Research on brain tumor segmentation algorithm based on multiple encoders
In order to improve the accuracy of automatic brain tumor segmentation,a multi-encoder based brain tumor segmentation algorithm is proposed.In this paper,the multi-encoder U-shaped network structure is firstly adopted to deeply mine the semantic information of different modes,then the convolution module in the down sampling uses cavity convolution to extract multi-scale features,and the fusion module based on channel attention is used to fuse the features of different modes.Finally,the Transformer module is used at the bottleneck layer of the network to fully perceive the global features.The algorithm is tested on the Brain Tumor Segmentation Challenge 2018(BraTS 2018)dataset,and the average Dice coefficients of the whole tumor,the tumor core and the enhanced tumor are 0.8907,0.7851 and 0.7487,respectively.The experimental results show that the algorithm can effectively utilize the difference of modal characteristics and improve the segmentation effect.
NETL
NSTL
万方数据
