Establishment And Evaluation Of The Predictive Model For N-staging Of Colorectal Cancer

Objective: This study is aimed to establish and evaluate the value of the predictive model for colorectal cancer N-staging.Method: The clinical,imaging and pathological data of 611 patients with colorectal cancer(419 patients in the training group and 192 patients in the validation group)were collected.A spearman correlation analysis was used to validate the relationship among these factors and pathological T-staging.A prediction model was trained with the random forest algorithm.T staging of the patients in the validation group was predicted by both prediction model and traditional method.The consistency,accuracy,sensitivity,specificity and area under the curve(AUC)were used to compare the efficacy of the two methods.SPSS software was used to test t-hypothesis of continuous variables(sex and location of tumors)respectively.Chi-square tests were performed on discrete variables(age,CEA,carbohydrate antigen 19-9,carbohydrate antigen 72-4,maximum diameter of tumors,enhancement rate).The features with the highest correlation were screened out.Machine learning model was built with the features obtained after dimensionality reduction.ROC curve and sensitivity specificity were used.Results: The N stage of colorectal cancer was correlated with CA199,A,V,A-P,V-P,A-P/P,V-P/P(P < 0.05),but not with age,maximum diameter,CEA,CA72-4.The Kappa coefficient of the training set was 0.78452(95% confidence interval was 0.69073-0.87830),and the Kappa coefficient of the test set was 0.58333(95% confidence interval was 0.39579-0.77088).The accuracy of the training set and the test set for predicting N stage of colorectal cancer was 0.971 and 0.792 respectively.Conclusion: Random Forest(RF)model based on CT images combined with clinical and pathological data can improve the imaging diagnostic efficiency of preoperative N staging of colorectal cancer.This model has a good efficacy when predict the preoperative N-staging of colorectal cancer.