| PART Ⅰ Development and Validation a Radiomics Nomogram for Diagnosing Occult Brain Metastases in Patients with Stage ⅣLung AdenocarcinomaBackgroundLung cancer is the most common cancer in the world,of which NSCLC accounts for about 80-85%.Lung adenocarcinoma（LADC）is a common pathological type of NSCLC,with an increasing incidence in recent years.About 50%of NSCLC are stageⅣ at first diagnosis due to the lack of specific symptoms.According to TNM stage,stage Ⅳ is the end stage of the tumor usually with poor prognosis.Once BM occurs,the prognosis is more poorer,and the mOS is no more than one year.Therefore,the presence or absence of BM determines the prognosis and treatment of lung cancer.Current researches have shown that risk factors associated with BM were age,sex,pathological type,and gene expression.Brain metastases are treated by brain radiotherapy,including whole brain radiotherapy（WBRT）and stereotactic radiosurgery（SRS）.There was no difference between SRS and WBRT in local tumor control rate,but SRS is only suitable for the small number and diameter of patients with early BMs,and it can significantly reduce the risk of neurocognitive decline in patients with radiotherapy compared to WBRT.Therefore,early detection of BM can not only improve the quality of life but also the prognosis of patients.Occult brain metastases are the early state of BMs,usually with a relatively small diameter of about 1cm,which cannot be distinguished by brain CT examination,but can be shown on MRI.At present,the early detection of occultation BMs is mainly by craniocerebral magnetic resonance imaging（MRI）.But MRI has its limitations,includinglong examination time,which is intolerant to patients with poor physical status,and is contraindicated for patients with metal objects in the body.It cannot be applied to all patients in stage Ⅳ lung cancer.Therefore,it is very important to find other methods to predict the risk of occult BMs in stage Ⅳ NSCLC patients in order to guide the treatment of patients,avoid excessive examination and reduce the economic pressures on public health.In recent years,radiomics as a new medical technology can extract the invisible information in medical images with high throughput through automated computing methods.It can convert the images of the regions of interest into quantitative data,and then analyze these feature data to reflect the phenotype of the entire tumor lesions.At present,it is widely used in tumor diagnosis,response evaluation and prognosis analyses.In the present study,we aimed to develop and validate aradiomics model using CT images acquired from the first diagnosisto estimate the status of occult BM in patients with stageⅣ lung adenocarcionoma.MethodsOne hundred and ninety-three patients who were first diagnosed with stage ⅣLADC were enrolled and divided into a training cohort（n=135）and validation cohort（n=58）.Then,725 radiomic features were extracted from contoured primary tumor volume of LADC.Intra-and inter-observer reproducibility test and LASSO were then implemented for feature selection.Subsequently,a radiomics signature was built.For better performance,a nomogram incorporating radiomics signature and independent clinical predictor were further constructed.Finally,the established signature and nomogram were assessed using receiver operating characteristic curve（ROC）and precision-recall curve（PRC）.Both the empirical and α-binomial model-based ROC curves and PRCs were plotted,and area under curve（AUC）and average precision（AP）of ROCs and PRCs were calculated and compared.ResultsA radiomics signaturewas constructed using eight radiomic features and has a significant correlation with the status of occult BM.The nomogram was developed by incorporating Rad-Score and primary tumor location.The nomogram yielded an optimal AUC of 0.911（95%confidence interval,CI,0.903-0.919）and AP of 0.885（95%CI,0.876-0.894）in training cohort,and AUC of 0.873（95%CI,0.866-0.80）and AP of 0.827（95%CI,0.820-0.834）in validation cohort using α-binomial model-based method.The calibration curve demonstrated that the nomogram had remarkable agreement with actual occult BMs（P=0.427）.ConclusionsThe nomogram incorporating a radiomics signature and a clinical risk factor achievedoptimal performance after holistic assessment using unbiased indexes for diagnosing occult BM of patientswho were first diagnosed with stage Ⅳ LADC.PART Ⅱ Prognostic value of Radiomics Model in Patients with StageⅣ Lung AdenocarcinomaBackgroundLung cancer is the leading cause of cancer-related mortality.At present,the prediction of the prognosis of NSCLC mainly depends on TNM stage.The later the stage is,the worse the prognosis is.However,TNM staging cannot accurately predict the prognosis of patients,because the prognosis of patients with the same staging is different even if they receive the same treatment plan.Current studies have shown that the prognosis of NSCLC is related to tumor heterogeneity and individual factors.Therefore,there is an urgent need for a more accurate method to evaluate the prognosis of patients except for tumor staging.Stage Ⅳ NSCLC is an end-stage tumor,with a median survival time（mOS）of 7-10 months,and a 5-year survival rate of 2%-13%.The main treatment strategy for stage Ⅳ NSCLC is platinum-containing chemotherapy.For patients with sensitive gene mutations,NCCN guidelines recommend small-molecule targeted drug therapy（TKI）as an effective treatment.Lung adenocarcinoma is more prone to genetic mutations than squamous cell carcinomas,while EGFR mutation is the most common type,and the corresponding small molecular targeted drugs currently have three generations.A number of large,multicenter,randomized trial showed that targeted therapy could significantly improve the survival of patients with stage Ⅳ NSCLC compared to chemotherapy,and mOS can be extended to more than 17 months.StageⅣ lung adenocarcinoma is a kind of tumor with great heterogeneity,which is characterized by diverse phenotypes and treatment methods,as well as great difference in survival prognosis.Effective prediction of patient prognosis is of great significance for guiding individual treatment of patients.Radiomics can extract high-throughput features of the entire tumor used for the diagnosis,gene phenotype,response evaluation and prognosis of NSCLC.Therefore,in this study,we attempted to explore radiomic labels through chest CT radiomic analysis of patients with stage Ⅳ lung adenocarcinoma,and established a nomogram combining radiomic labels with clinical risk factorsto evaluate the prognosis of paitents withNSCLC.MethodsIn this study,191 patients with stage Ⅳ lung adenocarcinoma were retrospectively analyzed and divided into the training cohort（n=134）and the validation cohort（n=57）.Age,sex,smoking history,maximum tumor diameter,thoracic lymph node metastasis,primary tumor location,pleural invasion,carcinoembryonic antigen（CEA）,brain metastasis,EGFR mutation,chemotherapy,brain radiotherapy,TKI therapy,and follow-up time and survival outcome were obtained through medical records.The gross tumor volumes（GTV）of the primary tumor on chest CT were contoured by two experienced radiotherapy oncologists who were blinded to all patients.Radiomic features were extracted in 3D slicer.A total of 725 features were extracted.Least absolute shrinkage and selection operator（LASSO）Cox regression analysis was used to select predictable features and a Rad-Score,was established,which was also known as radiomics signature.The patients were classified into the high-and low-risk subgroups according to the Rad-score at a cut-off point according to X-Tile,and the Kaplan-Meier survival analyses were then performed.For clinical factors,age and tumor diameter were continuous variables,and the optimal cutoff values were also determined by X-tile.Kaplan-meier survival analyses were performed for all clinical factors to determine the predictive clinical risk factors associated with survival.The Rad-Score and clinical risk factors were combined to construct the nomogram,which were then evaluated by c-index and calibration curves.ResultsNine features with non-zero coefficients were selected using Lasso-Cox regression analysis.The patients were classified into the high-and low-risk subgroups according to the Rad-score at a cut-off point of 0.58.Kaplan-Meier analysis revealed significantly different subgroup OS in both training cohort and validation cohort（p<0.0001）utilizing log-rank test.The Hazard Ratio（HR）was 6.332 for the training cohort and 2.926 for the validation cohort,and it was further confirmed that the Rad-score was good at distinguishing the the high-and low-risk subgroup.Kaplan-Meier survival analyses of all clinical factors showed that EGFR mutations and targeted therapy were clinical risk factors associated with survival.Nomogram was constructed by combining Rad-Score with EGFR mutations and targeted therapy.The C-indexes of the nomogram model were higher than that of Rad-Score in both training and validation cohort.The calibration curves of the nomogram for 1-,2-and 3-year survival demonstrated good consistency between predicted and actual survival（P-values were 0.3122、0.5716 and 0.2961 respectively）.ConclusionsIn this study,the radiomics signature could well distinguish the high-and low-risk subgroup of stage Ⅳ lung adenocarcinoma.The nomogram model established by incorporating the radiomics signature with clinical risk factors achieved better perfomance in predicting survival,and had good robustness and accuracy. |
PDF Full Text Request