张怡梦,冯吉雪,张晓杰,刘昊月,刘梦珂,李兴鹏,张明霞,王仁贵.MR淋巴管成像影像组学评估中央传导淋巴管畸形[J].中国医学影像技术,2024,40(11):1677~1681
MR淋巴管成像影像组学评估中央传导淋巴管畸形
MR lymphangiography radiomics for evaluating central conducting lymphatic anomaly
投稿时间:2024-03-13  修订日期:2024-06-11
DOI:10.13929/j.issn.1003-3289.2024.11.010
中文关键词:  胸导管  畸形  磁共振成像  影像组学
英文关键词:thoracic duct  abnormalities  magnetic resonance imaging  radiomics
基金项目:国家自然科学基金(61876216)。
作者单位E-mail
张怡梦 首都医科大学附属北京世纪坛医院放射科, 北京 100038  
冯吉雪 首都医科大学附属北京世纪坛医院放射科, 北京 100038  
张晓杰 首都医科大学附属北京世纪坛医院放射科, 北京 100038  
刘昊月 首都医科大学附属北京世纪坛医院放射科, 北京 100038  
刘梦珂 首都医科大学附属北京世纪坛医院放射科, 北京 100038  
李兴鹏 首都医科大学附属北京世纪坛医院放射科, 北京 100038  
张明霞 首都医科大学附属北京世纪坛医院放射科, 北京 100038  
王仁贵 首都医科大学附属北京世纪坛医院放射科, 北京 100038 wangrg@bjsjth.cn 
摘要点击次数: 405
全文下载次数: 111
中文摘要:
      目的 观察MR淋巴管成像(MRL)影像组学用于评估中央传导淋巴管畸形(CCLA)的价值。方法 回顾性分析41例CCLA (A组)、20例非CCLA (包括泛发性淋巴管异常及Gorham-Stout综合征相关淋巴管畸形)复杂淋巴管畸形患者及18名胸导管正常者(B组,n=38),于3D重T2W快速自旋回波序列颈胸部(必要时联合腹部) MRL中沿胸导管全程勾画ROI并提取其影像组学特征;利用5折交叉验证将A、B组数据集划分为K个子集,持续以(K-1)子集的并集为训练集、其余子集为验证集;基于支持向量机(SVM)算法构建模型;绘制受试者工作特征曲线,计算曲线下面积(AUC),分析SVM模型评估CCLA的效能。结果 A组胸导管主干分叉型、胸导管主干囊样变、胸导管主干扩张及胸导管末端扩张占比均高于,而胸导管主干及胸导管末端正常者占比均低于B组(P均<0.05);组间胸导管多干、胸导管主干右位型、胸导管主干部分未显示、胸导管末端多支、胸导管末端囊样变及胸导管末端蔓状占比差异无统计学意义(P均>0.05)。SVM模型评估训练集CCLA的敏感度、特异度、准确率、阳性预测值、阴性预测值及AUC分别为78.95%、97.56%、88.61%、96.77%、83.33%及0.920,在验证集分别为78.95%、83.57%、82.28%、83.33%、81.40%及0.833。结论 MRL影像组学可有效评估CCLA。
英文摘要:
      Objective To observe the value of MR lymphangiography (MRL) radiomics for evaluating central conducting lymphatic anomaly (CCLA). Methods Sixty-one patients with complex lymphatic anomaly, including 41 CCLA (group A), 20 non-CCLA (generalized lymphatic anomaly and Gorham-Stout disease) and 20 subjects with normal thoracic duct (group B, n=38) were retrospectively enrolled. Cervical and thoracic (combined with abdominal if necessary) MRL was acquired using three-dimensional heavily T2W fast spin echo sequence. ROI was delineated along overall thoracic duct, and radiomics features were extracted. Data sets of group A and B were divided into K subsets using 5-fold cross-validation. The union of (K-1) subset was always used as training set, while the other subsets were used as validation set. Radiomics model was constructed based on support vector machine (SVM) algorithm. Receiver operating characteristic curve was drawn, the area under the curve (AUC) was calculated to evaluate the efficacy of SVM model for assessing CCLA. Results The proportions of bifurcation, cystoid change and extension of main thoracic duct, and extension of terminal thoracic duct in group A were all higher, while of normal main and terminal thoracic duct in group A were both lower than those in group B (all P<0.05). No significant difference of proportions of multiple thoracic ducts, dextral thoracic duct, part of thoracic duct invisible, multiple terminal thoracic ducts, cystoid change of terminal thoracic duct nor terminal thoracic duct pampiniform was found between groups (all P>0.05). The sensitivity, specificity, accuracy, positive predictive, negative predictive and AUC of SVM model for evaluating CCLA in training set was 78.95%, 97.56%, 88.61%, 96.77%, 83.33% and 0.920, respectively, which in validation set was 78.95%, 83.57%, 82.28%, 83.33%, 81.40% and 0.833, respectively. Conclusion MRL radiomics could be used to effectively evaluate CCLA.
查看全文  查看/发表评论  下载PDF阅读器