变描述的高阶直观特征（High-Level Intuitive Features, HLIFs）[J]. IEEE生物医学工程汇刊, 2015, 62(3): 820-831）。第一个数据集为DermIS数据库（皮肤病学信息系统，可在线访问：http://www.dermis.net）的子集，其中包含43张经诊断为黑素瘤的病变宏观摄影图像，以及26张经诊断为非黑素瘤的病变图像。第二个数据集为DermQuest数据库（可在线访问：http://www.dermquest.com）的子集，其中包含76张黑素瘤病变图像与61张非黑素瘤病变图像。该数据集的图像存在多种伪影，例如强烈阴影效应与光照变化。我们还提供了金标准分割掩码（ground truth segmentation），用于可视化评估本方法所得结果的准确性。