Shengfeng Yu*, Xu Gao, Dong Hu
Department of Imaging, Taihe County People's Hospital, No. 21, Jiankang Road, Taihe, Fuyang, Anhui 236600, China
Introduction: To clarify the value of enhanced CT radiomics in diagnosing acute pancreatitis (AP) complicated with hepatic injury.
Materials and methods: CT enhanced images of 76 AP patients were retrospectively analyzed. The above patients were divided into two groups according to whether they had hepatic injury or not, including 42 cases in hepatic injury group and 34 cases in non-hepatic injury group. Region of interest (ROI) was manually delineated on patients' enhanced CT portal venous phase images using ITK-Snap software, and radiomics features in ROI images were extracted. The data were randomly divided into training set and validation set according to ratio of 7:3. The radiomic features were chosen in extracted features using Max-Relevance and Min-Redundancy (mRMR) combined with multivariate logistic regression to establish a radiomics model, and area under curve (AUC) value and Youden index evaluated the performance of model. The clinical efficacy of radiomics scoring model was assessed using decision curves.
Results: After feature screening, 7 optimal features were obtained, and the selected feature parameters were weighted and summed to establish a radiomics model. The AUC values of model in training set and validation set were 0.91 and 0.89, respectively; sensitivity and specificity in training set were 95.4% and 98.3%, and in validation set were 95.2% and 95.7%. Decision curves demonstrated that radiomics models all had a higher net benefit when high-risk threshold probability was in 0-1.
Conclusion: Enhanced CT image-based radiomics feature analysis can effectively diagnose AP complicated with hepatic injury.
Acute pancreatitis, hepatic injury, CT, radiomics.