摘要
目的建立用于软组织瘤磁共振成像(Magnetic Resonance Imaging,MRI)影像辅助诊断的计算机影像组学方法。方法收集了在中国医科大学附属辽宁省肿瘤医院就诊的75例软组织瘤患者的T1增强和T1平扫序列MRI影像数据,采用无监督k-means算法将肿瘤区域分割为高亮度和低亮度亚区。分别提取瘤内亚区及肿瘤完整区域的影像组学特征建立模型,评估各区域诊断价值,建立基于k-最邻近机器学习分类器的融合模型,绘制受试者工作特征(Receiver Operating Characteristic,ROC)曲线和校正曲线,计算ROC曲线下面积(Area Under Curve,AUC)、特异性和敏感性作为模型评价指标,通过决策曲线(Decision Curve Analysis,DCA)分析模型的潜在临床应用价值。结果T1增强序列和T1平扫序列的瘤内高亮度亚区诊断效果均优于低亮度亚区及肿瘤完整区域。通过特征级融合建立的影像组学融合预测模型在训练集和测试集上获得AUC值分别为0.865(特异性=0.763,敏感性=0.763)和0.856(特异性=0.867,敏感性=0.800),校正曲线表明建立的诺模图模型有较好的预测能力,DCA分析进一步表明模型具有良好的潜在临床应用价值。结论本研究提出的基于瘤内亚区分割的MRI影像组学融合模型能对软组织瘤进行有效辅助诊断,具有一定的临床医学应用价值。
Objective To develop and validate a radiomics model for diagnosis of soft-tissue tumors based on multi-parametric magnetic resonance imaging(MRI).Methods T1-CE and T1WI dual-sequence MRI date of 75 soft-tissue tumors patients obtained from Liaoning Cancer Hospital and Institute China Medical University.K-means algorithm was used to divide the soft tissue tumor date into enhanced intensity sub region and non-enhanced intensity sub region.The overall region of interest(ROI)and intratumor segmentation regions were extracted to establish the model.The fusion model was established based on k-nearest neighbor classifier model and drew the receiver operating characteristic(ROC)curve and correction curve,and calculated the area under curve(AUC),specificity and sensitivity as the model evaluation index.The potential clinical value of the model was analyzed by decision curve analysis(DCA).Results The enhanced intensity sub region of T1-CE and T1WI sequences were better than low brightness subregion and overall ROI in the diagnosis of intra-tumor segmentation regions.The AUC values of our computer prediction model were 0.865(sensitivity=0.763,sensitivity=0.763)and 0.856(specificity=0.867,sensitivity=0.800)in the training and test cohorts,respectively.The calibration curve showed that the nomogram model had great prediction ability,and DCA analysis confirmed that the model has good clinical value.Conclusion The dual-sequence MRI image date based on intra-tumor segmentation of soft tissue tumors in this study can effectively assist the diagnosis of soft tissue tumors,and has certain clinical medical application value.
作者
周晓娅
尚圣捷
王颖妮
董越
刘冠宇
罗娅红
蒋西然
ZHOU Xiaoya;SHANG Shengjie;WANG Yingni;DONG Yue;LIU Guanyu;LUO Yahong;JIANG Xiran(School of Basic Medicine,Jining Medical University,Jining Shandong 272067,China;School of Intelligent Medicine,China Medical University,Liaoning Shenyang 110122,China;Department of Medical Imaging,Liaoning Cancer Hospital and Institute,Cancer Hospital of China Medical University,Liaoning Shenyang 110042,China)
出处
《中国医疗设备》
2021年第9期86-90,共5页
China Medical Devices
基金
国家癌症中心攀登基金(NCC201806B011)
健康医疗大数据研究课题(HMB201903101)
2020沈阳市教科工委高校“双服务”重点项目(20200802)
济宁医学院青年教师科研扶持基金(JY2017KJ023)。
关键词
软组织肿瘤
磁共振成像
影像组学
计算机辅助诊断
soft-tissue tumor
magnetic resonance imaging
radiomics
computer-aided diagnosis
作者简介
通信作者:蒋西然,博士,副教授,主要研究方向为医学影像学诊断。邮箱:xrjiang@cmu.edu.cn。
