摘要
目的探讨基于多参数MRI的影像学特征、膀胱影像报告和数据系统(VI-RADS)评分构建的模型对预测膀胱癌病理分级的价值。方法回顾性分析明确病理分级的膀胱尿路上皮癌患者的临床及影像资料,利用多因素逻辑回归分析筛选出病理分级的独立预测因素,建立临床模型。从T_(2)WI、扩散加权成像(DWI)、动态对比增强(DCE)图像中提取影像组学特征,对特征进行降维和筛选后建立影像组学模型。联合独立预测因素与影像组学构建联合模型。结果筛选出VI-RADS评分为独立预测因素,建立了VI-RADS模型。T_(2)WI、DWI、DCE 3个序列最终分别筛选出10、12、12个影像组学特征。在训练集和验证集中,T_(2)WI模型的敏感度(90.78%、80.00%)、特异度(89.16%、80.56%)、准确度(90.18%、80.21%)以及ROC曲线下面积大小(AUC)(0.9632、0.9157)均高于DWI和DCE模型。根据VI-RADS评分及T_(2)WI序列的影像组学评分建立了联合模型。该模型在训练集和验证集中的敏感度分别为91.49%、81.67%;特异度分别为87.95%、88.89%;准确度分别为90.18%、84.38%;AUC分别为0.9633、0.9398。结论VI-RADS评分是膀胱癌病理分级的独立预测因素,VI-RADS评分越高,病理分级为高级别的风险越大。影像组学模型和联合模型可以在术前无创预测膀胱的病理分级。
Objective To investigate the value of models constructed based on the imaging characteristics of multiparametric MRI and the Vesical Imaging-Reporting and Data System(VI-RADS)score for predicting the pathological grading of bladder cancer.Methods Clinical and MRI data of patients with urothelial carcinoma of the bladder with clear pathological grading were retrospectively analyzed.Multifactorial logistic regression analysis was used to identify independent predictors of pathological grading and establish a clinical model.Radiomics features were extracted from T_(2)-weighted imaging(T_(2)WI),diffusion-weighted imaging(DWI),and dynamic contrast-enhanced imaging(DCE)sequences,and radiomics models were developed after dimensionality reduction and feature selection.A combined model was then constructed by integrating independent predictors with radiomics features.Results The VI-RADS score was identified as an independent predictor,and a VI-RADS model was developed.A total of 10,12,and 12 radiomics features were selected from T_(2)WI,DWI,and DCE sequences,respectively.In both the training and validation sets,the T_(2)WI model exhibited higher sensitivity(90.78%,80.00%),specificity(89.16%,80.56%),accuracy(90.18%,80.21%),and area under the receiver operating characteristic curve(AUC)(0.9632,0.9157)compared to the DWI and DCE models.A combined model was developed based on the VI-RADS score and the radiomics score(rad-score)of the T_(2)WI sequence.This combined model achieved sensitivity of 91.49%and 81.67%,specificity of 87.95%and 88.89%,accuracy of 90.18%and 84.38%,and AUC of 0.9633 and 0.9398 in the training and validation sets,respectively.Conclusion The VI-RADS score is an independent predictor of pathological grading in bladder cancer,with higher VI-RADS scores indicating a greater risk of high-grade pathology.The T_(2)WI radiomics model and the combined model can effectively predict the pathological grade of bladder cancer preoperatively.
作者
尚芸芸
刘灿丽
周小龙
郝金钢
SHANG Yunyun;LIU Canli;ZHOU Xiaolong(Dept.of Radiology,The 2nd Affiliated Hospital of Kunming Medical University,Kunming,Yunnan Province 650101,P.R.China)
出处
《临床放射学杂志》
北大核心
2025年第9期1699-1706,共8页
Journal of Clinical Radiology
基金
云南省教育厅科学研究基金项目(编号:2024J0348)
昆医联合专项-面上项目(编号:202401AY070001-337)。
