| BackgroundLung cancer is one of the most common malignant tumors globally,with high incidence and mortality rates,and has become the leading cause of cancer deaths worldwide.Non-small cell lung cancer(NSCLC)accounts for about 85%of cases,and almost half of patients are diagnosed at an advanced stage.Current treatment options include surgery,radiotherapy,chemotherapy,targeted therapy,immunotherapy,traditional Chinese medicine,and cryoablation.Cryoablation combined with traditional Chinese medicine is a new treatment mode for lung cancer guided by TCM theory,which uses cryoablation to locally attack lesions and TCM to regulate the overall physical condition.Previous studies have demonstrated that this treatment mode has good therapeutic effects on NSCLC patients,and multicenter randomized controlled trials are currently underway to explore its advantages in specific populations.Radiomics,proposed in 2012,aims to automatically and quantitatively extract and describe medical imaging features.It has been widely used in the field of tumor research,especially in survival prediction,because it can accurately evaluate local lesion conditions.Based on this,this study combined perioperative TCM treatment with radiomics methods and the team’s previous research to construct a clinical comprehensive prognostic model.The aim is to provide help and reference for preoperative assessment of cryoablation combined with traditional Chinese medicine treatment for NSCLC.Objectives1.Through radiomics analysis,the optimal method and parameters for cold ablation combined with traditional Chinese medicine treatment of non-small cell lung cancer(NSCLC)were determined,and a radiomics score was established.2.An evaluation standard was established for the exposure level of traditional Chinese medicine treatment during the perioperative period,and its short-and long-term efficacy and prognostic value were assessed.3.Based on radiomics research and previous studies,a clinical comprehensive prognostic prediction model was established for cold ablation combined with traditional Chinese medicine treatment of NSCLC.4.Based on the clinical comprehensive prognostic prediction model,the advantages of cold ablation combined with traditional Chinese medicine treatment over standard treatment in specific patient populations were explored.Materials and MethodsA total of 170 NSCLC patients who underwent CT-guided cold ablation at our hospital between 2013 and 2020 were retrospectively included.The lung CT’s region of interest was extracted using manual and automatic segmentation methods.Radiomics features were extracted using the pyradiomics package,and features related to prognosis were selected using various methods including ICC-based threshold feature selection,correlation-based network construction and feature selection,single-factor Cox regression-based threshold feature selection,and LASSO-based wrapper feature selection.A multi-factor Cox model was then established,and a radiomics score was calculated to evaluate the prognostic significance of the radiomics model.The dataset was divided for internal validation,and the model was compared with similar models.Next,unsupervised hierarchical clustering was performed based on the exposure level of traditional Chinese medicine treatment during the perioperative period.Finally,a clinical comprehensive prognostic prediction model was constructed based on variables such as radiomics score,exposure level of traditional Chinese medicine treatment during the perioperative period,M stage,neutrophil-to-lymphocyte ratio,fibrinogen content,and size of target lesions.The model was visualized using a column chart,and the optimal cutoff value was determined to distinguish between the advantage and non-advantage groups.Based on the "2022 Chinese Society of Clinical Oncology Non-Small Cell Lung Cancer Diagnosis and Treatment Guidelines," subgroup analysis of all subjects was conducted,and the survival benefits of cold ablation combined with traditional Chinese medicine versus standard treatment were evaluated using a single-group objective value method.Results1.Patients inclusionA total of 229 NSCLC patients with imaging records were screened,59 ineligible patients were excluded,and finally 170 patients were included.A total of 170 preoperative chest CT scans were collected.2.Consistency evaluation between automatic and manual segmentationThe intra-group consistency(ICC)of automatic segmentation was 1.0,indicating good consistency.The median ICC of manual segmentation was 0.9628(0.9077,0.9834),and 809 features(76.97%)had an ICC greater than 0.9.Compared with manual segmentation,the median ICC of automatic segmentation was 0.2496(0.0849,0.4396),indicating poor accuracy of automatic segmentation.3.Radiomics feature selection process809 features were selected through threshold-based feature selection method based on intra-group consistency with the threshold set at 0.9 out of a total of 1051 features.406 features with high centrality were selected through correlation-based network construction and feature selection with threshold set at the median value of betweenness centrality.253 features with prognostic significance were selected through univariate Cox regression analysis.Then,7 radiomics features and corresponding regression coefficients β were selected through LASSO-based wrapper feature selection method,including 6 risk factors and 1 protective factor.4.Establishment of Multivariate Cox Regression Model and Evaluation of Prognostic PerformanceThe radiomics model was established by incorporating the above 7 features into a multivariable Cox regression model,with a concordance index(C-index)of 0.695.The survival difference between groups stratified by the median radiomics score was statistically significant(P<0.0001,Log-rank test),and lower radiomics score was associated with better prognosis,with a hazard ratio of 0.5184(P<0.0001).The areas under the curve(AUC)of radiomics model for predicting 1-year,2-year,and 3-year survival were 0.750,0.762,and 0.760,respectively.Decision curve analysis(DCA)showed that the application of the radiomics model could improve the net benefit of clinical decision-making.5.Internal validation and horizontal comparison of the original modelThe original dataset was randomly divided in a 7:3 ratio for 10-fold cross-validation,with an average C-index of 0.620±0.038 and average AUCs of 0.653±0.052,0.691±0.071,and 0.665 ± 0.124 for predicting 1-year,2-year,and 3-year survival,respectively.Compared with a similar study 1,the radiomics model had better prognostic performance(C-index=0.620±0.038 vs 0.589 ± 0.024),and the difference in predicting 1-year survival was statistically significant(P=0.0021).Compared with a similar study 2,the radiomics model had slightly higher prognostic performance(C-index=0.695±0.023 vs 0.694 ±0.019)±and there was no statistically significant difference between the two models in predicting 1-year,2-year,and 3-year survival(1-year:AUC=0.750 vs 0.785,P=0.3480;2-year:AUC=0.762 vs 0.808,P=0.2879;3-year:AUC=0.760 vs 0.788,P=0.6188).6.Hierarchical clustering and exposure assessment based on perioperative TCM treatmentHierarchical clustering was performed on all samples based on traditional Chinese medicine treatment during the perioperative period,resulting in 2 clusters:cluster 1 included 126 samples and cluster 2 included 44 samples.There were significant inter-group differences(P<0.001)in the use of Chinese herbal decoctions,intravenous traditional Chinese medicine,oral traditional Chinese medicine,and external traditional Chinese medicine treatment,with a higher proportion of cluster 2 receiving treatment and thus belonging to the high exposure group.7.Evaluation of perioperative TCM treatment’s efficacy and prognostic valueThere was no significant difference in overall prognosis between the high and low exposure groups to traditional Chinese medicine treatment during the perioperative period(P=0.91),but the survival curves of the two groups significantly diverged within the first 6 months,suggesting that traditional Chinese medicine treatment during the perioperative period may improve short-term prognosis,but its long-term efficacy remains uncertain.Regarding prognostic value,there was a difference in predicting 1-year survival rate between the original imaging model and the imaging score model(P=0.026),but there was no significant difference in predicting 2-year(P=0.849)or 3-year(P=0.403)survival rates.After incorporating traditional Chinese medicine treatment during the perioperative period,the predictive performance slightly improved(mean AUC=0.709±0.059 vs 0.707±0.051),but there was no statistically significant difference.8.Establishment of a comprehensive clinical prognosis model and visualizaiton of a nomogramMultiple factor Cox regression analysis was performed by including variables such as imaging score,exposure level to traditional Chinese medicine treatment during the perioperative period,M stage,fibrinogen content,size of target lesions,and neutrophil-to-lymphocyte ratio to establish a comprehensive clinical prognostic prediction model.The C-index was 0.730±0.020 and the mean AUC was 0.806 ±0.03,which was higher than both the original imaging model(C-index=0.695±0.023,AUC=0.768 ±0.021,P<0.0001)and a similar model 2(C-index=0.694±0.019,AUC =0.795 ± 0.023,P=0.093).9.Identification of advantageous subgroups and comparison of therapeutic efficacy with standard treatmentAccording to the comprehensive clinical prognostic prediction model,the optimal cut-off value for the total score was determined to be-0.1568,with scores equal to or lower than this value belonging to the advantageous subgroup and scores above this value belonging to the non-advantageous subgroup.For inoperable stage ⅢA,ⅢB,and ⅢC NSCLC,the median survival period was 32 months for the advantageous group,which was significantly longer than the 16.5 months for first-line treatment(P=0.007).For stage Ⅳ driver gene-negative non-squamous NSCLC,the median survival period was 23.12 months for the advantageous group,which was significantly longer than the 15.8 months for first-line treatment(P=0.011).Conclusion1.In the delineation of local lesions,manual segmentation has better consistency and accuracy than automatic segmentation,and can be used as material for radiomics research.2.By using a series of feature selection pathways,7 radiomic features and corresponding regression coefficients β were obtained,including 6 risk factors and 1 protective factor,and the radiomic score was finally established,which can quantitatively evaluate the local lesion and accurately predict the subject’s prognosis.3.The exposure level of traditional Chinese medicine treatment during the perioperative period was evaluated by hierarchical clustering.Patients who received more traditional Chinese medicine treatment during the perioperative period had better short-term prognosis within 6 months and had a certain ability to predict prognosis.Long-term prognosis could not be accurately evaluated.4.A comprehensive clinical prognostic prediction model was established by incorporating variables such as radiomic score,exposure level to traditional Chinese medicine treatment during the perioperative period,M stage,fibrinogen content,neutrophil-to-lymphocyte ratio,and size of target lesions,which can accurately predict the subject’s 1-year,2-year,and 3-year survival probabilities.5.According to the comprehensive clinical prognostic prediction model,the advantageous population for cryoablation combined with traditional Chinese medicine treatment was determined.For inoperable stage ⅢA,ⅢB,and ⅢC NSCLC,and stage Ⅳdriver gene-negative non-squamous NSCLC patients,the advantageous group had significantly better prognosis than first-line treatment,with survival periods extended by 15.5 months and 7.32 months,respectively. |
PDF Full Text Request