Ting Zhou, Chao-Lian Xie1, Yong Chen2, Yan Deng3, Jia-Long Wu3, Rui Liang3, Guo-Dong Yang4, Xiao-Ming Zhang3. 1. From the Department of Radiology, Sichuan Cancer Hospital and Institute, Sichuan Cancer Center, School of Medicine, University of Electronic Science and Technology of China, Chengdu. 2. Department of Radiology, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai. 3. Sichuan Key Laboratory of Medical Imaging and Department of Radiology, Affiliated Hospital of North Sichuan Medical College, Nanchong. 4. Department of Gastroenterology, Affiliated Hospital of North Sichuan Medical College, Nanchong, China.
Abstract
OBJECTIVE: The aim of the study was to investigate radiomics models based on magnetic resonance imaging (MRI) for predicting early extrapancreatic necrosis (EXPN) in acute pancreatitis. METHODS: Radiomics features were extracted from T2-weighted images of extrapancreatic collections and late arterial-phase images of the pancreatic parenchyma for 135 enrolled patients (94 in the primary cohort, including 47 EXPN patients and 41 in the validation cohort, including 20 EXPN patients). The optimal features after dimension reduction were used for radiomics modeling through a support vector machine. A clinical model, the MR severity index score, and extrapancreatic inflammation on MRI were evaluated. RESULTS: Twelve optimal features from the extrapancreatic collection images and 10 from the pancreatic parenchyma images were selected for modeling. The pancreatic parenchyma-based and extrapancreatic collection-based radiomics models showed good predictive accuracy in both the training and validation cohorts. The areas under the curve of the extrapancreatic collection-based radiomics model (0.969 and 0.976) were consistent with those of the pancreatic parenchyma-based model (0.931 and 0.921) for both cohorts and better than those of the clinical model and imaging scores for both cohorts. CONCLUSIONS: The MRI-based radiomics models of both the extrapancreatic collections and the pancreatic parenchyma had excellent predictive performance for early EXPN.
OBJECTIVE: The aim of the study was to investigate radiomics models based on magnetic resonance imaging (MRI) for predicting early extrapancreatic necrosis (EXPN) in acute pancreatitis. METHODS: Radiomics features were extracted from T2-weighted images of extrapancreatic collections and late arterial-phase images of the pancreatic parenchyma for 135 enrolled patients (94 in the primary cohort, including 47 EXPN patients and 41 in the validation cohort, including 20 EXPN patients). The optimal features after dimension reduction were used for radiomics modeling through a support vector machine. A clinical model, the MR severity index score, and extrapancreatic inflammation on MRI were evaluated. RESULTS: Twelve optimal features from the extrapancreatic collection images and 10 from the pancreatic parenchyma images were selected for modeling. The pancreatic parenchyma-based and extrapancreatic collection-based radiomics models showed good predictive accuracy in both the training and validation cohorts. The areas under the curve of the extrapancreatic collection-based radiomics model (0.969 and 0.976) were consistent with those of the pancreatic parenchyma-based model (0.931 and 0.921) for both cohorts and better than those of the clinical model and imaging scores for both cohorts. CONCLUSIONS: The MRI-based radiomics models of both the extrapancreatic collections and the pancreatic parenchyma had excellent predictive performance for early EXPN.