基于T2WI影像组学模型评估帕金森病患者严重程度的研究  

Assessment of Parkinson's disease severity based on T2-weighted magnetic resonance imagingradiomic modeling

作　　者：韦福利 刘丹[1] 李鑫娅 曹婉君 彭永明 王芳 WEI Fuli;LIU Dan;LI Xinya;CAO Wanjun;PENG Yongming;WANG Fang(Department of Radiology,Yongchuan Hospital of Chongqing Medical University,Chongqing 402160,China;Shanghai United Imaging Intelligence Co.,Ltd,Shanghai 200232,China)

机构地区：[1]重庆医科大学附属永川医院放射科,重庆402160 [2]上海联影智能有限公司,上海200232

出　　处：《磁共振成像》2024年第11期12-16,38,共6页Chinese Journal of Magnetic Resonance Imaging

摘　　要：目的探究基于T2WI影像组学模型评估帕金森病(Parkinson's disease,PD)患者严重程度的价值。材料与方法回顾性分析201例临床诊断为PD患者的临床及影像资料。根据Hoehn-Yahr(H-Y)分级分为早期组113例与中晚期组88例,同时按7∶3比例将患者分为训练集(n=140)和测试集(n=61)。基于T2WI提取并筛选黑质(substantia nigra,SN)、红核(red nucleus,RN)、尾状核(caudatenucleus,CN)、壳核(putamen,PUT)及苍白球(globuspallidus,GP)组学特征,并采用逻辑回归(logistic regression,LR)分类器分别构建相应模型。将临床及影像变量依次纳入单因素和多因素logistic回归中,筛选出影响PD严重程度的独立危险因素并与最佳模型Radscore结合,从而构建联合模型(影像-临床模型)。受试者工作特征(receiver operating characteristic,ROC)曲线用于评估各模型效能,并结合校准曲线评估各模型的校准度。结果从SN中获得了19个最优影像组学特征,在训练集和测试集中AUC(0.817、0.733)均高于RN模型(0.758、0.704)、CN模型(0.712、0.643)、PUT模型(0.713、0.708)及GP模型(0.705、0.708)。脑白质高信号(white matter hyperintensities,WMH)负荷、病程为诊断PD患者严重程度的独立危险因素,与SN模型建立的联合模型AUC为0.865(训练集)、0.836(测试集),两种模型的诊断效能差异无统计学意义(P>0.05)。校准曲线显示6种模型诊断结果与实际结果的一致性较好。结论基于T2WI的SN影像组学特征对于评估PD严重程度具有良好的潜力,并且结合WMH负荷和病程构建的联合模型进一步提高了LR分类器的诊断性能,为临床及时干预治疗提供影像学指导。Objective:To explore the value of assessing the severity of patients with Parkinson's disease(PD)based on the T2WI radiology model.Materials and Methods:A total of 201 patients with a clinical diagnosis of PD were retrospectively collected,according to Hoehn-Yahr(H-Y)stage.They were divided into the early group(n=113)and middle-late group(n=88),into a training set of 140 cases and a test set of 61 cases using a 7∶3 ratio at the same time.We established by extracting and screening radiomics features from substantia nigra(SN),red nucleus(RN),caudate nucleus(CN),putamen(PUT),globus pallidus(GP),and used a logistic regression(LR)classifier to build the corresponding models separately.Clinical and imaging data were sequentially incorporated into univariate and multivariate logistic regression analysis to identify independent risk factors about PD severity.Receiver operating characteristic(ROC)curves were used to assess the efficacy of each model,and calibration curves were used to assess calibration accuracy.Results:19 optimal features on T2WI were selected from SN,and the AUC(0.817,0.733)in the training and test set were higher than those of the RN model(0.758,0.704),the CN model(0.712,0.643),the PUT model(0.713,0.708),and the GP model(0.705,0.708).The white matter hyperintensities burden and the duration of PD were independent risk factors for diagnosing the severity of PD patients,and the AUC of the combined model with the SN model was 0.865(training set)and 0.836(test set),but there is no significant difference in diagnostic performance between the two models.The calibration curve indicates good consistency between the diagnostic results of the six models and the actual outcomes.Conclusions:SN signatures on T2WI achieved better performance in assessing the severity of PD patients.The assessed efficacy of the combined model established by the combined WMH burden and duration of PD was further improved and provide imaging guidance for timely clinical intervention and treatment.

关 键 词：帕金森病 磁共振成像 T2加权成像 影像组学 脑白质高信号 

分 类 号：R445.2[医药卫生—影像医学与核医学] R742.5[医药卫生—诊断学]