多模态超声预测模型对肾占位性病变的 诊断价值研究

A Study on the Diagnostic Value of Multimodal Ultrasound Predictive Models for Renal Space-Occupying Lesions

摘要: 目的：比较常规超声、彩色多普勒超声与超声造影鉴别肾肿瘤良恶性的价值，并分别构建基于单一技术及联合技术的预测模型，评估其诊断效能。方法：回顾性收集174例经病理证实的肾占位性病变，包含恶性病变106例，良性病变68例。记录患者的二维超声、彩色多普勒及超声造影特征并进行标准化描述，通过单因素logistic回归筛选有统计学意义的特征纳入多因素logistic回归模型，绘制受试者工作特征曲线(ROC)比较不同预测模型的诊断效能。结果：整合常规超声、彩色多普勒和超声造影的多模态模型，是鉴别肾占位性病变良恶性的最优超声诊断方案，ROC曲线下面积、灵敏度、特异度分别为0.938、93.4%和83.8%。结论：基于多模态数据构建的Logistic回归风险模型可以较好地预测肾恶性占位病变的风险，为临床医师对肾占位性病变的诊断与决策提供了重要参考。

Abstract: Objective: To compare the value of conventional ultrasound, color Doppler ultrasound, and contrast-enhanced ultrasound in differentiating benign from malignant renal tumors, and to construct predictive models based on single techniques and combined techniques, and evaluate their diagnostic efficacy. Methods: A retrospective analysis was conducted on 174 cases of pathologically confirmed renal space-occupying lesions, including 106 malignant lesions and 68 benign lesions. The two-dimensional ultrasound, color Doppler, and contrast-enhanced ultrasound features were recorded and standardized. Statistically significant features were screened by univariate logistic regression and included in multivariate logistic regression models. Receiver Operating Characteristic (ROC) curves were drawn to compare the diagnostic efficacy of different predictive models. Results: The multimodal model integrating conventional ultrasound, color Doppler, and contrast-enhanced ultrasound is the optimal ultrasound diagnostic approach for differentiating benign and malignant renal space-occupying lesions, with an area under the ROC curve, sensitivity, and specificity of 0.938, 93.4%, and 83.8%, respectively. Conclusion: The logistic regression risk model based on multimodal data can effectively predict the risk of malignant renal space-occupying lesions, providing an important reference for clinicians in the diagnosis and decision-making of renal space-occupying lesions.