Application of MRI-Based Radiomics in the Diagnosis and Categorization of Liver Fibrosis
Objective:To explore the value of an imaging-based radiomics model constructed by consecutively delineating three layers of regions of interest (ROIs) based on T1-weighted imaging (T1WI) for the classification and prediction of liver fibrosis. Methods:Imaging data from 20 patients with liver fibrosis were selected,including 12 with non-significant fibrosis and 8 with significant fibrosis. The patients were randomly divided into a training group (14 cases) and a test group (6 cases) in a 7∶3 ratio. With the aid of image edge detection technology and manual delineation,ROIs were outlined in three consecutive layers to extract radiomic features. Through feature selection and dimensionality reduction,radiomics models based on machine learning algorithms,namely Support Vector Machine with Radial Basis Function kernel (SVM_RBF) and Random Forest (RF),were constructed. The predictive performance of both models was evaluated for distinguishing between non-significant and significant liver fibrosis,and validation was conducted on the test group. Results:The radiomics models constructed using SVM_RBF and RF achieved areas under the curve (AUC) of 0.84 and 0.85,respectively,in the training group. In the test group,the AUCs were 0.87 and 0.69,respectively,with the SVM_RBF-based model performing optimally. Conclusion:The combination of MRI-based radiomic features from three consecutive layers of ROIs with machine learning classification models can predict the presence of non-significant versus significant liver fibrosis.
万方数据
维普
