Multi-scale input mirror network for tuberculosis detection in CXR image  

用于CXR图像中结核病检测的多尺度输入镜像网络

作　　者：XING Guangxin FAN Jingjing ZHENG Yelong ZHAO Meirong 邢广鑫;樊晶晶;郑叶龙;赵美蓉(天津大学精密测试技术及仪器全国重点实验室,天津300072;中国人民解放军总医院医疗保障中心药剂科,北京100853)

机构地区：[1]State Key Laboratory of Precision Measuring Technology and Instruments,Tianjin University,Tianjin 300072,China [2]Department of Pharmacy,Medical Supplies Center,PLA General Hospital,Beijing 100853,China

出　　处：《Journal of Measurement Science and Instrumentation》2025年第1期1-10,共10页测试科学与仪器(英文版)

基　　金：supported by the Joint Fund of the Ministry of Education for Equipment Pre-research(No.8091B0203);National Key Research and Development Program of China(No.2020YFC2008700)。

摘　　要：Computer-aided diagnosis(CAD)can detect tuberculosis(TB)cases,providing radiologists with more accurate and efficient diagnostic solutions.Various noise information in TB chest X-ray(CXR)images is a major challenge in this classification task.This study aims to propose a model with high performance in TB CXR image detection named multi-scale input mirror network(MIM-Net)based on CXR image symmetry,which consists of a multi-scale input feature extraction network and mirror loss.The multi-scale image input can enhance feature extraction,while the mirror loss can improve the network performance through self-supervision.We used a publicly available TB CXR image classification dataset to evaluate our proposed method via 5-fold cross-validation,with accuracy,sensitivity,specificity,positive predictive value,negative predictive value,and area under curve(AUC)of 99.67%,100%,99.60%,99.80%,100%,and 0.9999,respectively.Compared to other models,MIM-Net performed best in all metrics.Therefore,the proposed MIM-Net can effectively help the network learn more features and can be used to detect TB in CXR images,thus assisting doctors in diagnosing.计算机辅助诊断可用于检测结核病例,为放射科医生提供更准确、更高效的诊断解决方案。结核病胸部X光(Chest Xray,CXR)图像中的各种干扰噪声是这一分类任务的主要挑战。本研究旨在提出一种结核病CXR图像检测的高性能模型,即基于CXR图像对称性的多尺度输入镜像网络(Multi-scale input mirror network,MIM-Net),它由多尺度输入特征提取网络和镜像损失组成:多尺度图像输入可增强特征提取,而镜像损失则通过自监督提高网络性能。该模型在一个公开的结核CXR图像分类数据集上通过5倍交叉验证进行了评估,其准确率、灵敏度、特异性、阳性预测值、阴性预测值和曲线下面积(Area under curve,AUC)分别达到了99.67%、100%、99.60%、98.00%、100%和0.9999。与其他模型相比,MIM-Net在所有指标上都表现最佳。因此,我们提出的MIM-Net可以有效帮助网络学习更多特征,并可用于检测CXR图像中的结核病,从而帮助医生做出诊断。

关 键 词：computer-aided diagnosis(CAD) medical image classification deep learning feature symmetry mirror loss 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]