MRI影像组学特征列线图在鉴别卵巢囊肿和单房囊腺瘤中的应用价值
杨洁 , 孙惠苗 , 全帅 , 胡磊 , 高凯 , 王思洁
山西医科大学学报 ›› 2025, Vol. 56 ›› Issue (10) : 1109 -1115.
MRI影像组学特征列线图在鉴别卵巢囊肿和单房囊腺瘤中的应用价值
Application of MRI radiomic features⁃based nomogram in identifying ovarian cysts and unilocular cystadenomas
目的 通过建立影像组学的方法将女性盆腔MR-T2WI及增强序列的图像信息模型绘制列线图评价其鉴别女性卵巢囊肿和单房囊腺瘤的诊断效能。 方法 回顾性收集113例因卵巢肿物行盆腔MRI平扫及增强扫描并手术证实为卵巢囊肿或单房囊腺瘤的患者,共121例病灶,其中卵巢囊肿49例,单房囊腺瘤72例。分析各卵巢病变的T2WI及增强序列的磁共振图像,同时收集患者的相关临床资料。基于T2WI及增强图像各提取1 316个、共2 632个影像组学特征,筛选最佳特征,计算影像组学评分并建立影像组学模型。按7∶3将患者随机分为训练组(n=80,共84个病灶)和验证组(n=33,共37个病灶)。应用Logistic回归分析筛选临床相关独立危险因素,建立临床模型,并构建形态学和影像组学评分联合模型,绘制列线图将之可视化。以受试者工作特征(ROC)曲线评估各模型预测囊肿和单房囊腺瘤的效能,以决策曲线分析(DCA)评价联合模型的价值。 结果 训练组与验证组患者年龄差异无统计学意义(P>0.05)。本研究基于T2WI和增强扫描两个序列的影像组学特征进行特征筛选,最终T2WI及增强序列各筛选出10个特征,并建立单序列影像组学模型;基于T2WI及增强序列的影像特征进行融合得到融合特征集后,再次进行特征筛选,最终剩余6个特征。基于筛选到的6个关键特征建立T2WI及增强序列联合影像组学模型在训练组患者中预测卵巢单房囊腺瘤的曲线下面积(AUC)为0.946,验证组的AUC为0.927。临床特征中病变前后径与上下径是鉴别囊肿与单房囊腺瘤的MRI影像学独立因素(P<0.05);与影像组学构建的联合nomogram模型训练集AUC为0.955,测试集AUC为0.942,高于影像组学模型(Z=-3.451,P<0.001)。联合模型在阈值概率0~1.0时的临床净获益大于影像组学模型。 结论临床特征与MR-T2WI及增强序列影像组学的联合模型对卵巢囊肿和单房囊腺瘤有很高的鉴别诊断能力。
Objective To evaluate the diagnostic efficacy of a radiomics-based nomogram using female pelvic MR-T2WI and contrast-enhanced sequences in distinguishing ovarian cysts and unilocular cystadenoma. Methods A retrospective study was conducted on 113 patients, 121 lesions(49 ovarian cysts and 72 unilocular cystadenomas), who underwent pelvic MRI scans(both non-contrast and contrast-enhanced) for ovarian masses and were surgically confirmed as ovarian cysts or unilocular cystadenomas. The MRI images of T2WI and contrast-enhanced sequences for each ovarian lesion were analyzed, and clinical data of patients were collected. From T2-weighted and contrast-enhanced images, 1 316 radiomic features were extracted per sequence, producing 2 632 features in total, to select the optimal features, calculate the radiomics score, and establish a radiomics model. Patients were randomly divided into training group(n=80, 84 lesions) and validation group(n=33, 37 lesions) at a ratio of 7∶3. Logistic regression was employed to identify independent clinical risk factors and construct a clinical model. Finally, a combined model of morphologic parameters and radiomics score was built and visualized as a nomogram.Receiver operating characteristic(ROC) curves were used to evaluate the discriminatory performance of each model in distinguishing ovarian cysts from unilocular cystadenomas, while decision curve analysis(DCA) was performed to assess the clinical value of the combined model. Results The difference in age was not statistically significant between the training set and the testing set(P>0.05). Based on the imaging characteristics of T2WI and enhanced sequences, 10 features were identified from each sequence and a single-sequence imaging omics model was established. After integrating the imaging features from both T2WI and enhanced sequences, a fusion feature set was created for selection of characteristic features, and finally 6 features were obtained. A combined imaging omics model for T2WI and enhanced sequences was established using these 6 key features. The model showed an area under the curve(AUC) of 0.946 in training group and 0.927 in validation group for predicting unilocular cysta⁃denoma. In the clinical features, the anteroposterior and craniocaudal diameters of the lesion were identified as independent MRI imaging factors for distinguishing cysts from unilocular cystadenomas(P<0.05).The AUC of the combined model based on APD, SID and radiomics was 0.955 in the training set and 0.942 in the testing set, higher than that of the imaging omics model(Z=-3.451, P<0.001). The combined model yielded a greater clinical net benefit than the radiomics-alone model across the entire threshold probability range of 0-1.0. Conclusion The radiomics model based on MR-T2WI and contrast-enhanced sequences is effective in the differential diagnosis of ovarian cysts and unilocular cystadenomas.
| [1] |
|
| [2] |
|
| [3] |
李建军,朱兰,郎景和, |
| [4] |
乔敏霞,时惠平,秦丹, |
| [5] |
谭红娜,武明辉,周晶, |
| [6] |
Society for Maternal⁃Fetal Medicine(SMFM), |
| [7] |
|
| [8] |
陈国亨,李梅莲,何金华, |
| [9] |
胡艳,刘洋,郑伊能, |
| [10] |
许贝贝.MRI增强扫描联合弥散加权成像鉴别诊断良恶性卵巢浆液性囊腺瘤的价值[J].医疗装备,2024,37(9):24-26. |
山西省高等学校科技创新项目(2022L200)
/
| 〈 |
|
〉 |