基于CT的影像组学列线图对布鲁氏菌性脊柱炎和脊柱结核的鉴别价值  

The value of radiomics nomogram based on CT for differentiating brucella spondylitis from spinal tuberculosis

作　　者：奇丽霞 张凤翔[1] 杜海[1] 贾艳荣[1] 云暄 吴奇飞 刘波 张芳[1] QI Lixia;ZHANG Fengxiang;DU Hai;JIA Yanrong;YUN Xuan;WU Qifei;LIU Bo;ZHANG Fang(CTMRI Room,Ordos Central Hospital,Ordos,Inner Mongolia 017000,China;Department of Radiology,Ordos Second People’s Hospital,Ordos,Inner Mongolia 017000,China)

机构地区：[1]鄂尔多斯市中心医院CT-MRI室,内蒙古鄂尔多斯017000 [2]鄂尔多斯市第二人民医院影像医学科,内蒙古鄂尔多斯017000

出　　处：《中国实验诊断学》2025年第4期440-448,共9页Chinese Journal of Laboratory Diagnosis

基　　金：鄂尔多斯市科技计划项目(鄂财教指[2019]501号)。

摘　　要：目的基于CT的影像组学列线图鉴别布鲁氏菌性脊柱炎(BS)和脊柱结核(TBS)。方法回顾性分析2018年1月至2023年3月94例布鲁氏菌性脊柱炎及脊柱结核患者的临床资料及CT图像,按照7∶3随机分为训练集和测试集。基于训练集数据,对临床及常规CT征象进行组间差异性分析,并采用多因素logistic回归构建临床模型。基于CT图像提取影像组学特征,并筛选出最优特征计算影像组学评分(Radscore)。联合Radscore、临床特征构建联合模型并绘制列线图。采用受试者工作特征曲线下面积(AUC)、特异性、敏感性评估模型的诊断效能。结果训练集中性别、死骨及腰大肌脓肿被纳入临床模型,筛选得到5个最优特征。最终列线图AUC、敏感性及特异性分别为0.894、82.1%、94.7%。测试集中列线图AUC、敏感性及特异性分别达到0.786、61.5%、93.3%;相较于Radscore(AUC=0.750,敏感性=66.6%,特异性=78.9%)及临床模型(AUC=0.714,敏感性=66.6%,特异性=73.7%),列线图的AUC及特异性均更高。结论基于CT构建的影像组学列线图对BS和TBS具有较高的鉴别价值,有助于辅助临床医师快速识别BS并给予及时干预。Objective To differentiate Brucellar Spondylitis(BS)from Tuberculous Spondylitis(TBS)using CTbased radiomics nomograms.Methods A retrospective analysis was conducted on the clinical data and CT images of 94 patients with brucella spondylitis and tuberculous spondylitis from January 2018to March 2023.The data were randomly divided into a training set and a testing set in a 7∶3ratio.Based on the training set data,the differences between clinical and conventional CT signs were analyzed between groups,and multi-factor logistic regression was used to construct a clinical model,and the optimal features were selected to calculate the radiomics score(Radscore).Finally,ajoint model was constructed based on Radscore and clinical characteristics and a nomogram was drawn.The area under the receiver operating characteristic curve(AUC),specificity,and sensitivity were used to evaluate the diagnostic performance of each model.Results In the training set,gender,sequestrum,and psoas abscess were included in the clinical model,5optimal features were selected.The final nomogram achieved an AUC,sensitivity,and specificity of 0.894,82.1%,and 94.7%,respectively.In the testing set,the nomogram achieved an AUC,sensitivity,and specificity of 0.786,61.5%,and 93.3%,respectively.Compared to the Radscore(AUC=0.750,sensitivity=66.6%,specificity=78.9%)and the clinical model(AUC=0.714,sensitivity=66.6%,specificity=73.7%),the nomogram demonstrated higher AUC and specificity.Conclusion The CT-based radiomics nomogram has a high identification value for BS and TBS,which can help clinicians quickly identify BS and provide timely intervention.

关 键 词：布鲁氏菌性脊柱炎 脊柱结核 影像组学 列线图 计算机断层扫描 

分 类 号：R445[医药卫生—影像医学与核医学]