| Part I Analysis of risk factors for lymph node metastasis of rectal cancer and construction of clinical-ultrasound nomogram model ObjectiveThis part of the study mainly analyzes the risk factors of rectal cancer lymph node metastasis,and establishes a clinical-ultrasound nomogram(CUN)model based on independent preoperative risk factors to predict the status of rectal cancer lymph node metastasis.MethodsThis part of the study retrospectively analyzed 79 patients with rectal cancer.The patients’ general clinical data,laboratory biochemical data,transrectal three-dimensional ultrasound examination data and pathological examination data weregrouped according to the positive and negative status of pathological lymph nodemetastasis.Logistic regression analysis was performed to obtain independent riskfactors related to lymph node metastasis in patients with rectal cancer.Find the independent risk factors related to lymph node metastasis of rectal cancer.According to independent preoperative risk factors,a clinical-ultrasound nomogram model was constructed to predict the status of lymph node metastasis in patients with rectal cancer.ResultsLymph node status reported by ultrasound,tumor pathological T staging are independent risk factor for lymph node metastasis.Combined with preoperative CEA,Ca199 and lymph node status reported by ultrasound,a clinical-ultrasound nomogram model was formed.The results of the prediction model calibration curve show that the model is well calibrated(p-values are 0.948 and 0.999),and the ROC curve results show that the model has a good predictive ability(AUC is 0.734 and 0.692 in the training set and test set).The results of the H-L test testing the nomogram model show that the difference between the predicted value of the model and the actual observation data is not statistically significant(the p-values in the training set and test set are 0.504 and 0.307).ConclusionLymph node status reported by ultrasound,tumor pathological T staging are independent risk factors related to lymph node metastasis of rectal cancer.The clinical-ultrasound nomogram model composed of preoperative CEA,Ca199 and lymph node status reported by ultrasound has a better performance in predicting lymph node metastasis status,and can be used to predict lymph node metastasis in patients with rectal cancer before surgery.Part Ⅱ Application of Transrectal Three-dimensional Ultrasound Radiomics in Predicting Lymph Node Metastasis in Patients with Rectal Cancer ObjectiveThis part of the study mainly explores the clinical application value of the radiomics method based on transrectal three-dimensional ultrasound in the prediction of lymph node metastasis in patients with rectal cancer.MethodsThis part of the study retrospectively analyzes 79 patients with rectal cancer,and extracts the radiomics features of their transrectal three-dimensional ultrasound images.Subsequently,independent sample t-test and Least absolute shrinkage and selection operator regression(LASSO regression)were used for feature selection.According to the final screened features,six machine learning classifier models are established to predict the status of rectal cancer lymph nodes,including K-NearestNeighbor(KNN),Logistics Regression(LR),and Muti-Layer Perception,MLP),Random Forest(RF),Decision Tree(TREE)and Support Vector Machine(SVM).Draw the receiver operating characteristic curve(ROC)and calculate the area under the curve(AUC),accuracy,sensitivity,specificity and Youden index(Youden index) to evaluate the diagnostic performance of each model.Perform Delong Test to check the difference in diagnostic performance between models.ResultsIn this part of the study,a total of 8 categories and 1694 imaging omics features were extracted from the transrectal three-dimensional ultrasound images.Among them,34 imaging omics characteristics were significantly related to the status of lymph node metastasis(P <0.05).After feature selection,10 features are finally screened out to build a machine learning classifier model.According to the comprehensive analysis,the LR model and MLP model have the best diagnostic performance.Their AUCs in the training set are 0.817 and 0.908.Their AUCs in the test set are 0.958 and 0.958.Conclusion34 radiomics features were significantly correlated with lymph node metastasis status(P <0.05).The LR and MLP models constructed based on the radiomics features after screening have the best performance in predicting the status of lymph nodes in rectal cancer.The application of radiomics can accurately assess the status of lymph node metastasis in patients with rectal cancer before surgery.Part Ⅲ The construction of clinical-ultrasound-radiomics nomogram model and the performance comparison between clinical-ultrasound-radiomics nomogram model and clinical-ultrasound nomogram model ObjectiveThis part of the study combined clinical risk factors and ultrasound Radiomic signature to construct a clinical-ultrasound-radiomics nomogram(CURN)model to predict the status of lymph node metastasis in patients with rectal cancer,and compared the diagnostic performance of clinical-ultrasound-radiomics nomogram model and clinical-ultrasound nomogram model.MethodsThis part of the study retrospectively analyzed 79 patients with rectal cancer,combined with their independent preoperative risk factors and Radiomic signature to construct a clinical-ultrasound-radiomics nomogram model to predict the status of lymph node metastasis.The Delong Test was performed to examine the difference in diagnostic performance between the clinical-ultrasound-radiomics nomogram model and the clinical-ultrasound nomogram model.ResultsCombined with preoperative CEA,Ca199 level,lymph node status reported by ultrasound and Radiomic signature,a clinical-ultrasound-imaging nomogram model was constructed.The ROC curve results show that the model has good predictive ability(AUC is 0.908 and 0.888 in the training set and the test set),and the prediction model calibration curve results show that the model is well calibrated(p-values are 0.921 and 0.848).The results of the H-L test testing the nomogram model show that the difference between the predicted value of the model and the actual observation data is not statistically significant(the p-values in the training set and test set are 0.885 and 0.389).The results of Delong Test showed that the diagnostic performance difference between the clinical-ultrasound model and the clinical-ultrasound-radiomics model was statistically significant(P=0.009),the area of difference between the two ROC curves was 0.173,the standard error was 0.067, and the Z number was 2.608.ConclusionThe clinical-ultrasound-imaging nomogram model constructed by combining preoperative CEA,Ca199,lymph node status reported by ultrasound,and Radiomic signature has excellent performance in predicting lymph node metastasis status.The diagnostic performance of the clinical-ultrasound-radiomics nomogram is better than that of the clinical-ultrasound nomogram model. |
PDF Full Text Request