新型冠状病毒是一种RNA病毒.利用人工智能技术分析新冠肺炎患者肺部图像,辅助医生快速有效地诊断与评估感染的严重程度,可以有效防止病情恶化.因此,提出一种改进的U-Net模型,在编码器部分采用预训练好的EfficientNet模型替换原有结构,在解码器部分添加scSE(spatial and channel Squeeze & Excitation)空间注意力模块.实验结果表明,本模型在新冠肺炎CT图像分割方面优于同等参数级别的其他模型,其DSC、MIoU、ACC、SEN和SPE五个指标均高于U-Net的相应指标,可视化病灶区域分割效果良好.
Research on the Segmentation Method of COVID-19 Image Based on Improved U-Net Network
COVID-19 is an RNA virus.Analyzing the images of COVID-19 of lung patients by applying artificial intelligence technology to assist doctors to quickly diagnose and evaluate the severity of infection,which can effectively prevent the deterioration of the disease.This paper proposes an improved U-Net model,in which the original encoder is replaced by the pre-trained EfficientNet model and add the spatial attention module scSE(spatial and channel Squeeze & Excitation)to the decoder.The experimental results show that the proposed model is superior to other models in the same parameter level in CT image segmentation of COVID-19.The five indexes of DSC,MIoU,ACC,SEN and SPE are higher than those of U-Net,and indicating good visual segmentation effect of focal region.
NETL
NSTL
万方数据
