Lei Yan, Tang Hanzhou, Liu Lianlian, Zheng Tingting, Zhang Yuan, Chen Tong, Shen Junkang, Song Bin
Department of Radiology, Minhang Hospital, Fudan University, 170 Xinsong Road, Shanghai, 201199, Shanghai, People's Republic of China.
Department of Radiology, The Second Affiliated Hospital of Soochow University, Soochow University, 1055 Sanxiang Road, Suzhou, 215004, Jiangsu, People's Republic of China.
Int J Med Sci. 2025 Feb 26;22(6):1425-1436. doi: 10.7150/ijms.107452. eCollection 2025.
To investigate the diagnostic performance of texture analysis using multi-parameter MRI in distinguishing between benign and malignant lesions with ovarian-adnexal magnetic resonance imaging report and data system (O-RADS MRI) score 4. A retrospective analysis was conducted of 57 lesions with an O-RADS MRI score of 4, of which 26 were benign and 31 were malignant. Based on the T2WI, ADC, and CE_T1WI, the textural features of the entire lesion were extracted. The minimum redundancy maximum relevance (mRMR) method was used to select features, and the random forest (RF) algorithm was used to construct four prediction models: T2WI, ADC, CE_T1WI, and the combined models. Ten-fold cross-validation was used to verify the model prediction performance, and receiver operating characteristic (ROC) analysis was used to evaluate the model performance, including area under the curve (AUC), accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). 3474 texture features were extracted from the ADC, T2WI, and CE_T1WI images. ADC, T2WI, CE_T1WI, and combined models were constructed. Each model contained ten texture features. The AUC of the ADC, T2WI, CE_T1WI, and combined models were 0.749 (95% CI: 0.621-0.876), 0.671 (95% CI: 0.524-0.818), 0.786 (95% CI: 0.662-0.909), and 0.860 (95% CI: 0.76-0.959), respectively. The AUC of the combined model was significantly higher than those of the other three groups. The accuracy, sensitivity, specificity, PPV, and NPV of the combined model in distinguishing benign and malignant lesions with an O-RADS MRI score of 4 were 75.9%, 77.8%, 74.1%, 72.4%, and 79.3%, respectively. Texture analysis of multi-parameter MRI can improve the diagnostic efficiency of distinguishing benign and malignant lesions with an O-RADS MRI score of 4 and provide some help in clinical decision-making.
利用多参数磁共振成像（MRI）进行纹理分析，以鉴别卵巢附件磁共振成像报告和数据系统（O-RADS MRI）评分为4的良性和恶性病变的诊断性能。对57例O-RADS MRI评分为4的病变进行回顾性分析，其中26例为良性，31例为恶性。基于T2WI、ADC和CE_T1WI，提取整个病变的纹理特征。采用最小冗余最大相关性（mRMR）方法选择特征，并使用随机森林（RF）算法构建四个预测模型：T2WI、ADC、CE_T1WI和联合模型。采用十折交叉验证来验证模型的预测性能，并使用受试者操作特征（ROC）分析来评估模型性能，包括曲线下面积（AUC）、准确性、敏感性、特异性、阳性预测值（PPV）和阴性预测值（NPV）。从ADC、T2WI和CE_T1WI图像中提取了3474个纹理特征。构建了ADC、T2WI、CE_T1WI和联合模型。每个模型包含十个纹理特征。ADC、T2WI、CE_T1WI和联合模型的AUC分别为0.749（95%CI：0.621-0.876）、0.671（95%CI：0.524-0.818）、0.786（95%CI：0.662-0.909）和0.860（95%CI：0.76-0.959）。联合模型的AUC显著高于其他三组。联合模型在鉴别O-RADS MRI评分为4的良性和恶性病变时的准确性、敏感性、特异性、PPV和NPV分别为75.9%、77.8%、74.1%、72.4%和79.3%。多参数MRI的纹理分析可以提高鉴别O-RADS MRI评分为4的良性和恶性病变的诊断效率，并为临床决策提供一些帮助。
