基于18F-FDG PET/CT影像组学的非小细胞肺癌病理亚型分类

目的：旨在建立一种基于¹⁸F-FDG PET/CT的临床—影像组学相结合的综合模型用于区分非小细胞肺癌中的腺癌和鳞癌。方法：回顾性收集上海交通大学附属胸科医院120例经病理学验证为腺癌(65例)和鳞癌(55例)的患者，从预处理的CT图像和PET图像中分别提取1218、108个影像组学特征，并纳入10个临床特征因素；卡方检验和Wilcoxon检验用于对临床特征进行筛选，并使用Relief算法和最小绝对收缩和选择算子(LASSO)对影像组学特征进行筛选；通过6种机器学习分类器分别建立临床、影像组学、综合模型。通过受试者工作特征(ROC)曲线及曲线下面积(AUC)来评价模型的分类能力。结果：综合模型在训练集和测试集中均表现出最高的AUC值和准确率，其中随机森林(RF)和Bagging分类器表现出的分类效果最佳。经五折交叉验证后，训练集中RF和Bagging的AUC值和准确率分别为0.92±0.03、0.86±0.06和0.92±0.02、0.83±0.02；测试集中RF和Bagging的AUC值和准确率分别为0.92、0.81和0.91、0.86。结论：结合^1...

Classification of pathological subtypes of non-small-cell lung cancer based on 18F-FDG PET CT radiomics

Abstract: Objective To establish a comprehensive clinical-radiomics model based on 18F-FDG PET/CT for differentiating adenocarcinoma and squamous cell carcinoma in non-small-cell lung cancer. Methods A total of 120 cases of pathologically verified adenocarcinoma (n=65) and squamous cell carcinoma (n=55) from Shanghai Chest Hospital were collected retrospectively. In addition to 1 218 and 108 radiomics signatures extracted from the preprocessed CT images and PET images, 10 clinical features were included. Chi-square test and Wilcoxon test were used to screen clinical features, and radiomic signatures were screened using Relief algorithm and least absolute shrinkage and selection operator. Six machine learning classifiers were used to build clinical, radiomics, and comprehensive models. The classification ability of the model was evaluated using receiver operating characteristic (ROC) curve and area under curve (AUC). Results The comprehensive model exhibited the highest AUC and accuracy in both training and test sets, with random forest (RF) and Bagging classifiers showing the best classification results. After 5-fold cross-validation, the AUC and accuracy of RF in the training set were 0.92±0.03, 0.86±0.06, while those of Bagging were 0.92±0.02, 0.83±0.02. In the test set, RF and Bagging also had the optimal classification performances (RF: AUC=0.92, accuracy=0.81 Bagging: AUC=0.91, accuracy=0.86). Conclusion The classification prediction model combining 18F-FDG PET/CT clinical features and radiomics signatures can be well used to distinguish adenocarcinoma and squamous cell carcinoma.