Accurate segmentation of cell nuclei is the basic work of pathological diagnosis,and for the problems of current segmentation algorithms such as difficult extraction of fine features and much detail loss,a segmentation network based on generative adversarial network (GAN) with ResUNet is proposed in this paper.Firstly,the ResUNet network is used as the generative network,and the LeakyReLU activation function is used to enable the activation of negative-valued features,followed by the discriminative loss value of the discriminative network to guide the generative network to learn better.The experimental results show that the network in this paper achieves 82%,83%, 95% and 90%,90%,97% of the evaluation indexes of MioU, Dice and Acc on the breast cancer cell nucleus dataset and DSB dataset, respectively, which is 2.5%, 3.3%,0.7% and 0.7%,1.5%, 0.8% improvement over the ResUNet network, respectively.At the same time,the segmentation results of six commonly used segmentation network, such as SegNet and FCN8s,are improved, and the results proved that the improved network has better segmentation accuracy,which can provide an important basis for pathological diagnosis work.