摘要
目的 探讨平扫CT影像组学预测肺浸润性腺癌脉管癌栓的价值。资料与方法 回顾性分析2018年1月—2021年4月潍坊市人民医院经病理证实的肺浸润性腺癌195例,按照7∶3分为训练集136例和验证集59例。根据术前平扫CT图像,使用3D-slicer手动勾画感兴趣区并提取影像组学特征850个,然后使用最小绝对收缩和选择算法回归模型进行特征降维,构建预测模型并评价模型效能。结果 使用最终保留的17项特征参数建立组学评分模型,训练集和验证集的曲线下面积分别为0.810、0.809,训练集的特异度和敏感度分别为76.0%和76.9%;验证集的特异度和敏感度分别为75.3%和75.0%,两模型差异无统计学意义(P=0.678)。结论 基于平扫CT影像组学特征建立的模型对肺浸润性腺癌脉管癌栓状态具有较好的预测价值,影像组学特征可以用作潜在的生物标志物。
Purpose To explore the value of CT radiomics model in predicting the status of lymphovascular tumor embolus in patients with lung invasive adenocarcinomas.Materials and Methods A total of 195 patients with lung invasive adenocarcinoma confirmed pathologically from January 2018 to April 2021 in Weifang People's Hospital were retrospectively analyzed,including 136 cases in the training set and 59 cases in the validation set according to 7∶3.According to the preoperative CT images,the 3D-slicer was used to manually delineate the area of interest.A total of 850 quantization radiomics features were extracted,and then least absolute shrinkage and selection operator(LASSO)was used to perform feature dimensionality reduction to build a predictive model,and the effectiveness of the model was evaluated.Results The Radscore was established via the 17 feature parameters obtained after dimensionality reduction of the LASSO regression model.The area under curve of the training group and the validation group were 0.810 and 0.809 respectively,the specificity and sensitivity of the training group and the validation group were 76.0%,76.9%;75.3%and 75.0%respectively,there was no statistical difference between the two models in the above indicators(P=0.678).Conclusion The model established based on the features of CT radiomics that may be a potential biomarker,which has a good predictive value for the status of lymphovascular tumor embolus in patients with lung invasive adenocarcinoma.
作者
刘俊忠
王琦
褚玉静
寇介丽
康立清
LIU Junzhong;WANG Qi;CHU Yujing;KOU Jieli;KANG Liqing(Department of Radiology,Weifang People's Hospital,the First Affiliated Hospital of Weifang Medical University,Weifang 261041,China;不详)
出处
《中国医学影像学杂志》
CSCD
北大核心
2023年第5期487-491,共5页
Chinese Journal of Medical Imaging
基金
潍坊市科技发展计划项目(2020YX073)。