基于多参数磁共振成像影像组学融合模型对骨肉瘤肺转移预测价值  

Predictive value of a multiparametric magnetic resonance imaging radiomics model for lung metastases in osteosarcoma

作　　者：潘丹 林恒山 鲁璎 林敏贵 林鹤 Pan Dan;Lin Hengshan;Lu Ying;Lin Minggui;Lin He(Department of Radiology,Fuzhou Second General Hospital,Fujian 350000,China)

机构地区：[1]福州市第二总医院影像科,福州350000

出　　处：《实用医技杂志》2025年第4期250-254,I0001,I0002,共7页Journal of Practical Medical Techniques

基　　金：福州市科技计划项目(2022-S-056);福建省自然科学基金(2023J011513)。

摘　　要：目的探讨基于多参数核磁共振成像(MRI)影像组学融合模型对骨肉瘤肺转移预测价值。方法回顾性分析2017年1月至2023年9月通过病理学检查明确诊断为骨肉瘤94例患者临床及影像资料,按照7∶3比例随机分为训练集与验证集。纳入常规影像特征构建临床影像学模型,通过多因素logistic回归分析筛选出MR常规影像特征独立预测因子,提取多参数的MRI图像影像组学特征,在训练集中采用最小绝对收缩和选择算子(LASSO)筛选出与骨肉瘤肺转移最相关特征用于模型构建。构建基于联合MRI常规影像特征与影像组学模型融合模型,通过验证组进行验证。采用受试者工作特征(ROC)曲线下面积(AUC)评估模型效能、决策曲线评估临床价值。结果骨肉瘤肺转移与无肺转移患者间,常规MRI影像学特征中血管周围浸润及关节侵犯差异有统计学意义。基于多参数MRI,行LASSO逻辑回归分析,提取了4个与骨肉瘤肺转移相关的影像组学特征构成影像组学标签。训练组中,影像组学标签预测骨肉瘤肺转移ROC曲线AUC为0.699,准确率为0.929,灵敏度为0.652;验证组中AUC为0.731,准确率为0.667,灵敏度为0.786。融合常规影像学特征(血管周围浸润和关节侵犯)和影像组学标签构建列线图,在训练组、验证组预测骨肉瘤肺转移的ROC AUC分别为0.699和0.731。决策曲线表明,当风险阈值>23%时,采用融合模型预测骨肉瘤肺转移方面比其他单序列模型具有更高净收益。结论基于多参数MRI影像组学模型及MRI常规影像学特征,构建预测模型,对骨肉瘤进行个体化肺转移具有良好预测效能,能够为骨肉瘤患者临床诊疗决策制定提供有价值指导信息,从而实现骨肉瘤精准诊疗评估。Objective To evaluate predictive value of a multiparametric magnetic resonance imaging(MRI)radiomics model for lung metastases in Osteosarcoma.Methods The imaging data of patients with osteosarcoma from January 2017 to September 2023 were retrospectively analyzed,a total of 94 osteosarcoma cases were randomly divided into training and validation cohorts at 7∶3 ratio.The features of conventional images were incorporated into the clinical imaging model,independent predictors of MRI features of conventional images were screened by multivariate logistic regression analysis,image radiomics features of multiparametric MRI images were extracted.In the training set,least absolute shrinkage and selection operators were used to screen out the features most associated with osteosarcoma lung metastasis for model construction.The nomogram model based on conventional MRI image and the fusion model were constructed and verified.The area under the receiver operating characteristic curve(AUC)and calibration curve were used to evaluate the efficacy and clinical benefit of the model.Results There were statistically significant differences in the routine MRI features of joint invasion and perivascular infiltration between patients with osteosarcoma with lung metastasis and those without lung metastasis.Based on multiparametric MRI and LASSO logistic regression analysis,the radiomics signature was built using 4 valuable selected features that were significantly associated with recurrent lung metastases in osteosarcoma.In the training group,the ROC curve AUC for predicting osteosarcoma lung metastasis was 0.699,the accuracy was 0.929,and the sensitivity was 0.652.In the verification group,the AUC was 0.731,the accuracy was 0.667,and the sensitivity was 0.786.A nomogram was constructed by combining conventional imaging features(perivascular infiltration and joint invasion)and radiomics labels.Predicting osteosarcoma lung metastasis of ROC curve in the training group the validation group respectively was 0.699 and 0.731.Decision curve

关 键 词：骨肉瘤 磁共振成像 影像基因组学 肺转移 

分 类 号：R738.1[医药卫生—肿瘤]