基于灰阶超声影像组学在鉴别膀胱尿路上皮癌病理分级中的应用价值

目的 通过灰阶超声影像组学特征鉴别膀胱尿路上皮癌病理分级。 方法 回顾性分析2016年4月~2023年5月山西白求恩医院153例经病理证实的膀胱尿路上皮癌患者。灰阶超声图像手工勾画肿瘤感兴趣区并提取组学特征，LASSO特征降维后采用3种机器学习方法建模并选出最优影像组学模型。采用ROC曲线对模型性能评估，采用Hosmer-Lemeshow适合度检验评价模型的拟合度，并绘制校正曲线，采用决策曲线分析进一步探讨模型的临床应用价值。 结果 3种机器学习模型中的支持向量机算法模型性能表现最优，此模型在训练集和测试集的曲线下面积分别为0.858(95% CI：0.787~0.928)和0.832(95% CI：0.708~0.936)，校准曲线显示出良好的一致性。决策曲线分析结果显示具有较高的净收益。 结论 基于灰阶超声影像组学在鉴别膀胱尿路上皮癌病理分级具有术前诊断价值, 有助于临床精准诊疗。

Intermediate value of grayscale ultrasound image-based radiomics in discriminating the pathological grade of bladder urothelial carcinoma

Objective To discriminate the pathological grade of bladder urothelial carcinoma through grayscale ultrasound image- based radiomics analysis. Methods A retrospective analysis was conducted on 153 patients with bladder urothelial carcinoma confirmed by pathology in our hospital from April 2016 to May 2023. The grayscale ultrasound images were manually delineated to outline the tumor region of interest and extract radiomics features. LASSO feature selection was utilized for dimensionality reduction, followed by modeling with three machine learning methods to identify the best model. The performance of the models was evaluated using ROC curves, and the goodness-of-fit was assessed using the Hosmer-Lemeshow test and calibration curve. Furthermore, decision curve analysis was conducted to explore the clinical utility of the model. Results Among the three machine learning models, the Support vector machine model exhibited the best performance, with an AUC of 0.858 (95% CI: 0.787-0.928) on the training set and 0.832 (95% CI: 0.708-0.936) on the test set. The calibration curve demonstrated good consistency. The decision curve analysis also showed a high net benefit. Conclusion Grayscale ultrasound image- based radiomics has preoperative diagnostic value in distinguishing the pathological grading of bladder urothelial carcinoma, which contributes to precise clinical diagnosis and treatment.