摘要
计算机辅助算法在医学图像疾病诊断中发挥着重要作用,一套准确的诊断系统是极其重要的。虽然现有肺结节分类模型性能已经有了很大提升,但在提取特征、提高准确率和降低假阳率方面存在不足。为了解决深度学习网络结构与肺结节图像的匹配问题,将3D多尺度作为输入,以DPN作为主干网络,能够更有效地提取图像特征信息,文中构造一种多尺度决策层融合网络模型,来区分肺结节的恶性和良性,该模型能够从原始图像中自动提取全面的图像特征。在肺图像数据库联盟图像采集(LIDC-IDRI)数据库上进行了一系列实验,结果表明,所提出的多尺度决策融合模型准确率要高于其他分类模型,具有良好的稳定性和鲁棒性。
Computer aided algorithm plays an important role in medical image disease diagnosis.An accurate diagnosis system is extremely important.Although the performance of existing pulmonary nodule models has been greatly improved,However,there are deficiencies in extracting features,improving accuracy and reducing false positive rate.In this paper,3D multi⁃scale is used as input,and DPN is used as the backbone network to extract image feature information more effectively.In order to solve the matching problem between deep learning network structure and lung nodule image,a multi⁃scale decision level fusion network model is constructed to distinguish malignant and benign lung nodules.The model can automatically extract comprehensive image features from the original image.A series of experiments are carried out on the lung image database alliance image acquisition(LIDC-IDRI)database.The results show that the accuracy of the multi⁃scale decision fusion model proposed in this paper is higher than other classification models,and has good stability and robustness.
作者
池洪泽
杨静
CHI Hongze;YANG Jing(Institute of Information and Computer,Taiyuan University of Technology,Jinzhong 030600,China;Information Center,Taiyuan University of Technology,Taiyuan 030024,China)
出处
《电子设计工程》
2022年第22期188-193,共6页
Electronic Design Engineering
基金
赛尔网络下一代互联网技术创新项目(NGI 120180311)。
