双参数MRI影像组学对前列腺癌国际泌尿病理协会分级的预测价值  

Radiomic signature based on bi-parametric MRI predicting International Society of Urological Pathology grading in prostate cancer

作　　者：张永胜[1] 葛玉杰 李志平 瞿华[1] 高晨 崔凤[1] 陈明涛[4] 许茂盛[3] ZHANG Yongsheng;GE Yujie;LI Zhiping;QU Hua;GAO Chen;CUI Feng;CHEN Mingtao;XU Maosheng(Department of Radiology,Hangzhou TCM Hospital Affiliated to Zhejiang Chinese Medical University,Hangzhou 310007,China;Department of Radiology,Hangzhou Fuyang Hospital of Traditional Chinese Medicine,Hangzhou 311400,China;Department of Medical Imaging,the First Affiliated Hospital of Zhejiang Chinese Medical University,Hangzhou 310006,China;Department of Pathology,the First Affiliated Hospital of Zhejiang Chinese Medical University,Hangzhou 310006,China)

机构地区：[1]浙江中医药大学附属杭州市中医院放射科,浙江杭州310007 [2]杭州市富阳区中医院放射科,浙江杭州311400 [3]浙江中医药大学附属第一医院医学影像科,浙江杭州310006 [4]浙江中医药大学附属第一医院病理科,浙江杭州310006

出　　处：《实用放射学杂志》2023年第12期1995-2000,共6页Journal of Practical Radiology

基　　金：浙江省中医药科技计划项目(2022ZA102,2022KY996)。

摘　　要：目的探讨基于术前双参数MRI影像组学对前列腺癌(PCa)国际泌尿病理协会(ISUP)分级的预测价值。方法回顾性分析经病理证实的165例PCa患者。根据ISUP分级系统将PCa患者分为5个亚组:G1组(Gleason评分=6分)、G2组(Gleason评分=3+4分)、G3组(Gleason评分=4+3分)、G4组(Gleason评分=8分)、G5组(Gleason评分=9或10分)。从每例患者的T2WI、扩散加权成像(DWI)和表观扩散系数(ADC)图中共提取3948个影像组学特征。对纳入的PCa患者根据Gleason评分≥4+3分或≤3+4分进行2分类,采用最小冗余最大相关(mRMR)和最小绝对收缩和选择算子(LASSO)2种算法进行特征降维,然后构建影像组学标签(Rad-score)。采用Spearman秩相关分析评价Rad-score与ISUP分组间的相关性。采用Kruskal-Wallis检验比较Rad-score在ISUP分级系统内两两比较的差异性。结果纳入3948个影像组学特征进行mRMR和LASSO降维,最终筛选出11个价值较大的影像组学特征及其相关系数组成Rad-score。Rad-score与ISUP分级之间有良好的相关性(r=0.53,P<0.05),Rad-score在G1、G2组与G3、G4、G5组间差异有统计学意义(P<0.05),除此之外ISUP分级系统内任2组间差异均无统计学意义(P>0.05)。Rad-score诊断ISUP G1~G5组的受试者工作特征(ROC)曲线的曲线下面积(AUC)分别为0.827、0.762、0.563、0.657、0.698。结论双参数MRI影像组学可用于预测ISUP分级系统中G1~G2级PCa患者。Objective To investigate the predictive value of preoperative bi-parametric MRI radiomics for the International Society of Urological Pathology(ISUP)grading of prostate cancer(PCa).Methods One hundred and sixty-five patients with PCa confirmed by pathology were analyzed retrospectively.According to the ISUP grading system,PCa patients were divided into five subgroups:G1 group(Gleason score=6),G2 group(Gleason score=3+4),G3 group(Gleason score=4+3),G4 group(Gleason score=8)and G5 group(Gleason score=9 or 10).A total of 3948 radiomics features were extracted from T2WI,diffusion weighted imaging(DWI),and apparent diffusion coefficient(ADC)images of each patient.Patients were classified into two categories based on Gleason score≥4+3 or≤3+4.A radiomics signature(Rad-score)was constructed after reduction of dimension by the minimum redundancy maximum relevance(mRMR)and the least absolute shrinkage and selection operator(LASSO).The Spearman rank correlation test was used to evaluate the correlation between Rad-score and ISUP grading groups.The Kruskal-Wallis test was used to compare the difference of Rad-score among the five groups.Results Eleven most valuable features were selected as the Rad-score after reducing the dimension by mRMR and LASSO algorithm.Moderate correlation existed between Rad-score and ISUP grading(r=0.53,P<0.05).There were significant differences in Rad-score between G1 and G2 groups and G3,G4 and G5 groups(P<0.05),no significant difference existed between the remained two groups(P>0.05).The area under the curve(AUC)of receiver operating characteristic(ROC)curve for Rad-score were 0.827,0.762,0.563,0.657,0.698 for G1,G2,G3,G4 and G5 groups,respectively.Conclusion The radiomics based on bi-parametric MRI can be used to predict grade 1,2 PCa patients in the ISUP grading system.

关 键 词：前列腺癌 磁共振成像 影像组学 国际泌尿病理协会分级 

分 类 号：R737.25[医药卫生—肿瘤] R445.2[医药卫生—临床医学] R445