| Objective: To develop and validate CT radiomics models predicting the response of colorectal liver metastases after combined with bevacizumab chemotherapy,and to compare the changes of CT texture parameters before and after chemotherapy.Materials and Methods: A total of 141 liver metastases were selected from 50 patients(18patients in response and 32 patients in non-respond)with colorectal cancer liver metastases who received chemotherapy combined with bevacizumab.792 texture features were obtained from the arterial and portal phases with each patient.The texture parameters were selected by ANOVA,Spearman correlation test and LASSO regression.SVM and Logistics regression were used to establish the radiomics model of the selected parameters and then validation was performed by test cohort.The radiomics model is evaluated by accuracy,sensitivity,specificity,and accuracy.Paired t test was used to compare the changes of CT texture parameters before and after chemotherapy in 135 liver metastases.Result: The imaging ensemble model of 12 texture features was established by SVM and logistic regression with high accuracy,sensitivity,specificity and precision.The training set of SVM: 0.95,0.98,0.89,0.83;testing set,0.84,0.93,0.67,0.83;logistic regression model training set: 0.88,0.96,0.74,0.79;testing set 0.84,0.89,0.73,0.79.The area under the ROC curve for the two model training sets were 0.9 and 0.79,respectively.After chemotherapy,211 texture parameters changed significantly in the response cohort,and 130 texture parameters of the non-response cohort changed significantly.Among them,the same 62 texture parameters significant changes in both two cohort.Conclusion: The radiomics model can predict the response of patients with colorectal liver metastases after chemotherapy combined with bevacizumab.The texture analysis can evaluate the changes of the internal structure of liver metastases after chemotherapy,which can be used as a monitor of chemotherapy combined bevacizumab. |
PDF Full Text Request