Zhou Qi, Ma Lu, Yu Yanhang, Zhang Chuanao, Ouyang Jun, Mao Caiping, Zhang Zhiyu
Department of Urology, The First Affiliated Hospital of Soochow University, Suzhou, China.
Department of Reproductive Medicine Center, The First Affiliated Hospital of Soochow University, Suzhou, China.
Front Oncol. 2025 Aug 27;15:1661979. doi: 10.3389/fonc.2025.1661979. eCollection 2025.
This study aimed to develop a preoperative predictive model for pathological grading of bladder urothelial carcinoma by integrating multi-parameter, thin-slice enhanced computed tomography (CT) texture features with relevant clinical indicators.
CT images and clinical data were retrospectively collected from 372 individuals diagnosed with bladder urothelial carcinoma at our institution between January 2015 and October 2020. The cohort was categorized into high-grade urothelial carcinoma (HGUC; n = 190) and low-grade urothelial carcinoma (LGUC; n = 182). Participants were randomly assigned to a training group (n = 259) and a validation group (n = 113) in a 7:3 ratio. Regions of interest (ROIs) were delineated on all enhanced CT images using 3D-Slicer software, and 1,223 texture features encompassing first-order, second-order, high-order, and filtered attributes were extracted. Features with an intraclass correlation coefficient (ICC) above 0.75 were retained for further analysis via least absolute shrinkage and selection operator (LASSO) regression. A logistic regression model was constructed based on the selected features to develop a clinical prediction tool. The model's performance was evaluated using the concordance index (C-index), calibration curve, receiver operating characteristic (ROC) curve, and decision curve analysis (DCA).
Eleven radiomics features demonstrated significant associations with the pathological grade of bladder urothelial carcinoma. Among the models evaluated, the logistic regression model exhibited the highest discriminative power, with an area under the curve (AUC) of 0.858. Multivariate analysis identified age and proteinuria as independent predictors. The integrated model, incorporating both clinical and imaging features, outperformed models based on clinical or radiomic data alone (AUC = 0.864).
This study presents the first CT-based nomogram that integrates multiparametric radiomic features with comprehensive clinical indicators to preoperatively predict pathological grade in bladder urothelial carcinoma. The model offers a robust, accurate, and non-invasive tool that can facilitate individualized treatment planning and enhance clinical decision-making.
本研究旨在通过整合多参数薄层增强计算机断层扫描（CT）纹理特征与相关临床指标，建立一种用于膀胱尿路上皮癌病理分级的术前预测模型。
回顾性收集了2015年1月至2020年10月期间在本机构诊断为膀胱尿路上皮癌的372例患者的CT图像和临床数据。该队列被分为高级别尿路上皮癌（HGUC；n = 190）和低级别尿路上皮癌（LGUC；n = 182）。参与者以7:3的比例随机分配到训练组（n = 259）和验证组（n = 113）。使用3D-Slicer软件在所有增强CT图像上勾勒出感兴趣区域（ROI），并提取了包括一阶、二阶、高阶和滤波属性在内的1223个纹理特征。类内相关系数（ICC）高于0.75的特征通过最小绝对收缩和选择算子（LASSO）回归保留用于进一步分析。基于所选特征构建逻辑回归模型以开发临床预测工具。使用一致性指数（C-index）、校准曲线、受试者工作特征（ROC）曲线和决策曲线分析（DCA）评估模型的性能。
11个放射组学特征与膀胱尿路上皮癌的病理分级显示出显著相关性。在评估的模型中，逻辑回归模型表现出最高的判别能力，曲线下面积（AUC）为0.858。多变量分析确定年龄和蛋白尿为独立预测因素。结合临床和影像特征构建的综合模型优于仅基于临床或放射组学数据的模型（AUC = 0.864）。
本研究提出了首个基于CT的列线图，该列线图整合了多参数放射组学特征与全面的临床指标，用于术前预测膀胱尿路上皮癌的病理分级。该模型提供了一种强大、准确且无创的工具，可促进个体化治疗计划并加强临床决策。
