摘要
肺部CT能够较准确地鉴定新冠肺炎病例,但医生工作量较大,本研究提出一种基于改进阈值的VGG网络的新冠肺炎CT图像自动诊断算法,通过该模型可快速准确地完成新冠肺炎病例的自动识别,为进一步控制其传播提供帮助。通过比较卷积神经网络VGG中的VGG-11、VGG-13、VGG-16,获得准确率较高的新冠肺炎CT图像自动诊断模型VGG-13,并在此基础上通过改进阈值的方式使准确率由86%提高到了89%,进一步提高诊断的准确性。
Lung CT can accurately identify COVID-19, but the workload of doctors is relatively large. An automatic diagnosis algorithm for COVID-19 CT image using improved threshold-based VGG network is proposed. The model can quickly and accurately complete the automatic identification of COVID-19 cases, and provide help for further control of its spread. By comparing VGG-11, VGG-13 and VGG-16 in convolutional neural network VGG, the automatic diagnosis model for COVID-19 CT image with high accuracy is obtained. On this basis, the accuracy is enhanced from 86% to 89% by modifying the threshold, further improving the accuracy of diagnosis.
作者
翁羽洁
李忠贤
姬宇程
薄素玲
梁莹
WENG Yujie;LI Zhongxian;JI Yucheng;BO Suling;LIANG Ying(College of Computer and Information,Inner Mongolia Medical University,Hohhot 010110,China)
出处
《中国医学物理学杂志》
CSCD
2022年第6期731-736,共6页
Chinese Journal of Medical Physics
基金
内蒙古自治区自然科学基金(2020LH01012)
内蒙古医科大学青年创新基金(YKD2018QNCX014)。
