影像组学列线图在预测糖尿病肾病患者心血管疾病发生风险中的应用  

Application of radiomics nomogram in predicting risk of cardiovascular diseases in patients with diabetic kidney disease

作　　者：李乐 王硕[1] 庞衍平 王秀艳[1] LI Le;WANG Shuo;PANG Yanping;WANG Xiuyan(Department of Ultrasound,Tongji Hospital Affiliated to Tongji Unversity,Shanghai 200065,China)

机构地区：[1]同济大学附属同济医院超声科,上海200065

出　　处：《外科研究与新技术（中英文）》2024年第3期186-191,共6页Surgical Research and New Technique

摘　　要：目的建立并验证一种基于多模态超声影像组学的列线图模型以预测糖尿病肾病(DKD)患者的心血管疾病(CVD)风险。方法回顾性分析167例DKD患者的基线临床数据和多模态超声图像。其中,超声图像采用3D Slicer软件进行图像分割并提取影像组学特征。通过最小绝对收缩和选择算子(LASSO)算法给每位患者生成影像组学评分。基于这些评分和CVD相关的临床预测因子,建立列线图模型以预测DKD患者未来发生CVD的风险,并通过受试者工作特征(ROC)曲线、校准曲线以及决策曲线分析(DCA)进行内部验证。结果基于LASSO算法,从每位患者的双幅弹性成像中提取5个与CVD发病密切相关的影像组学特征。基于这些特征构建的影像组学评分预测CVD发病风险的ROC曲线下面积(AUC)为0.764。在临床特征方面,年龄、体重指数、低密度脂蛋白胆固醇和糖化血红蛋白是CVD发病的独立影响因子。基于这些因子和影像组学评分建立的影像组学列线图,其预测CVD发病风险的AUC值为0.897,显著高于影像组学评分的单独预测。校准曲线显示列线图预测的CVD发生概率与实际发生率具有较高的一致性。DCA进一步证明了该列线图在不同阈值概率下均能实现临床净收益。结论本研究建立的影像组学列线图模型,其预测DKD患者发展为CVD风险的性能表现出色,有望帮助临床实现DKD患者的个体化管理和早期干预。Objective To establish and validate a nomogram model based on multimodal ultrasound radiomics to predict the risk of cardiovascular diseases(CVD)in patients with diabetic kidney disease(DKD).Methods A retrospective analysis of baseline clinical data and multimodal ultrasound images from 167 DKD patients was conducted.The ultrasound images were segmented and radiomic features were extracted using 3D Slicer software.A radiomics score was generated for each patient using the least absolute shrinkage and selection operator(LASSO)algorithm.Based on these scores and clinical predictors related to CVD,a nomogram model was established to predict the risk of CVD in DKD patients.The model was internally validated using the receiver operating characteristic(ROC)curve,calibration curve,and decision curve analysis(DCA).Results Utilizing the LASSO algorithm,five radiomic features closely related to the incidence of CVD were extracted from the dual mode elastography of each patient.The radiomics score,which was constructed based on these features,predicted the risk of CVD with an area under ROC curve(AUC)of 0.764.In terms of clinical features,age,body mass index,low-density lipoprotein cholesterol,and glycated hemoglobin were identified as independent risk factors for developing CVD.A radiomics nomogram,built based on these factors and the radiomics score,significantly outperformed the radiomics score alone in predicting the risk of CVD,with an AUC of 0.897.The calibration curve demonstrated a high consistency between the predicted probability of CVD occurrence by the nomogram and the actual incidence.DCA further validated that the nomogram achieved a net clinical benefit at different threshold probabilities.Conclusion The radiomics nomogram established in this study demonstrates excellent performance in predicting the risk of DKD patients developing CVD.It holds promise to aid clinicians in achieving individualized management and early intervention for DKD patients.

关 键 词：糖尿病肾病 心血管疾病 剪切波弹性成像 影像组学 列线图 

分 类 号：R587.2[医药卫生—内分泌]