Gathering data from image segmentation for breast cancer, a critical work analysis in medical field. This can be used to diagnose and treat diseases related to breast cancer. Convolutional neural networks (CNNs) have shown exceptional performance in recent years in several medical picture segmentation tasks, such as breast image segmentation. This study evaluates image segmentation performance of the histopathological breast cancer images using three CNN models. First model is U-Net. And other two are its variants U-Net++, and U-Net3+. The latter two models developed are to overcome the limitations of the basic U-Net model. The dataset used for this assessment is publicly available. Although the experiments show that U-Net3+ outperformed in finding the surface distance and dice coefficient, other two models also achieved good results in terms of accuracy. Based on our results U-Net3+ is a best model for image segmentation of breast cancer that improves the accuracy in detection of disease.