Abstract: Aimed at the problem of poor segmentation accuracy due to fuzzy edge of nuclei and overlapping nuclei, an edge and position information-based nuclei segmentation method is proposed. The proposed method used the nuclei edge enhancement network to obtain and enhance the boundary information of the mask, and used the enhanced edge information to guide the encoder of convolutional neural network to extract the edge information. In the encoder, the edge refinement module was used to obtain more accurate nuclei edge. Between the encoder and the decoder, the position information was obtained through the dual attention mechanism. In network training, the loss function combined with edge information was used. The experimental results show that this method can improve the accuracy of nuclear segmentation by using edge and position information.