基于CT影像组学结合机器学习预测复杂性与非复杂性急性阑尾炎  

Prediction of complicated and uncomplicated acute appendicitis using CT-based radiomics combined with machine learning models

作　　者：李文武 卜慧萍 姚明哲 竺仕林 杨雅量 葛星月 唐乾利[3] Li Wenwu;Bu Huiping;Yao Mingzhe;Zhu Shilin;Yang Yaliang;Ge Xingyue;Tang Qianli(Graduate School,Youjiang Medical University for Nationalities,Baise 533000,China;Radiology Department,First Affiliated Hospital of Guangxi Medical University,Nanning 530021,China;Affiliated Hospital of Youjiang Medical University for Nationalities,Baise 533000,China)

机构地区：[1]右江民族医学院研究生学院,广西百色533000 [2]广西医科大学第一附属医院放射科,广西南宁530021 [3]右江民族医学院附属医院,广西百色533000

出　　处：《右江民族医学院学报》2024年第1期57-64,共8页Journal of Youjiang Medical University for Nationalities

基　　金：广西特聘专家项目(桂人才通字〔2019〕13号);广西医学高层次领军人才培养“139”计划项目资助(桂卫科教发[2018]22号)。

摘　　要：目的探讨和验证不同影像组学模型在复杂性与非复杂性急性阑尾炎的术前鉴别诊断中的价值。方法回顾性分析212例经手术病理证实为急性阑尾炎患者的临床资料及CT平扫图像,从CT图像中提取影像组学特征,经过特征的降维和筛选,分别采用Logistic回归、支持向量机(SVM)和随机森林等算法构建影像组学模型,通过比较受试者工作特征(ROC)曲线下面积(AUC)、准确度、95%置信区间(95%CI)等指标获得最佳的影像组学模型。此外,应用单因素和多因素Logistic回归分析来筛选临床特征并建立临床模型。通过多变量逻辑回归将影像组学标签与临床标签相结合,构建一个组合模型。最后,采用ROC曲线分析来评估模型的性能,并利用决策曲线分析(DCA)来评估模型的临床价值。结果最终筛选出年龄和C反应蛋白2个临床特征。从每个患者CT图像共提取出1834个影像组学特征,并确定了16个最有价值的影像组学特征。在影像组学模型中,SVM表现出最佳的预测效率和稳定性,训练集和测试集的AUC分别为0.916(95%CI为0.862~0.970)和0.842(95%CI为0.739~0.945)。在所有模型中,组合模型的诊断效能最佳,训练集和测试集的AUC分别为0.943(95%CI为0.896~0.990)和0.855(95%CI为0.759~0.951)。DCA提示组合模型具有更好的预测性能和临床价值。结论结合影像组学特征与临床特征的组合模型对复杂性与非复杂性急性阑尾炎具有良好的预测能力,可以为临床决策提供了一种无创、有效的方法,避免不必要的手术切除。Objective To explore and validate the value of various radiomics models in the preoperative discrimination between complicated and uncomplicated acute appendicitis.Methods retrospective analysis of clinical data and CT plain images from 212 surgically confirmed acute appendicitis cases was conducted.Radiomic features were extracted from CT images,following feature reduction and selection.Various algorithms including Logistic Regression,Support Vector Machine(SVM),and Random Forest were employed to construct radiomics models.Model performance was evaluated by comparing metrics such as the Area Under the Receiver Operating Characteristic(ROC)Curve(AUC),accuracy,and 95%confidence intervals(95%CI)to determine the optimal radiomics model.Additionally,univariate and multivariate Logistic Regression analyses were performed to select clinical features and establish a clinical model.A combined model was developed by integrating radiomic labels with clinical labels using multivariate logistic regression.Finally,ROC curve analysis was conducted to assess the model's performance,and Decision Curve Analysis(DCA)was conducted to evaluate its clinical utility.Results The final selection included age and C-reactive protein as the two clinical features.From each patient's CT images,a total of 1834 radiomic features were extracted,16 most valuable features identified.Among the radiomics models,SVM exhibited the highest predictive efficiency and stability,with AUCs of 0.916(95%CI:0.862~0.970)in the training set and 0.842(95%CI:0.739~0.945)in the test set.In all models,the combined model showed the best diagnostic performance,with AUCs of 0.943(95%CI:0.896~0.990)in the training set and 0.855(95%CI:0.759~0.951)in the test set.DCA suggested that the combined model had superior predictive performance and clinical value.Conclusion The combined model integrating radiomic and clinical features demonstrates robust predictive ability in distinguishing between complicated and uncomplicated acute appendicitis,providing a non-invasive and effe

关 键 词：影像组学 机器学习 急性阑尾炎 临床特征 

分 类 号：R445[医药卫生—影像医学与核医学] R656.8[医药卫生—诊断学]