A novel clinical radiomics nomogram to predict disease activity in Crohn’s disease

OBJECTIVE: To develop and validate a novel CT enterography (CTE)-based radiomics nomogram for predicting disease activity in Crohn's disease (CD). METHODS: The CTE images and clinical data of 133 CD patients were retrospectively collected. The CTE-based radiomics features were extracted and screened by t-test and least absolute shrinkage and selection operator regression algorithm. Significant clinical variables were identified by multifactor logistic regression analysis. Then a combined model of clinical and radiomics features was established by multifactorial logistic regression, and a nomogram was plotted. RESULTS: 11 and 16 best radiomics features were screened based on CTE venous phase and arterial phase images, respectively. The area under ROC curve (AUC) of the venous radiomics model was higher than that of the arterial radiomics model on both the training and test sets (0.948 vs 0.927, 0.915 vs 0.878). Venous CT value, erythrocyte sedimentation rate and C-reactive protein were clinically relevant independent predictors of CD activity, and the AUC of the clinical model constructed from the 3 predictors was 0.873 and 0.822 on the training set and test set, respectively. The combined model had AUCs of 0.968 and 0.944 on the training and test sets, respectively. And the accuracy, sensitivity, and specificity were 0.900, 0.913, and 0.882 on the test set, respectively, which were higher than the other models. CONCLUSIONS: We develop a novel clinical radiomics nomogram to predict CD activity, which can assist clinicians in individualized treatment. ADVANCES IN KNOWLEDGE: This study is a novel attempt to establish a combined clinical-imaging graph model to predict the CD activity.