Biomedical Image Segmentation Based on Classification Supervision.

Convolutional neural networks (CNN) has been widely used in the biomedical image segmentation (BIS) for their remarkable feature representation capability. However, there are often segmentation errors and missing segmentation problems in biomedical image segmentation based on deep learning. In this paper, we propose a full convolutional neural network, which is assisted by classification supervision based on segmentation network. The algorithm first obtains a segmentation result through a basic segmentation network. Then a Classification Supervision Module (CSM) is designed to enable the network to judge whether each slicer contains lesions from the perspective of classification. In this way we allow the network to take advantage of more global information. Experimental results on several available databases demonstrate the effectiveness and advancement of the proposed method.