Objective To predict the status of microsatellite instability (MSI) of rectal carcinoma (RC) using different machine learning algorithms based on tumoral and peritumoral radiomics combined with clinicopathological characteristics. Methods There were 497 RC patients enrolled in this retrospective study. The tumoral and peritumoral CT-based radiomic features were calculated after tumor segmentation. The radiomic features from two radiologists were compared by way of inter-observer correlation coefficient (ICC). After methods of variance, correlation, and dimension reduction, six machine learning algorithms of logistic regression (LR), Bayes, support vector machine, random forest, k-nearest neighbor, and decision tree were conducted to develop models for predicting MSI status of RC. The relative standard deviation (RSD) was quantified. The radiomics and significant clinicopathological variables constituted the radiomics-clinicopathological nomogram. The receiver operator curve (ROC) was made by DeLong test, and the area under curve (AUC) with 95% confidence interval (95% CI) was calculated to evaluate the performance of the model. Results The venous phase of CT examination was selected for further analysis because the proportion of radiomic features with ICC greater than 0.75 was higher. The tumoral and peritumoral model by LR algorithm (M-LR) with minimal RSD showed good performance in predicting MSI status of RC with the AUCs of 0.817 and 0.726 in the training and validation set. The radiomic-clinicopathological nomogram performed better in both the training and validation set with AUCs of 0.843 and 0.737. Conclusion The radiomics-clinicopathological nomogram demonstrated better predictive performance in evaluating the MSI status of RC.

Background To construct clinical and machine learning nomogram for predicting the lymph node metastasis (LNM) status of rectal carcinoma (RC) based on radiomics and clinical characteristics. Methods 788 RC patients were enrolled from January 2015 to January 2021, including 303 RCs with LNM and 485 RCs without LNM. The radiomics features were calculated and selected with the methods of variance, correlation analysis, and gradient boosting decision tree. After feature selection, the machine learning algorithms of Bayes, k-nearest neighbor (KNN), logistic regression (LR), support vector machine (SVM), and decision tree (DT) were used to construct prediction models. The clinical characteristics combined with intratumoral and peritumoral radiomics was taken to develop a radiomics and machine learning nomogram. The relative standard deviation (RSD) was used to predict the stability of machine learning algorithms. The area under curves (AUCs) with 95% confidence interval (CI) were calculated to evaluate the predictive efficacy of all models. Results To intratumoral radiomics analysis, the RSD of Bayes was minimal compared with other four machine learning algorithms. The AUCs of arterial-phase based intratumoral Bayes model (0.626 and 0.627) were higher than these of unenhanced-phase and venous-phase ones in both the training and validation group.The AUCs of intratumoral and peritumoral Bayes model were 0.656 in the training group and were 0.638 in the validation group, and the relevant Bayes-score was quantified. The clinical-Bayes nomogram containing significant clinical variables of diameter, PNI, EMVI, CEA, and CA19-9, and Bayes-score was constructed. The AUC (95%CI), specificity, and sensitivity of this nomogram was 0.828 (95%CI, 0.800-0.854), 74.85%, and 77.23%. Conclusion Intratumoral and peritumoral radiomics can help predict the LNM status of RCs. The machine learning algorithm of Bayes in arterial-phase conducted better in consideration of terms of RSD and AUC. The clinical-Bayes nomogram achieved a better performance in predicting the LNM status of RCs.

Purpose To predict the status of microsatellite instability (MSI) of rectal carcinoma (RC) using different machine learning algorithms based on tumoral and peritumoral radiomics combined with clinicopathological characteristics. Methods There were 487 RC patients enrolled in this retrospective study. The tumoral and peritumoral CT-based radiomic features were calculated after tumor segmentation. The radiomic features from two radiologists were compared by the method of inter-observer correlation coefficient (ICC). After methods of variance, correlation, and dimension reduction, six machine learning algorithms of logistic regression (LR), Bayes, support vector machine, random forest, k-nearest neighbor, and decision tree were conducted to develop models in predicting MSI status of RC. The relative standard deviation (RSD) was quantified. The radiomics and significant clinicopathological variables were integrated the radiomics-clinicopathological nomogram. The receiver operator curve (ROC) was made by Delong test and the area under curve (AUC) with 95% confidence interval (95%CI) was calculated to evaluate the performance of the model. Results The venous phase of CT examination was selected for further analysis because of the proportion of radiomic features with ICC greater than 0.75 was higher. The tumoral and peritumoral model by LR algorithm (M-LR) with minimal RSD showed good performance in predicting MSI status of RC with the AUCs of 0.817 and 0.726 in the training and validation set. The radiomic-clinicopathological nomogram conducted better in both the training and validation set with the AUCs of 0.843 and 0.737. Conclusion The radiomics-clinicopathological nomogram demonstrated the better predictive performance in evaluating the MSI status of RC.