| Objective: The latest national cancer statistics released by the National Cancer Center in2022 show that the morbidity and mortality of lung cancer rank first in China.Non-small cell lung cancer(NSCLC)is the most common pathological type,and most patients have been diagnosed as advanced and lost the opportunity of surgery.In recent years,immune checkpoint inhibitors(ICIs)are increasingly used in the clinical treatment of advanced NSCLC,significantly improving the survival of patients.However,not all patients can benefit from immunotherapy,so it is particularly important to find suitable predictive indicators to screen patients with effective immunotherapy more accurately.At present,there have been many studies on the establishment of models based on clinical parameters to predict the efficacy of immunotherapy.In this study,we first established an immunotherapy efficacy prediction model by combining the non-invasive radiomics characteristics with clinical parameters to explore whether it can improve the prediction efficiency.In addition,the spleen,as an important peripheral immune organ outside the primary tumor tissue,can properly regulate the interaction between the tumor and the host,and promote the development of the tumor to a certain extent.At the same time,the spleen has histological and pathophysiological interactions.The purpose of this study is to incorporate the radiomics characteristics of spleen to explore the value in predicting the efficacy of immunotherapy for primary lung tumors.Materials and methods: 137 patients with NSCLC confirmed by pathology from July2017 to April 2022 were retrospectively collected.We obtain the clinical data and the baseline enhanced CT images within one month before the first immunotherapy,extract the venous phase to delineate the three-dimensional area of interest of the lung lesions and spleen,and apply the K best absolute shrinkage and selection operator(LASSO)regression model method for feature screening,and establish three groups of models respectively based on the screened image features related to the prediction of the efficacy of immunotherapy,including: clinical parameter model(clinical group),clinical parameter combined with the radiomics characteristics model of primary lung lesions(clinical-lesion group),and clinical parameter combined with the radiomics characteristics model of primary lung lesions and spleen(clinical-lesion-spleen group).13 classifiers were used and then,we select multiple classifiers to predict the immunotherapy efficacy.Finally,the decision curve was used to evaluate the clinical value of the classifier.Results: We established three sets of models.Compared with the classifier,the AUC value of the clinical-lesion-spleen group model is generally higher than that of the clinical-lesion group model.Then,the comparison among various classifiers in the clinical-lesion-spleen group model showed that the AUC values of logistic regression and partial least squares discriminant analysis classifiers were 0.905(95% CI)and 0.901(95% CI)respectively,and the De Long test confirmed that there was significant statistical significance(P<0.05).Decision curve showed that in the clinical-lesion-spleen group model,the four classifiers of logistic regression,support vector machine,Gaussian process and quadratic discriminant analysis are used to predict the non-small cell lung cancer immunotherapy efficacy,which is superior to the "full prediction" or "no prediction" decision.Conclusion: This study establishes a prediction model for the efficacy evaluation of advanced NSCLC immunotherapy by combining non-invasive radiomics characteristics with clinical parameter,and further incorporates the spleen radiomics features.Based on the results of multiple classifiers,it reflects the predictive value of the combination of radiomics characteristics and clinical parameter in the efficacy of NSCLC immunotherapy.The radiomics characteristics of spleen are expected to be a new method to predict the immunotherapy efficacy for advanced NSCLC. |
PDF Full Text Request