PET-CT Imaging-based Preoperative Diagnosis Of Lymph Node Metastases In Non-small Cell Lung Cancer

Objective:The primary tumor and lymph node were analyzed using the radiomics model in order to confirm the effectiveness of preoperative identification of lymph node metastasis in non-small cell lung cancer.Methods:From January 2017 to December 2018,115 patients at our hospital who had non-small cell lung cancer that had been verified by pathology were retrospectively examined.At a 7:3 ratio,patients were randomly split into the cross validation set(83cases,including positive lymph node group: 37 primary lesions,73 lymph nodes;negative lymph node group: 46 primary lesions,94 lymph nodes)and the test set(32cases,including positive lymph node group: 17 primary lesions,34 lymph nodes;negative lymph node group: 15 primary sites,33 lymph nodes).We gathered the patients’ clinical information and preoperative PET-CT images.Intelli Space Discovery software was utilized to identify the primary tumor and lymph nodes based on pathology,and RIAS software was used to extract the image radiomics features of the Region of interest(ROI).Based on the reduced dimensional radiomics,lymph node models and primary tumor-lymph node combination models were developed.The diagnostic effectiveness of the model was then assessed using the Precision-Recall(PR)curve 、 Receiver Operating Characteristic(ROC)curve 、Area Under the Curve(AUC)and Average Precision(AP).Results:The patients’ gender and age did not statistically affect the diagnosis(P>0.05).A total of 234 lymph nodes(107 positive and 127 negative)were included.1427 radiomics features were extracted from each VOI,and the dimensions were reduced for the construction of the radiomics model.In the test set,the combined model’s AUC(95%CI)was 0.914(0.844-0.983),while the lymph node model’s AUC(95%CI)was 0.840(0.737-0.943).A significant difference between the two was revealed by the De Long test(P=0.0225).As a result,the combined model had higher accuracy in identifying preoperative lymph node metastases in non-small cell lung cancer patients.Conclusion:In this study,the model established by using lymph node imaging features has a good predictive effect on preoperative diagnosis of lymph node metastasis in patients with non-small cell lung cancer,and the introduction of primary tumor-lymph node radiomics model for primary tumors can significantly improve the diagnostic efficiency.Therefore,we can establish the primary tumor-lymph node radiomics model through PET-CT to help the individual preoperative diagnosis of lymph node metastasis in patients with non-small cell lung cancer.