基于特征复用网络的医学图像分割

Medical Image Segmentation Based on Feature Multiplexing Network

随着计算机技术的发展,深度学习技术在众多领域发挥着重要作用。为提高医学图像分割的精度和实时性,采用可分离卷积设计了一种特征图复用结构的编码器-解码器式的图像语义分割网络RUNet。通过多特征融合的方法节约通道数,进而极大的减少参数量,同时增加网络深度提升模型表示特征能力,在EM数据集和LUNA数据集下测试RUnet平均Dice系数分别为0.9692和0.9877,分割效果优于U-net,但是计算量仅为U-net的1/75。

With the development of computer technology,deep learning technology plays an important role in many fields.In or⁃der to improve the accuracy and real-time performance of medical image segmentation,an encoder-decoder image semantic segmenta⁃tion network RUNet with a feature map multiplexing structure is designed using separable convolution.The method of multi-feature fu⁃sion saves the number of channels,thereby greatly reducing the amount of parameters,and increasing the depth of the network to im⁃prove the feature representation ability of the model.The average Dice coefficient of RUnet tested under the EM data set and LUNA data set is 0.9692 and 0.9877,respectively,and the segmentation effect Better than U-net,but the calculation amount is only 1/75 of U-net.