Cell Nucleus Segmentation in Color Histopathological Imagery Using Convolutional Networks

Recent studies have shown that convolutional networks can achieve a great deal of success in high-level vision problems such as objection recognition. In this paper, convolutional networks are used to solve a typical low-level image processing task, image segmentation. Here, the convolutional networks are trained using gradient descent techniques to solve the problem of segmenting the cell nuclei from the background in the histopathology images. Using a dataset with 58 H&E stained breast cancer biopsy images, we find that the convolutional networks, with 3 hidden layers and 8 feature maps per hidden layer, provide superior performance to other pixel classification methods including FLDA and SVM. We also show two important properties of the convolutional networks as a segmentation method. First, as a machine learning approach, the convolution networks encode enough high-level domain-specific knowledge into the final segmentation strategy by learning the training data. Second, the convolutional networks can use appropriate amount of context information in segmenting by optimizing the weights of the filters in the networks through the learning process. In the end of this paper, several possible directions for future research are also proposed.