Region segmentation in histopathological breast cancer images using deep convolutional neural network

Computer aided diagnosis of breast cancers often relies on automatic image analysis of histopathology images. The automatic region segmentation in breast cancer is challenging due to: i) large regional variations, and ii) high computational costs of pixel-wise segmentation. Deep convolutional neural network (CNN) is proven to be an effective method for image recognition and classification. However, it is often computationally expensive. In this paper, we propose to apply a fast scanning deep convolutional neural network (fCNN) to pixel-wise region segmentation. The fCNN removes the redundant computations in the original CNN without sacrificing its performance. In our experiment it takes only 2.3 seconds to segment an image with size 1000 × 1000. The comparison experiments show that the proposed system outperforms both the LBP feature-based and texton-based pixel-wise methods.