| Objective: Lung cancer is one of the most common malignant tumors.Early-stage lung adenocarcinoma grows slowly with a high probability of operative resection and a good prognosis.However,about 60% of patients with adenocarcinoma are found to be in the intermediate or advanced stages for the lack of specific symptoms.Therefore,,and a large part has lost the opportunity of operative treatment with survival time greatly shortened.Over the last decade,advances in molecularly targeted drugs for thoracic oncology have led to a new emphasis on accurate analyses of biomolecular markers in a subset of lung adenocarcinoma.Patients with advanced lung adenocarcinoma harboring epidermal growth factor receptor(EGFR)-activating mutations showed a significant progression-free survival(PFS)benefit with reduced side effects from treatment with tyrosine kinase inhibitor(TKIs).TKI therapy had already been employed as first-line systemic therapy before chemotherapy.Biopsy is the only widely used means to identify mutations of EGFR in unresectable lesions,but some patients refuse the procedure due to the risk of hemorrhage and pneumothorax.Furthermore,it is sometimes difficult to obtain tissue samples from inaccessible locations in some cases.Therefore,automatic,non-invasive,and cost-effective alternatives are desirable.Radiomics refers to the systematic extraction and analysis of features from digital medical images with the intent of creating mineable databases to aid in diagnosis and treatment.The aim of first part is to develop a radiogenomic approach to identify EGFR mutations in advanced lung.Treatment methods for advanced lung cancer mostly include radiotherapy,chemotherapy,targeting,immunization,etc.,and application sequences are split into first-line,second-line,and even third-line.It is urgently needed a more accurate prognostic assessment method than tumor staging to estimate the patient’s survival time to choose therapeutic regimens rationally.Previous studies have shown correlations between radiomics and tumor prognosis.The second part of this study attempts to predict the overall survival time of advanced lung adenocarcinoma by radiomics.Methods: This study involved 201 patients in Part Ⅰ and 165 patients in Part II with advanced lung adenocarcinoma.A total of 396 features were extracted from manual segmentation based on enhanced and non-enhance CT imaging after preprocessing.Part Ⅰ: The Lasso algorithm was used for feature selection,6 machine learning methods were used to construct radiomics models.Receiver operating characteristic(ROC)curve analysis was applied to evaluate the performance of the radiomic signature between different data and methods.A nomogram was developed using clinical factors and the radiomic signature,then it was analyzed based on its discriminatory ability and calibration.Decision curve analysis(DCA)was implemented to evaluate the clinical utility.Part II: In the training group,Lasso-Cox regression model was used to screen features.Then they were substituted into multivariate Cox proportional hazards regression model.The training group and the validation group were divided into long and short survival groups with cutoff-points of 1,2,3 years,and then the ROC curve was used to estimate the predictive ability.The survival curves for high and low risk in the training and validation group were plotted.An overall prediction model was created as a form of nomogram to predict the patient’s 1-,2-,and 3-year survival,then it was analyzed based on its discriminatory ability and calibration.Results: Part Ⅰ : Ten features for contrast data and eleven features for noncontrast data were selected through LASSO algorithm.The performance of the radiomic signature for contrast images was better than that for noncontrast images in all of the 6 different machine learning methods.Finally,the best radiomics signature was built with logistic regression method based on enhanced CT imaging with an area under the curve(AUC)of 0.851(95% CI,0.750 to 0.951)in the validation cohort.A nomogram was developed using the radiomic signature and sex with a C-index of 0.908(95% CI,0.862 to 0.954)in the training cohort and 0.835(95% CI,0.825 to 0.845)in the validation cohort.It showed good discrimination and calibration(Hosmer-Lemeshow test,P = 0.621 for the training cohort and P = 0.605 for the validation cohort).Part II : In the training group,6 survival-related features were screened from 396 features.The multivariate Cox proportional hazards regression model had a C-index of 0.721(95% CI = 0.672,0.770)in the training group and 0.676(95% CI = 0.583,0.769)in the validation group.The training group and the validation group were separated to long and short survival groups with cutoff-points of 1,2,3 years,and then the ROC curve was used to evaluate the predictive ability.The AUC were 0.738,0.836,0.759 in the training group and 0.729, 0.682,0.699 in the validation group.In terms of predicting the high and low risk of death,according to the X-tile software,1.81 was selected as the cutoff-point,and the survival curves for high and low risk in the training and validation group were plotted,which were statistically significant.Three clinical features(EGFR sensitive mutations,ECOG,and brain metastases)were screened trough univariate Cox regression.An overall prediction model was established with them and radiomic signature as a form of nomogram to predict the patient’s 1-,2-,and 3-year survival.The C-index of the training group was 0.798(95% CI = 0.758,0.838);the C-index of the validation group was 0.709(95% CI = 0.626,0.792).Calibration curves also showed good coincidence.Conclusion: Radiomics signature could help distinguish between EGFR positive and wild-type advanced adenocarcinomas.Radiomics model based on enhanced CT imaging provided more predictive efficacy.Radiomics signature showed ability to predict the OS of such patients. |
PDF Full Text Request