摘要
目的:利用深度学习技术建立影像质控分类算法模型,提高对X线胸片异物的检出效能。方法:利用深度学习卷积神经网络ResNet-50开发全自动正位X线胸片影像异物检测模型,实现正常X线胸片影像与存有异物影像的分类,并通过热力图的形式显示具体质控识别点的位置和范围。结果:该模型在验证集的准确率达96.9%,AUC为0.994(95%CI:0.993~0.995),采用AI辅助质控对每张X线胸片进行分类的速度较人工质控快10倍。结论:模型的准确性、速度和轻量级架构使得该模型适合嵌入医院PACS系统中进行质量控制管理。
Objective To establish a classification algorithm model for influencing quality control by using deep learning technology,so as to improve the efficiency of chest X-ray foreign object detection.Methods A fully-automatic orthotopic chest X-ray films foreign body detection model was developed using the deep learning convolutional neural network ResNet-50,which can classify the normal X-ray chest images from those with foreign objects,and display the location and range of specific quality control identification points by means of thermal diagram.Results The accuracy of the model in the validation set was 96.9%,with AUC(areas under the curve) of 0.994,95% CI(confidence interval)of 0.993 to 0.995.Compared with manual quality control,AI-assisted quality control is 10 times faster in classifying each chest X-ray film.Conclusion The accuracy,speed,and lightweight architecture of the model make it suitable for being embedded into the hospital PACS system for quality control and management.
作者
李裴
刘慧
钱宝鑫
朱莉
Li Pei;Liu Hui;Qian Baoxin;Zhu Li(Information Department,the Affiliated Hospital of Xuzhou Medical University,Xuzhou 221000,Jiangsu Province,China;Huiying Medical Technology Co.,Ltd.)
出处
《中国数字医学》
2022年第11期32-37,共6页
China Digital Medicine
关键词
医学影像
质量控制
异物检测
Medical imaging
Quality control
Foreign object detection