余鎏,黄玲玲,袁振亚,赵泉,印宏坤,于朋鑫,曹颖.基于CT平扫影像组学模型预测食管癌淋巴结转移[J].中国医学影像技术,2021,37(9):1333~1337
基于CT平扫影像组学模型预测食管癌淋巴结转移
Radiomics model based on plain CT for predicting lymph node metastasis of esophageal cancer
投稿时间:2020-09-11  修订日期:2021-06-13
DOI:10.13929/j.issn.1003-3289.2021.09.014
中文关键词:  食管肿瘤  淋巴结转移  体层摄影术,X线计算机  影像组学
英文关键词:esophageal neoplasms  lymphatic metastasis  tomography, X-ray computed  radiomics
基金项目:
作者单位E-mail
余鎏 重庆市开州区人民医院放射科, 重庆 405400  
黄玲玲 重庆市开州区人民医院放射科, 重庆 405400  
袁振亚 重庆市开州区人民医院放射科, 重庆 405400  
赵泉 重庆市开州区人民医院放射科, 重庆 405400 1179292219@qq.com 
印宏坤 推想医疗科技股份有限公司, 北京 100025  
于朋鑫 推想医疗科技股份有限公司, 北京 100025  
曹颖 推想医疗科技股份有限公司, 北京 100025  
摘要点击次数: 1290
全文下载次数: 390
中文摘要:
      目的 评估术前基于胸部CT平扫影像组学模型预测食管癌患者淋巴结转移的价值。方法 回顾性分析368例经术前内镜活检及术后病理确诊的食管癌患者,其中100例淋巴结转移、268例无淋巴结转移,按比例3 ∶ 1将其分为训练组(包括201例无淋巴结转移和75例淋巴结转移)和验证组(67例无淋巴结转移和25例淋巴结转移)。自胸部CT中提取食管癌病灶的影像组学特征,并以最小绝对收缩和选择算子(LASSO)回归进行降维,筛选与食管癌淋巴结转移关联度高的特征;采用支持向量机构建预测淋巴结转移的影像组学模型,并以受试者工作特征(ROC)曲线分析模型的诊断效能。结果 共提取1 046个组学特征参数,经LASSO降维筛选出11个特征参数用于建立预测淋巴结转移模型。影像组学模型预测训练组淋巴结转移的曲线下面积(AUC)为0.84,敏感度为84.00%,特异度为75.12%,准确率为77.54%;于验证组的AUC为0.82,敏感度为80.00%,特异度为77.61%,准确率为78.26%。结论 术前基于胸部CT平扫影像组学模型预测食管癌患者淋巴结转移具有较高价值。
英文摘要:
      Objective To explore the value of radiomics model based on chest plain CT in preoperative predicting lymph node metastasis of esophageal cancer. Methods Data of 368 patients with esophageal cancer diagnosed with preoperative endoscopic biopsy and postoperative pathology were retrospectively analyzed including 100 cases with lymph node metastasis and 268 cases without lymph node metastasis. The patients were classified as lymph node metastasis and non-lymph node metastasis and were divided into training group (201 cases without lymph node metastasis and 75 cases with lymph node metastasis) and validation group (67 cases without lymph node metastasis and 25 cases with lymph node metastasis) at the ratio of 3 ∶ 1. Then radiomics features of esophageal cancer on chest CT were extracted, the least absolute shrinkage and selection operator (LASSO) regression was used for dimensionality reduction, and the features with high correlation with lymph node metastasis of esophageal cancer were selected. Support vector machine classifier was used to construct a radiomics model for predicting lymph node metastasis. Then the diagnostic efficacy of the radiomics model was estimated with the receiver operating characteristic (ROC) curve. Results A total of 1 046 radiomics characteristic parameters were extracted. After dimension reduction with LASSO, 11 characteristic parameters were selected to establish model of predicting lymph node metastasis. In training group, the area under the curve (AUC) of the model for predicting lymph node metastasis was 0.84, the diagnostic sensitivity was 84.00%, the specificity was 75.12% and the accuracy was 77.54%. In validation group, AUC of the radiomics model was 0.82, with sensitivity of 80.00%, specificity of 77.61%. Conclusion Radiomics model based on plain CT had good diagnostic value for pre-operative predicting lymph node metastasis of esophageal cancer.
查看全文  查看/发表评论  下载PDF阅读器