Chao Zhu1, Yongmei Yu2, Shihui Wang2, Xia Wang1, Yankun Gao1, Cuiping Li1, Jianying Li3, Yaqiong Ge3, Xingwang Wu1. 1. Department of Radiology, The First Affiliated Hospital of Anhui Medical University, Hefei, 230022, People's Republic of China. 2. Department of Radiology, The First Affiliated Hospital of Wannan Medical College, Wuhu, 241001, People's Republic of China. 3. GE Healthcare China, Shanghai, 210000, People's Republic of China.
Abstract
PURPOSE: To establish a clinical radiomics nomogram to differentiate Crohn's disease (CD) from intestinal tuberculosis (ITB). PATIENTS AND METHODS: Ninety-three patients with CD and 67 patients with ITB were recruited (111 in training cohort and 49 in test cohort). The region of interest (ROI) for the lesions in the ileocecal region was delineated on computed tomography enterography and radiomics features extracted. Radiomics features were filtered by the gradient boosting decision tree (GBDT), and a radiomics score was calculated by using the radiomics signature-based formula. We constructed a clinical radiomics model and nomogram combining clinical factors and radiomics score through multivariate logistic regression analysis, and the internal validation was undertaken by ten-fold cross validation. Analyses of receiver operating characteristic (ROC) curves and decision curve analysis (DCA) were used to evaluate the prediction performance. DeLong test was applied to evaluate the performance of the clinical, radiomics and combined model. RESULTS: The clinical radiomics nomogram, which was based on the 9 radiomics signature and two clinical factors, indicated that the clinical radiomics model had an area under the ROC curve (AUC) value of 0.96 (95% confidence interval [CI]: 0.93-0.99) in the training cohort and 0.93 (95% CI: 0.86-1.00) in validation cohort. The clinical radiomics model was superior to the clinical model and radiomics model, and the difference was significant (P = 0.006, 0.004) in the training cohort. DCA confirmed the clinical utility of clinical radiomics nomogram. CONCLUSION: CTE-based radiomics model has a good performance in distinguishing CD from ITB. A nomogram constructed by combining radiomics and clinical factors can help clinicians accurately diagnose and select appropriate treatment strategies between CD and ITB.
PURPOSE: To establish a clinical radiomics nomogram to differentiate Crohn's disease (CD) from intestinal tuberculosis (ITB). PATIENTS AND METHODS: Ninety-three patients with CD and 67 patients with ITB were recruited (111 in training cohort and 49 in test cohort). The region of interest (ROI) for the lesions in the ileocecal region was delineated on computed tomography enterography and radiomics features extracted. Radiomics features were filtered by the gradient boosting decision tree (GBDT), and a radiomics score was calculated by using the radiomics signature-based formula. We constructed a clinical radiomics model and nomogram combining clinical factors and radiomics score through multivariate logistic regression analysis, and the internal validation was undertaken by ten-fold cross validation. Analyses of receiver operating characteristic (ROC) curves and decision curve analysis (DCA) were used to evaluate the prediction performance. DeLong test was applied to evaluate the performance of the clinical, radiomics and combined model. RESULTS: The clinical radiomics nomogram, which was based on the 9 radiomics signature and two clinical factors, indicated that the clinical radiomics model had an area under the ROC curve (AUC) value of 0.96 (95% confidence interval [CI]: 0.93-0.99) in the training cohort and 0.93 (95% CI: 0.86-1.00) in validation cohort. The clinical radiomics model was superior to the clinical model and radiomics model, and the difference was significant (P = 0.006, 0.004) in the training cohort. DCA confirmed the clinical utility of clinical radiomics nomogram. CONCLUSION: CTE-based radiomics model has a good performance in distinguishing CD from ITB. A nomogram constructed by combining radiomics and clinical factors can help clinicians accurately diagnose and select appropriate treatment strategies between CD and ITB.
Authors: A Dignass; G Van Assche; J O Lindsay; M Lémann; J Söderholm; J F Colombel; S Danese; A D'Hoore; M Gassull; F Gomollón; D W Hommes; P Michetti; C O'Morain; T Oresland; A Windsor; E F Stange; S P L Travis Journal: J Crohns Colitis Date: 2010-01-15 Impact factor: 9.071
Authors: R Caprilli; M A Gassull; J C Escher; G Moser; P Munkholm; A Forbes; D W Hommes; H Lochs; E Angelucci; A Cocco; B Vucelic; H Hildebrand; S Kolacek; L Riis; M Lukas; R de Franchis; M Hamilton; G Jantschek; P Michetti; C O'Morain; M M Anwar; J L Freitas; I A Mouzas; F Baert; R Mitchell; C J Hawkey Journal: Gut Date: 2006-03 Impact factor: 23.059
Authors: V Pratap Mouli; K Munot; A Ananthakrishnan; S Kedia; S Addagalla; S K Garg; J Benjamin; V Singla; R Dhingra; V Tiwari; S Bopanna; S Hutfless; G Makharia; V Ahuja Journal: Aliment Pharmacol Ther Date: 2016-11-04 Impact factor: 8.171