TL;DR本研究提出了一种基于全卷积神经网络的方法,将临床显微镜图像中的皮肤病变区域分割出来,利用该方法可以有效地探讨临床皮肤显微镜特征识别问题,并在 ISIC-ISBI Part 2 比赛中取得了显著的优异表现。
Abstract
The presence of certain clinical dermoscopic features within a skin lesion
may indicate melanoma, and automatically detecting these features may lead to
more quantitative and reproducible diagnoses. We reformulat