基于多尺度结合全局和局部注意力的肺结节良恶性分类网络  

Benign and malignant pulmonary nodule classification network based on multi-scale combined with global and local attention

作　　者：李炳淑 吕楠 孙福振[1] LI Bingshu;LÜNan;SUN Fuzhen(College of Computer Science and Technology,Shandong University of Technology,Zibo 255000,China;Shandong Provincial Chronic Disease Hospital(Shandong Rehabilitation Center),Qingdao 266071,China)

机构地区：[1]山东理工大学计算机科学与技术学院,山东淄博255000 [2]山东省慢性病医院(山东省康复中心),山东青岛266071

出　　处：《山东科技大学学报（自然科学版）》2023年第6期97-105,共9页Journal of Shandong University of Science and Technology(Natural Science)

基　　金：国家自然科学基金项目(61841602);山东省自然科学基金项目(ZR2018PF005)。

摘　　要：由于肺结节具有不同尺寸大小,基于固定感受野的网络结构无法充分捕获肺结节的多尺度特征。针对该问题,提出一种基于多尺度结合全局和局部注意力的肺结节良恶性分类网络,首先通过设计的多尺度层动态学习肺结节不同大小尺度特征,增强网络对于多尺度特征的学习能力,然后使用注意力机制对通道特征以及空间特征进行增强,提升网络对于全局以及局部特征的信息捕获,最后在公开数据集LIDC-IDRI上进行性能验证,并与其他先进方法进行对比。实验结果表明,本研究提出的肺结节良恶性分类网络可以取得较好的分类结果。Due to the different sizes of pulmonary nodules,the existing fixed receptive field network structures cannot fully capture the multi-scale features of pulmonary nodules.To address this problem,this paper proposed a benign and malignant pulmonary nodule classification network based on multi-scale combined with global and local attention.Firstly,through the designed multi-scale network layer,the network dynamically learned the characteristics of different scales of pulmonary nodules to enhance the network’s learning ability for multi-scale representations.then,the attention mechanism was used to enhance the channel and spatial information so as to improve the ability of capturing global and local features.Finally,in order to verify the effectiveness of the network,the performance was validated on the public LIDC-IDRI dataset and compared with other state-of-the-art methods.The experimental results show that the designed network can achieve preferable results on benign and malignant classification tasks for pulmonary nodules.

关 键 词：多尺度 注意力机制 卷积神经网络 肺结节 分类 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]