| 李靖,高剑波,王睿,方梦捷,许春苗,黎海亮.基于增强CT影像组学术前预测胃腺癌淋巴结转移[J].中国医学影像技术,2022,38(6):878~883 |
| 基于增强CT影像组学术前预测胃腺癌淋巴结转移 |
| Radiomics based on enhanced CT for preoperative predicting lymph node metastasis of gastric adenocarcinoma |
| 投稿时间：2021-07-23 修订日期：2021-10-19 |
| DOI：10.13929/j.issn.1003-3289.2022.06.020 |
| 中文关键词: 胃肿瘤 淋巴结转移 体层摄影术,X线计算机 影像组学 |
| 英文关键词:stomach neoplasms lymphatic metastasis tomography, X-ray computed radiomics |
| 基金项目:河南省科技厅重点科技攻关项目（202102310736）、河南省中青年卫生健康科技创新优秀青年人才培养项目（YXKC2021054）。 |
|
| 摘要点击次数: 881 |
| 全文下载次数: 322 |
| 中文摘要: |
| 目的 观察增强CT影像组学模型术前预测胃腺癌淋巴结转移（LNM）的价值。方法 回顾性分析193例经术后病理证实的单发胃腺癌的腹部双期增强CT资料，将其分为训练集（n=97）和验证集（n=96），比较LNM （+）与LNM （－）肿瘤CT表现的差异。分别于增强动脉期和静脉期CT提取病灶影像组学特征，构建相应影像组学标签；将单因素分析有统计学意义的CT参数及其影像组学标签纳入多因素logistic回归分析，筛选胃腺癌LNM的独立预测因素，分别建立临床模型及影像组学列线图。采用受试者工作特征（ROC）曲线评估各模型预测胃腺癌LNM的效能，计算曲线下面积（AUC），比较其差异。结果 训练集含54例LNM （+）和43例LNM （－），验证集含58例LNM （+）和38例LNM （－）。LNM （+）患者肿瘤厚度和阳性淋巴结占比均高于LNM （－）者（P均<0.05）。肿瘤厚度及淋巴结状态均为LNM的独立预测因素（P均<0.01）并用于构建临床模型。淋巴结状态和静脉期影像组学标签是胃腺癌LNM的独立预测因素（P均<0.01），以之构建的影像组学列线图在训练集和验证集中的AUC分别为0.810和0.778，与临床模型AUC差异均无统计学意义（0.772、0.762，Z=1.11、0.27，P=0.27、0.78）。结论 基于增强CT影像组学模型术前预测胃腺癌LNM效能较佳。 |
| 英文摘要: |
| Objective To explore the value of radiomics model based on enhanced CT for preoperative predicting lymph node metastasis (LNM) of gastric adenocarcinoma. Methods Dual-phase enhanced CT data of 193 patients with single gastric adenocarcinoma lesion confirmed by surgical pathology were retrospectively analyzed. The patients were divided into training set (n=97) and testing set (n=96). CT manifestations of with LNM (-) and LNM (+) tumors in training set and testing set were compared. The radiomics features of lesions were abstracted based on dual-phase enhanced CT with machine learning method to construct radiomics signatures. Logistic regression analysis was performed on CT parameters being statistical different in univariate analysis and relative radiomics signatures to screen independent predictor for LNM, respectively, and the clinical and radiomics models were developed. Receiver operating characteristic (ROC) curve was used to evaluate the efficacy of each model for predicting LNM of gastric adenocarcinoma, and the areas under the curve (AUC) were calculated and compared. Results There were 54 LNM (+) and 43 LNM (-) patients in training set, 58 LNM (+) and 38 LNM (-) patients in testing set. Tumor thickness and the proportion of positive lymph node in LNM (+) patients were all higher than those in LNM (-) ones (all P<0.05), and both were independent predictors for LNM (both P<0.01) and used to construct clinical model. Lymph node status and radiomics signature of venous phase were independent predictors for LNM gastric adenocarcinoma (both P<0.01) and were used to construct radiomics nomogram. AUC of radiomics nomogram in training set and testing set were 0.810 and 0.778, respectively, not significantly different from that of clinical model (0.772, 0.762, Z=1.11, 0.27, P=0.27, 0.78). Conclusion Radiomics model based on enhanced CT had good efficacy for preoperative predicting LNM of gastric adenocarcinoma. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
