| Part Ⅰ.Computer tomography semantic features for assessing cell proliferation activity in stage I lung adenocarcinomaObjective:Surgically resected stage I lung adenocarcinoma has a wide variation in prognosis.Identifying high-risk patients is significant for optimizing therapeutic strategy.As a reflection of cell proliferation activity,Ki-67 labeling index(LI)is a well-established prognosticator for lung adenocarcinoma.This study aimed to investigate the relationships of clinicopathologic features and computer tomography(CT)semantic features with Ki-67 in stage I lung adenocarcinoma.Methods:A total of 182 patients with pathologically-confirmed stage I lung adenocarcinoma were retrospectively enrolled.The Ki-67 LI was acquired by immunohistochemistry.The patients were divided into two categories of low Ki-67 expression(Ki-67 LI<10%)and high Ki-67 expression(Ki-67 LI>10%).Two radiologists retrospectively analyzed preoperative CT images and independently evaluated CT semantic features.Univariate analysis and multivariate logistic regression analysis were used to investigate the relationships of clinicopathologic features and CT semantic features with Ki67 expression in stage I lung adenocarcinoma.Results:There were 73(40.1%)low-expressers and 109(59.9%)Ki-67 high-expressers out of 182 patients.High Ki-67 expression was more frequently found in(a)males and patients with smoking history,higher pack-year,CEA>5 ug/L,CYFRA21-1>3.3 ug/L,and TMI>1;(b)patients with poorly differentiated or stage IB tumors;(c)tumors with larger long-axis and short-axis diameters,lower tumor shadow disappearance rate(TDR),lobulation number three or more,spiculation,vacuolation,vascular invasion,vascular convergence,thickened bronchovascular bundle,pleural attachment and peripheral fibrosis;and(d)solid tumors(all P<0.05).Solid predominant lung adenocarcinoma had the highest Ki-67 LI,followed by micropapillary,papillary and acinar predominant lung adenocarcinoma,while lepidic predominant lung adenocarcinoma had the lowest Ki-67 LI(P<0.001).TDR was negatively correlated with Ki-67 LI(r=-0.601,P<0.001).Multivariate logistic regression analysis revealed that high Ki-67 expression was independently associated with male,poor differentiation,lower TDR and solid tumors in patients with stage I lung adenocarcinoma.Conclusions:Ki-67 expression differed distinctly according to lung adenocarcinoma histological subtypes and histological grades.Gender,histological grade,TDR and attenuation type are independent factors associated with Ki-67 expression.Part Ⅱ.Development and validation of a deep learning signature for predicting lymph node metastasis in lung adenocarcinomaObjective:Accurate prediction for lymph node metastasis in lung adenocarcinoma is crucial to formulate individualized therapeutic options.This study aimed to develop and validate a deep learning signature for predicting lymph node metastasis in patients with lung adenocarcinoma.Methods:A total of 612 patients with pathologically-confirmed lung adenocarcinoma were retrospectively enrolled and were randomly divided into training cohort(n=489)and internal validation cohort(n=123).Besides,108 patients were enrolled and constituted the independent test cohort(n=108).Patients’clinicopathological data was collected and CT semantic features were evaluated.The Swin Transformer architecture was adopted to develop a deep learning signature predictive of lymph node metastasis.The radiomics features were derived from presurgical contrast-enhanced CT images.After preliminarily excluding unstable features and features unrelated to lymph node metastasis,the remaining radiomics features were selected by multivariate logistic regression analysis to generate a radiomics signature.Multivariate logistic regression analysis was also performed to construct a clinical-semantic model based on clinical features and CT semantic features.Model predictive performance was evaluated by area under the receiver operating characteristic curve(AUC),calibration curve and decision curve analysis.The comparisons of AUCs were conducted by the DeLong test.Results:The proposed deep learning signature yielded the AUCs significantly higher than clinical-semantic model in training cohort(0.961 vs.0.823,P<0.001),internal validation cohort(0.948 vs.0.781,P<0.001)and independent test cohort(0.960 vs.0.853,P=0.002).Similarly,deep learning signature achieved the AUCs significantly higher than radiomics signature in training cohort(0.961 vs.0.884,P<0.001),internal validation cohort(0.948 vs.0.863,P=0.019)and independent test cohort(0.960 vs.0.886,P=0.029).The calibration curves show that the predicted probabilities using deep learning signature were in good agreement with the actual observed probabilities.According to decision curve analysis,deep learning signature had the highest net benefit compared with both the clinicalsemantic model and radiomics signature.The combination of radiomics signature and deep learning signature with clinical-semantic model revealed a significant improvement in predictive performance over clinical-semantic model in all three cohorts.Conclusions:The proposed deep learning signature based on Swin Transformer achieved a promising performance in predicting lymph node metastasis in lung adenocarcinoma and could confer important information in noninvasive mediastinal lymph node staging,which is of great value to formulate appropriate and individualized therapeutic options.Part Ⅲ,Development and validation of a clinical-radiomics nomogram for predicting survival and adjuvant chemotherapy benefit in resected early-stage lung adenocarcinomaObjective:To construct a clinical-radiomics nomogram for predicting disease-free survival(DFS)in node-negative,early-stage(I-II)lung adenocarcinoma,and to explore the potential of clinical-radiomics nomogram in evaluating adjuvant chemotherapy benefit.Methods:A total of 310 eligible patients with node-negative,early-stage lung adenocarcinoma were retrospectively enrolled and divided into training cohort(n=186)and validation cohort(n=124).The radiomics features were derived from presurgical contrast-enhanced CT images.After preliminarily excluding unstable and redundant features,the remaining features were selected by least absolute shrinkage and selection operator Cox regression to generate a radiomics signature associated with disease-free survival.The univariate and multivariate Cox regression analyses were performed to identify the independent clinical risk predictors and construct a clinical nomogram.The radiomics signature was incorporated to construct a clinical-radiomics nomogram along with the independent clinical risk predictors.The risk score(RS)for clinical-radiomics nomogram was calculated.Patients were categorized into high RS and low RS subgroup.The association between adjuvant chemotherapy and survival benefit was respectively assessed in high RS and low RS subgroup.Model performance was evaluated with reference to discrimination quantified by Harrell concordance index(C-index),integrated discrimination improvement(IDI),and net reclassification index(NRI),calibration,and clinical utility.Kaplan-Meier curve and Log-rank test were performed to compare the survival differences between subgroups.Results:Histological grade and tumor diameter were finally included to construct the clinical nomogram.Histological grade,tumor diameter and radiomics signature were finally included to construct the clinical-radiomics nomogram.The clinical-radiomics nomogram achieved C-indexes of 0.822(95%CI:0.769,0.876)in training cohort and 0.802(95%Cl:0.716,0.888)in validation cohort.The incorporation of radiomics signature into clinicalradiomics nomogram showed an incremental benefit over clinical nomogram according to the improved NRI and IDI.The calibration curves and decision curve analysis further verified the good calibration and clinical utility of clinical-radiomics nomogram.Further,ACT was significantly associated with survival benefit in high RS subgroup(HR:0.489;95%Cl:0.262,0.911;P=0.024),whereas ACT was not beneficial to patients in low RS subgroup(HR:0.324;95%Cl:0.042,2.510;P=0.281).Conclusions:The clinical-radiomics nomogram approach can accurately predict prognosis in patients with node-negative,early-stage lung adenocarcinoma.This nomogram also has a potential to identify the beneficiaries of adjuvant chemotherapy,which might serve as a helpful tool in informing therapeutic decision-making. |
PDF Full Text Request