Abstract
To obtain more semantic information with small samples for medical image segmentation, this paper proposes a simple and efficient dual-rotation network (DR-Net) that strengthens the quality of both local and global feature maps. The key steps of the DR-Net algorithm are as follows (as shown in Fig. 1). First, the number of channels in each layer is divided into four equal portions. Then, different rotation strategies are used to obtain a rotation feature map in multiple directions for each subimage. Then, the multiscale volume product and dilated convolution are used to learn the local and global features of feature maps. Finally, the residual strategy and integration strategy are used to fuse the generated feature maps. Experimental results demonstrate that the DR-Net method can obtain higher segmentation accuracy on both the CHAOS and BraTS data sets compared to the state-of-the-art methods.
Highlights
Computer vision is widely used in medical tasks: medical image semantic segmentation [1,2,3], medical image classification [4,5,6], bioengineering recognition [7,8,9], threedimensional reconstruction [10,11,12] and others
The section describes the framework of the entire model, the following section describes the details of the partial feature maps (PFMs) and dilated partial feature maps (DPFMs), the section describes the different strategies used in the encoder and decoder, and the following section introduces the environmental details of the dual-rotation network (DR-Net) algorithm (Fig. 2)
To further obtain more semantic information, in this part, we introduced the multiscale convolution in PFMs and generated 16 sets of feature maps with different semantic information
Summary
Computer vision is widely used in medical tasks: medical image semantic segmentation [1,2,3], medical image classification [4,5,6], bioengineering recognition [7,8,9], threedimensional reconstruction [10,11,12] and others. More semantic information can be obtained through the feature maps generated after the fusion of multiple strategies, the total number of parameters will increase, which greatly enhances the computational complexity of the algorithm. The advantage of converting high-dimensional feature maps to low-dimensional feature maps is that the number of parameter calculations can be reduced Their algorithm obtained good results on the ImageNet-1 k data set, proving that group convolution has good feature learning capabilities. To obtain richer semantic information without increasing the complexity of the algorithm and reducing the preprocessing of the data, this paper proposes a faithful deep learning algorithm: the double rotation network (DR-Net). We give a summary of this paper and further research goals
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
