ADVERSARIAL LEARNING WITH MULTI-SCALE LOSS FOR SKIN LESION SEGMENTATION

Inspired by classic Generative Adversarial Networks (GAN), we propose a novel end-to-end adversarial neural network, called SegAN, for the task of medical image segmentation. Since image segmentation requires dense, pixel-level labeling, the single scalar real/fake output of a classic GAN's discriminator may be ineffective in producing stable and sufficient gradient feedback to the networks. Instead, we use a fully convolutional neural network with new activation function in the last layer as the segmentor to generate segmentation label maps, and propose a novel adversarial critic network with a multi-scale L-1 loss function to force the critic and segmentor to learn both global and local features that capture long-and short-range spatial relationships between pixels. We show that such a SegAN framework is more effective in the segmentation task and more stable to train, and it outperforms current state-of-the-art segmentation methods in the ISBI International Skin Imaging Collaboration (ISIC) 2017 challenge, Part I Lesion Segmentation.