非小细胞肺癌淋巴结转移预测模型研究

在非小细胞肺癌的治疗过程中，淋巴结转移状态是决定治疗方案的重要因素。为了辅助临床医生制定更精确的治疗方案，开发并验证了一种基于CT影像组学非小细胞肺癌淋巴结转移的预测模型。从TCIA数据库的NSCLCRadiogenomics公共数据集中选取了134例符合试验要求的患者数据，然后从每例患者的CT影像数据中提取了1 648个特征，并用特征优化方法进行特征降维和选择，然后用朴素贝叶斯、线性判别分析、支持向量机和高斯过程5种机器学习方法建立预测模型，最后使用上海市胸科医院收集的44例患者数据进行外部验证。其中，最优淋巴结转移预测模型在训练集和测试集上准确率分别为0.802和0.795,AUC值分别为0.852和0.810。试验结果表明，所提出的预测模型分类性能良好，可以辅助医生更准确地评估淋巴结转移状态，从而制定出更精准的个性化治疗方案。

Prediction model for lymph node metastasis in non-small cell lung cancer

Abstract: In the treatment of non-small cell lung cancer, the state of lymph node metastasis is an important factor when deciding the therapeutic schedule. In order to assist clinicians in formulating a more accurate therapeutic schedule, a CT radiomics-based prediction model for lymph node metastasis in non-small cell lung cancer is developed and validated. The patient data of 134 cases meeting the requirement of experiment are selected from the NSCLC-Radiogenomics public data set in TCIA database, and then 1 648 features are extracted from the CT image data of each patient. After feature reduction and selection by feature optimization method, prediction models are established by different machine learning methods, namely na?e Bayes, linear discriminant analysis, support vector machine and Gaussian process. Finally, patient data collected from 44 cases in Shanghai Chest Hospital are used for external validation. Among the established models, the optimal model for predicting lymph node metastasis has an accuracy of 0.802 and 0.795 in training set and test set, respectively, and the AUC is 0.852 and 0.810, respectively. The experimental results show that the proposed prediction model has good classification performances, and can assist doctors in assessing the state of lymph node metastasis more accurately, so as to develop a more precise and personalized therapeutic schedule.