[Diagnostic Efficacy of an Ultrasound Quantitative Method in a Rat Model of Experimental Liver Fibrosis].

Ming-Li Chen,Yi Wang,Yong-Ming Yang,Qian-Miao Qin,Ye-Hua Cai,Guo-Hui Zhou
DOI: https://doi.org/10.3760/cma.j.issn.1007-3418.2012.02.011
2012-01-01
Abstract:To evaluate the efficacy of an ultrasound-based quantitative method to diagnose liver fibrosis using a rat model. Ultrasonography was performed on the livers of 90 Sprague-Dawley rats with or without thioacetamide-induced fibrosis. The liver capsule thickness and 13 texture parameters of gray level co-occurrence matrix were extracted from the standard sonograms. After sacrifice, severity of liver fibrosis (S0-S4 classification) was diagnosed by histopathology. Analysis of variance and correlation statistical tests were used to analyze the differences between groups and determine the relationships between each of the 14 quantitative ultrasound index points and the histological results, respectively. Discriminant analysis models were developed for quantitative diagnosis of liver fibrosis, and the leave-one-case-out method was used to verify the efficiency of models. All 14 indices were significantly correlated with the histological stages of fibrosis (P less than 0.05). The accuracy of the discriminant model for S0, S1, S2, S3 and S4 was 83.3%, 84.2%, 70.0%, 50.0% and 88.2%, respectively. In addition, 73.3% of cross-validated rats were accurately classified. Grouping S0 as no fibrosis, S1 as mild fibrosis, S2 with S3 as moderate to severe fibrosis and S4 as early cirrhosis increased the accuracy of the discriminant model for these four groups (respectively, 91.7%, 84.2%, 69.0% and 88.2%) and allowed for 78.9% of cross-validated rats to be correctly identified. Ultrasonography combined with texture analysis was a novel and accurate method to diagnose liver fibrosis in a rat model; further studies may provide insights into its applicability for quantitating liver fibrosis in other animal models or in clinic.
What problem does this paper attempt to address?