📝 SummarXiv

Abstract

Objectives: The role of advanced diffusion-weighted imaging (DWI) in chronic liver disease (CLD) has not been fully studied. Chronic liver disease (CLD) is a progressive deterioration of liver functions, caused by one or more etiology. This study was aimed to investigate whether radiomics features extracted from individual or combined magnetic resonance imaging sequences, such as T1-weighted, T2-weighted images, or quantitative maps from chemical shift encoded, diffusion-weighted imaging, can effectively classify inflammation and fibrosis in CLD. Method: Seventy-seven patients with CLD were enrolled in this study. Each participant underwent both MRI examinations and liver biopsy. The biopsy procedure was applied to quantitatively or semi-qualitatively analyze several histology features, steatosis, inflammation, and fibrosis. Radiomic features were extracted, selected, reduced, and used to train the inflammation and fibrosis classification based on random forest models. The performances of classifiers were evaluated by the receiver operating characteristic curve (ROC), accuracy, precision, sensitivity, specificity, and DeLong tests. Result: The random forest model achieved the area under the curve (AUC) of 0.85 and 0.86 for inflammation and fibrosis classification, respectively. Conclusion: This study demonstrated that the MRI-based radiomics features hold potential in the inflammation and fibrosis classification.