基于DenseNet的肺部医学影像智能检测方法研究  

Research on intelligent detection method of pulmonary medical imaging based on DenseNet

作　　者：郭邵宁 张伟[1] 董延华[1] GUO Shao-ning;ZHANG Wei;DONG Yan-hua(School of Mathematics and Computer,Jilin Normal University,Siping 136000,China)

机构地区：[1]吉林师范大学数学与计算机学院,吉林四平136000

出　　处：《哈尔滨商业大学学报（自然科学版）》2023年第2期164-168,共5页Journal of Harbin University of Commerce：Natural Sciences Edition

基　　金：中国高校产学研创新基金-新一代信息技术创新项目(2020ITA05017)。

摘　　要：研究基于智能检测医学图像分类的肺部X线胸片检测可快速准确的对患者进行临床分析和诊断,能够实现病症早发现早治疗,为临床决策提供高效可靠的支持.以包含医学图像的Chest X-ray的数据集为基础,对其进行了缩放、随机裁剪、水平翻转、维度转换、像素值归一化处理等预处理后,创新地利用DenseNet-121模型进行了肺炎的分类实验.实验证明基于DenseNet的肺部医学图像的测试准确率达到97.4%,召回率、F1值等量化指标也优于已有的ResNet方法,证明DenseNet可作为肺炎医学影像智能检测的一种有效方式.Under the current global situation,this paper studied on lung X-ray detection based on intelligent detection of medical image classification could quickly and accurately conduct clinical analysis and diagnosis of patients,realize early detection and early treatment of symptoms,and provide efficient and reliable support for clinical decision-making.Based on the dataset of Chest X-raycontaining medical images,after scaled,random cropped,horizontal flipped,dimension transformation,pixel value normalization,etc.It innovatively used the DenseNet-121 model to perform a pneumonia classification experiment.Experiments showed that the test accuracy rate of pulmonary medical images based on DenseNet reached 97.4%,and the quantitative indicators such as Recall and F 1 score were also better than the existing ResNet methods,which proved that DenseNet could be used as an effective method for intelligent detection of pneumonia medical images.

关 键 词：医学影像 深度学习 图像分类 卷积神经网络 肺部 DenseNet网络 

分 类 号：TP311.1[自动化与计算机技术—计算机软件与理论]