天津科技大学学报2024，Vol.39Issue(3)：49-55,7.DOI:10.13364/j.issn.1672-6510.20230170
基于CSNet网络的肺炎图像分类研究
Research on Pneumonia Image Classification Based on CSNet Network
摘要
Abstract
X-ray images are an important imaging basis for the diagnosis of pneumonia disease.Due to the diversity of lung diseases,the accuracy of pneumonia diagnosis needs to be further improved.Based on the ConvNeXt network model,a new convolutional neural network model CSNet is proposed for four classification of X-ray pneumonia images.Based on the ConvNeXt network model,the CSNet network changed the original image preprocessing part and pooling layer,and added a feature channel attention module to the convolution block to highlight the pneumonia information in the feature map.On the basis of the attention module,the activation function was modified and compared,and finally the SMU activation function was selected.The effectiveness of each module was proved by ablation experiments,and the effectiveness of the network was verified by comparative experiments with 7 networks.Experimental results show that compared with other network models,the proposed model has the highest accuracy of 99.0%,and the precision and recall are higher.关键词
肺炎/卷积神经网络/图像分类/X射线图像Key words
pneumonia/convolutional neural network/image classification/X-ray image分类
信息技术与安全科学引用本文复制引用
刘玉良,白玉洁..基于CSNet网络的肺炎图像分类研究[J].天津科技大学学报,2024,39(3):49-55,7.基金项目
国家自然科学基金资助项目(52378254) （52378254）
