Recurrent Residual Convolutional Neural Network based on U-Net (R2U-Net)\n for Medical Image Segmentation

Abstract

Deep learning (DL) based semantic segmentation methods have been providing\nstate-of-the-art performance in the last few years. More specifically, these\ntechniques have been successfully applied to medical image classification,\nsegmentation, and detection tasks. One deep learning technique, U-Net, has\nbecome one of the most popular for these applications. In this paper, we\npropose a Recurrent Convolutional Neural Network (RCNN) based on U-Net as well\nas a Recurrent Residual Convolutional Neural Network (RRCNN) based on U-Net\nmodels, which are named RU-Net and R2U-Net respectively. The proposed models\nutilize the power of U-Net, Residual Network, as well as RCNN. There are\nseveral advantages of these proposed architectures for segmentation tasks.\nFirst, a residual unit helps when training deep architecture. Second, feature\naccumulation with recurrent residual convolutional layers ensures better\nfeature representation for segmentation tasks. Third, it allows us to design\nbetter U-Net architecture with same number of network parameters with better\nperformance for medical image segmentation. The proposed models are tested on\nthree benchmark datasets such as blood vessel segmentation in retina images,\nskin cancer segmentation, and lung lesion segmentation. The experimental\nresults show superior performance on segmentation tasks compared to equivalent\nmodels including U-Net and residual U-Net (ResU-Net).\n

Ähnliche Arbeiten