摘要
目的:探讨基于磁共振ADC图的影像组学模型对诊断前列腺癌侵袭度的价值。方法:回顾性分析2018年1月至2019年5月在建湖医院,经手术病理证实且能确定Gleason分级的42例患者的ADC图像,将癌灶分为高危组(Gleason评分≥8)和低中危组(Gleason评分≤7)2组。其中中低危21例、高危21例。应用ITK-SNAP软件勾化感兴趣区(ROI),将ADC图像导人Analysis-Kinetics分析软件,进行影像特征提取。采用Lasso回归分析进行特征降维。通过LASSO降维筛选出的特征和相应加权系数乘积的线性组合来建立鉴别中低危、高危前列腺癌的模型,绘制ROC曲线评价模型鉴别中低危、高危前列腺癌的预测效能。结果:共提取396个影像组学特征,通过特征筛选后最后筛选出7个影像组学特征。建模后影像组学特征对鉴别中低危、高危前列腺癌具有较好的预测效能,预测模型在训练组中鉴别效能的曲线下面积、准确度、敏感度、特异度、阳性预测值和阴性预测值分别为0.97、93.3%、93.3%、93.3%、0.93和0.93;在验证组中的曲线下面积、准确度、敏感度、特异度、阳性预测值和阴性预测值分别为0.97、91.7%、83.3%、100.0%、1和0.86;结论:基于磁共振ADC图的影像组学模型对前列腺癌Gleason分级具有诊断价值。
Purpose:To develop and validate a ADC-based radiomics predictive model for predicting aggressive prostate cancer.Methods:The ADC maps of 42 patients(low and intermediate risk group:21 cases;high risk group:21 cases)with histologically confrmed prostate cancer fom January 2018 to March 2019 were analyzcd retrospectively.The lesions were clasified into high risk group(Gleason score≥8)and Low-intermediate risk group(Gleason score≤7).Among them,21 cases were with low-intermediate risk and 21 cases were with high risk.The software ITK-SNAP was used to draw the region of interest(ROD),and the radiomic features based on ADC map were generated automatically from Analysis-Kinetics(GE Healthcare).LASSO regression model was used for data dimension reduction.The linear combination of the features selected by LASSO dimensionality reduction screening and the corresponding weighted coefficient product was used to establish prediction model.The model performance was assessed with respect to discrimination using the area under the curve(AUC)of receiver operating characteristic(ROC)analysis.Results:.Three hundred and ninety-six radiomics features were extracted automatically by software and 7 features were left after redundancy reduction step.Radiomics features after modeling had a good predictive value for the identification of low-risk and high-risk prostate cancer.The prediction model showed good discrimination in both primary dataset(AUC=0.97,95%;Accuracy=93.3%,sensitivity=93.3%,specificity=93.3%,positive predictive vaIue=0.93,negative.predictive value=0.93)and independent validation dataset(AUC=0.97,95%;Accuracy=91.7%,sensitivity=83.3%,specificity=93.3%,positive predictive value=0.93,negative predictive value=0.93).Conclusion:The radiomics model could provide important reference in assessment of prostate cancer aggressiveness.
作者
许晴
陆大军
袁为标
李海峰
许明明
高辉
XU Qing;Lu Da-jun;YUAN Wei-biao;LI Hai-feng;XU Ming-ming;GAO Hui(Department of Radiology,Northern Jiangsu People's Hospital,Clinical Medical School of Yangzhou University;Department of Radiology,Jiangsu jianhu People's Hospital,Affliated Hospital of Nantong University)
出处
《中国医学计算机成像杂志》
CSCD
北大核心
2020年第2期149-153,共5页
Chinese Computed Medical Imaging
基金
扬州市科技计划项目:YZ2017066
盐城市科技计划项目:YK2015076。