基于MR影像学指标鉴别卵巢良恶性肿瘤列线图的构建和验证

Construction and Verification of a Nomogram for Differentiating Benign and Malignant Ovarian Tumors Based on MR Imaging Indicators

摘要: 目的：探讨动态增强磁共振(DCE-MRI)动态增强时间–信号强度曲线(TIC)类型、半定量参数增强扫描后60 s强化率(SI60%)、增强扫描后200 s内达峰时间(TTP200s)及临床指标(最大径、CA125等)在鉴别卵巢良、恶性肿瘤中的价值。方法：收集2023年7月至2025年7月济宁医学院附属医院确诊的60例卵巢病变患者的临床及影像学资料(恶性25例，良性35例)，按2:1随机划分为训练集(40例)和测试集(20例)。采用单因素及多因素Logistic回归分析病变良恶性与临床及影像指标的关系，并构建预测列线图模型评估诊断效能。结果：单因素分析显示CA125 (P = 0.004)、SI60% (P = 0.002)及TTP200s (P = 0.005)与病变良恶性显著相关。多因素分析进一步确认CA125 (OR = 1.12, P = 0.018)、SI60% (OR = 1.16, P = 0.036)及TTP200s (OR = 2.20, P = 0.031)为独立预测因素。基于这些变量构建的列线图模型在训练集中的AUC值为0.92，优于单一因素模型，校准曲线及决策曲线分析(DCA)均显示模型具有良好的预测性能和临床应用价值。结论：基于CA125、SI60%及TTP200s的多参数列线图模型可有效鉴别卵巢良恶性肿瘤，对术前风险评估具有一定价值。

Abstract: Objective: To investigate the value of dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) time-intensity curve (TIC) types, semi-quantitative parameters [including the enhancement rate at 60 seconds post-contrast (SI60%) and time to peak within 200 seconds (TTP200s)], and clinical indicators (such as maximum diameter and CA125) in differentiating between benign and malignant ovarian tumors. Methods: Clinical and imaging data from 60 patients with ovarian lesions (25 malignant, 35 benign) confirmed at Jining Medical University Affiliated Hospital between July 2023 and July 2025 were collected. The patients were randomly divided into a training set (40 cases) and a test set (20 cases) in a 2:1 ratio. Univariate and multivariate logistic regression analyses were used to examine the relationship between lesion malignancy and the clinical/imaging indicators. A predictive nomogram model was constructed and its diagnostic performance was evaluated. Results: Univariate analysis showed that CA125 (P = 0.004), SI60% (P = 0.002), and TTP200s (P = 0.005) were significantly associated with lesion malignancy. Multivariate analysis further confirmed CA125 (OR = 1.12, P = 0.018), SI60% (OR = 1.16, P = 0.036), and TTP200s (OR = 2.20, P = 0.031) as independent predictive factors. The nomogram model constructed based on these variables achieved an AUC of 0.92 in the training set, outperforming single-factor models. Both the calibration curve and decision curve analysis (DCA) demonstrated that the model has good predictive performance and clinical application value. Conclusion: The multiparameter nomogram model based on CA125, SI60%, and TTP200s can effectively differentiate between benign and malignant ovarian tumors and holds value for preoperative risk assessment.