现代电子技术2024，Vol.47Issue(1)：62-68,7.DOI:10.16652/j.issn.1004-373x.2024.01.011
改进ConvNeXt的肝囊型包虫病超声图像五分类研究
Study on five classifications of ultrasonic images of hepatic cystic echinococcosis based on improved ConvNeXt
摘要
Abstract
Hepatic echinococcosis is an important zoonotic parasitic disease that seriously harms human and animal health.Ultrasonic examination is the first choice for hepatic echinococcosis.The deep learning technology can effectively reduce manual error,reduce cost and improve diagnostic efficiency.The proposed method is based on the ConvNeXt model and combined with focal loss function and Lion optimizer.The CBAM(convolutional block attention module)is introduced to construct CLCFNet model to realize early screening and accurate diagnosis of hepatic cystic echinococcosis(HCE).The ablation experiment on ultrasonic image data set of HCE shows that the classification accuracy,precision,recall,specificity and F1 index of the proposed model are improved by 4.3％,21％,25％,4％ and 26％ on average in comparison with those of the benchmark model,and the comparison experiment also verifies that the proposed method is superior to the existing popular methods.The improved algorithm reduces the model reasoning time significantly,enhances the steady-state performance of model training,so it can realize rapid and accurate classification and identification of HCE.关键词
肝囊型包虫病/超声图像/ConvNeXt/焦点损失函数/Lion优化器/注意力机制Key words
HCE/ultrasonic image/ConvNeXt/focal loss function/Lion optimizer/attention mechanism分类
信息技术与安全科学引用本文复制引用
热娜古丽·艾合麦提尼亚孜,米吾尔依提·海拉提,王正业,茹仙古丽·艾尔西丁,严传波..改进ConvNeXt的肝囊型包虫病超声图像五分类研究[J].现代电子技术,2024,47(1):62-68,7.基金项目
国家自然科学基金项目(地区项目):新疆高发病肝包虫疾病计算机辅助诊断方法的研究(81560294) （地区项目）
省部共建中亚高发病成因与防治国家重点实验室开放课题(SKLHIDCA-2020-YG2) （SKLHIDCA-2020-YG2）
