LU-NET: An improved U-net for ventricular segmentation

Abstract

In order to solve the problem of low accuracy of the U-Net in cardiac ventricular segmentation, we propose an improved U-net named LU-Net by the following three methods. First, in order to improve the efficiency and effectiveness of extracting the features of the original image, we combine U-net with SE-Net model. This model reweights the channels of the feature map, which can give higher weight to the useful information and lower weight to the invalid information. Second, in order to alleviate the extent of losing the pixel-location information when using the encoder to dawn sample, we combine multi-scale input with U-net's encoder. Third, in order to solve the problem of low accuracy in traditional U-net, we replace the transposed convolution layer, used by the traditional U-Net's encoder during upsampling, with an unsampling layer. During the process of unsampling, it can put pixels to their original location using the pixel-location information reserved by the encoder during the sampling process, which can reduce errors caused by losing pixel-location information. Besides, using the unsampling layer during unsampling can also avoid producing checkerboard artifacts during transposed convolution and improve the segmentation accuracy. To verify the effectiveness of LU-Net, we apply it to the ACDC Stacom 2017 dataset. The experimental results show that the evaluation criteria of prediction results are 92.4%, 86.4%, and 92.5% on Dice coefficient, Jaccard similarity coefficient, and F1-ccore respectively, which are better than U-Net, SegNet, and IU-Net and remarkably better than the traditional neural convolution network model, FCN8s.