针对肝脏肿瘤分割面临的病灶形状、大小和位置差异明显等问题，文章提出了一种基于空间通道注意力的三维肝脏肿瘤分割方法。在 3D U-Net的基础上融合了Transformer，提出成对全局和局部注意力PGLA(Paired Global Local Attention)模块替代Transformer中的传统注意力模块，并在尺度变换前引入CBAM(Convolutional Block Attention Module)模块。在肝脏肿瘤分割挑战赛数据集上的实验结果显示该方法在肿瘤分割的Dice系数上达到了69。18%，这些成绩均优于当前流行的模型，这证明了该方法在提高肝脏肿瘤分割精度方面的有效性。
Liver Tumor Segmentation Based on Spatial Channel Attention
Aiming at the problems of obvious differences in lesion shape,size and location in liver tumor segmentation,this paper proposes a 3D liver tumor segmentation method based on spatial channel attention.It integrates Transformer based on 3D U-Net,proposes the PGLA module to replace the traditional attention module in the Transformer,and introduces the CBAM before scale transformation.The experimental results on the liver tumor segmentation challenge dataset show that the Dice coefficient of tumor segmentation of the proposed method reaches 69.18%.These results are better than the current popular models,which proves the effectiveness of the proposed method in improving the accuracy of liver tumor segmentation.
3D liver tumor segmentation3D U-NetTransformerPaired Global Local Attention moduleConvolutional Block Attention Module
何琼、陆雪松
展开 >
中南民族大学 生物医学工程学院,湖北 武汉 430074
3D肝脏肿瘤分割 3D U-Net Transformer 成对全局和局部注意力模块 卷积注意力模块
国家科技期刊平台
NSTL
万方数据
