摘要
目的U-Net是医学图像分割领域中应用最为广泛的基础分割网络,然而U-Net及其各种增强网络在跳跃连接时仅利用相同尺度特征,忽略了具有互补信息的多尺度特征对当前尺度特征的指导作用。同时,跳跃连接时编码器特征和解码器特征所处的网络深度不同,二者直接串联会产生语义特征差距。针对这两个问题,提出了一种新型分割网络,以改进现有网络存在的不足。方法首先,将编码器不同层级具有不同尺度感受野的特征进行融合,并在融合特征与编码器各层级特征间引入加性注意力对编码器特征进行指导,以增强编码器特征的判别性;其次,在编码器特征和解码器特征间采用加性注意力来自适应地学习跳跃连接特征中的重要特征信息,以降低二者间的语义特征差距。结果在多模态脑肿瘤数据集BraTS2020(multimodal brain tumor segmentation challenge 2020)上评估了所提出的网络模型,并进行了消融实验和对比实验。实验结果表明,所提出的网络在BraTS2020验证数据集上关于整个肿瘤、肿瘤核心和增强肿瘤的平均Dice分别为0.8875、0.7194和0.7064,优于2D网络DR-Unet104(deep residual Unet with 104 convolutional layers)的分割结果,其中肿瘤核心和增强肿瘤的分割结果分别高出后者4.73%和3.08%。结论所提出的分割网络模型,通过将编码器中具有互补信息的多尺度特征进行融合,然后对当前尺度特征进行加性注意力指导,同时在编码器和解码器特征间采用加性注意力机制来降低跳跃连接时二者间的语义特征差距,能更精准地分割MR(magnetic resonance)图像中脑肿瘤子区域。
Objective U-Net can be as the basic network in medical image segmentation.For U-Net and its various aug⁃mented networks,the encoder can extract features from input images in terms of a series of convolution and down-sampling operations.With the convolution and down-sampling operations at each layer of the encoder,the feature map sizes are decreased and the receptive field sizes can be remained to increase.For the network training,each level of the encoder can learn discriminative feature information at the current scale.To improve its feature utilization,the augmented U-Net schemes can melt skip connections between the encoder features and the decoder features into feature information-reused of shallow layers.However,the same scale are concatenated the features only via the skip-connected channel,and the role of multi-scale features with complementary information can be ignored.In addition,encoder features are oriented at a rela⁃tively shallow position in the overall network structure,while decoder features are based on a relatively deep position.As a result,a semantic feature gap is required to be bridged between encoder features and decoder features when skip connec⁃tions are made.To optimize the U-Net and its augmented networks model,a novel segmentation network model is devel⁃oped.Method We construct a segmentation network in terms of multi-scale feature fusion and additive attention mecha⁃nism.First,the features are fused in relevant to multi-scale receptive fields at different levels of the encoder.To guide the encoder features and enhance their discrimination ability,additive attention is introduced between the fused features and the encoder features at each level of the encoder.Second,to bridge the gap between the two semantic features,encoder and decoder features-between additive attention is used to learn important feature information in skip connections features adaptively.Experiments are carried out based on five-fold cross-validation.Multimodal magnetic resonance(MR)images of 234 high-grade
作者
孙家阔
张荣
郭立君
汪建华
Sun Jiakuo;Zhang Rong;Guo Lijun;Wang Jianhua(Faculty of Electrical Engineering and Computer Science,Ningbo University,Ningbo 315211,China;Affiliated Hospital of Medicine School of Ningbo University,Ningbo University,Ningbo 315211,China)
出处
《中国图象图形学报》
CSCD
北大核心
2023年第4期1157-1172,共16页
Journal of Image and Graphics
基金
浙江省公益技术研究项目(LGF21F020008)
浙江大学CAD&CG国家重点实验室开放项目(A2119)。
