Research On Evaluation And Prediction Models Of First-Line Chemotherapy Efficacy For Non-small Cell Lung Cancer Using CT Images

Lung cancer is one of the most common types of cancer and the leading cause of cancerrelated deaths worldwide.Among these patients,non-small cell lung cancer(NSCLC)patients are more common and account for over 85%of all lung cancer patients.Early NSCLC patients are asymptomatic,while most are diagnosed in the advanced stage due to the high invasiveness and lack of early screening tools.Standard treatment options for NSCLC patients include surgical resection,chemotherapy,radiation therapy,targeted therapy,and immunotherapy.Although the clinical treatments are increasing and constantly optimized,the therapeutic effect of NSCLC patients is unsatisfactory due to tumor heterogeneity,physical condition,drug resistance,and other factors.Therefore,it is important to effectively diagnose the patients’ condition and design reasonable and effective regimens for NSCLC patients.The platinum-based chemotherapy is the first-line treatment for advanced NSCLC patients,which effectively improves the survival rate of NSCLC patients.However,the objective remission rate of chemotherapy is limited,and the treatment responses vary greatly among patients.Chemotherapy can cure patients,but it also has some toxic side effects.Patients taking long-term chemotherapy may also develop drug resistance,which limits the efficacy of chemotherapy.Therefore,it is of great clinical significance to predict chemotherapy response for NSCLC patients.An effective prediction model can support the clinical personalized treatment decisions,which helps improve the overall survival rate and prolong the life of patients.Computed tomography(CT)is an important imaging method for detecting,diagnosing,and reviewing lung cancer in clinical practice.Continuous CT follow-ups reflect the patients’ tumor progression during treatment,potentially predictive of treatment efficacy.In this study,with CT images of advanced NSCLC patients,four main parts of work are performed to predict the chemotherapy response as follows:(1)Evaluation of tumor responses of NSCLC patients during treatment according to the Response Evaluation Criteria in Solid Tumors(RECIST)and tumors’ volumetric changes.RECIST assesses tumor progression simply and effectively,but it is rough,which affects the accuracy of evaluation and results in errors in practical application.(2)Prediction of tumor progression for NSCLC patients after two cycles of chemotherapy based on radiomics analysis.Radiomics features are selected and modeled with machine learning algorithms to predict the tumor progression after chemotherapy.The experimental results demonstrate that the radiomics features can effectively predict the tumor progression of patients.(3)Prediction of tumor progression for NSCLC patients after two cycles of chemotherapy based on transfer learning of a convolutional neural network.ResNet is transferred to learn CT images’ deep features and predict tumor progression after chemotherapy.The experimental results show that the deep learning models can achieve better prediction results than those of radiomics analysis.(4)Prediction of final chemotherapy response on NSCLC patients using delta-radiomics analysis.The delta-radiomics features are calculated using the first two CT images during patient treatment.After feature selection,machine learning models are constructed to predict the final chemotherapy effect on patients.The experimental results show that delta-radiomics features are potentially predictive of prognosis.