摘要
目的:探讨基于多参数MRI影像组学建立模型在鉴别子宫内膜癌与子宫黏膜下肌瘤中的价值。方法:回顾性收集本院2013年8月-2022年1月经术后病理证实为子宫内膜癌和子宫黏膜下肌瘤病例93例,其中子宫内膜癌51例,子宫黏膜下肌瘤42例。应用Python软件Pyradiomics包将T_(2)WI、ADC、对比增强T_(1)WI(CE-T_(1)WI)图像勾画的ROI提取影像组学特征,按照7:3的比例随机分为训练集(n=65)和测试集(n=28),应用最小绝对收缩和选择算子(LASSO)回归对训练集数据进行特征降维,筛选最佳影像组学特征,代入Logistic回归机器学习方法构建模型,通过对受试者工作特征曲线、准确率、敏感度、特异度进行分析,探讨T_(2)WI、ADC、CE-T_(1)WI及T_(2)WI、ADC联合CE-T_(1)WI四组影像组学模型的诊断效能。结果:经过降维和筛选,分别在T_(2)WI、ADC、CE-T_(1)WI组中提取24、27、26个影像学特征。构建的鉴别诊断子宫内膜癌和子宫黏膜下肌瘤模型中,T_(2)WI、ADC、CE-T_(1)WI及T_(2)WI、ADC联合CE-T_(1)WI组的训练集和测试集ROC曲线下面积分别为0.97、0.88、0.93、0.99和0.90、0.67、0.85、0.94。结论:基于T_(2)WI、ADC和CE-T_(1)WI建立影像组学模型对鉴别子宫内膜癌和子宫黏膜下肌瘤有一定价值。
Objective:To investigate the value of establishing radiomics model based on multi-parameter MRI imaging in the differential diagnosis between endometrial cancer and submucous myoma.Methods:A total of 93 patients with endometrial cancer and submucous myoma confirmed by postoperative pathology from August 2013 to January 2022 in our hospital were retrospectively collected,including 51 patients with endometrial cancer and 42 patients with submucous myoma.The ROI delineated by T_(2)WI,ADC,and contrast-enhanced T_(1)WI(CE-T_(1)WI)images were extracted for imaging characteristics using the Python software Pyradiomics package and randomly divided into the training set(n=65)and the testing set(n=28)according to a ratio of 7:3.The least absolute shrinkage and selection operator regression(LASSO)regression was used to reduce the characteristics of the training set data,select the best imaging characteristics,and substitute them into logistic regression machine learning method to construct the model.The diagnostic efficacy of the four groups of imaging models,T_(2)WI,ADC,CE-T_(1)WI,and T_(2)WI and ADC combined with CE-T_(1)WI,was investigated by analyzing the receiver operating characteristic curve,accuracy,sensitivity,and specificity.In addition,the Delong Test was used to compare the models.Results:After dimension reduction and screening,24,27 and 26 imaging features were extracted from T_(2)WI,ADC and CE-T_(1)WI groups,respectively.In the constructed model of differential diagnosis for endometrial carcinoma and submucosal myoma,the areas under the ROC curves of the training and test sets for the T_(2)WI,ADC,CE-T_(1)WI,and T_(2)WI and ADC combined with CE-T_(1)WI groups were 0.97,0.88,0.93,0.99,0.90,0.67,0.85,and 0.94,respectively.Conclusion:The establishment of radiomics model based on T_(2)WI,ADC and CE-T_(1)WI has a certain value in the differential diagnosis of endometrial carcinoma from submucous myoma.
作者
严文杰
许传芳
杜灏蓝
林晓南
兰鹏
钟群
YAN Wen-jie;XU Chuan-fang;DU Hao-lan(Department of Radiology,the Affiliated People's Hospital of Fujian University of Traditional Chinese Medicine,350004 Fuzhou,China)
出处
《放射学实践》
CSCD
北大核心
2023年第7期931-936,共6页
Radiologic Practice
