This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Animal models provide a useful platform for developing and testing new drugs to treat liver fibrosis. Accordingly, we developed a novel automated system to evaluate liver fibrosis in rodent models. This system uses second-harmonic generation (SHG)/two-photon excited fluorescence (TPEF) microscopy to assess a total of four mouse and rat models, using chemical treatment with either thioacetamide (TAA) or carbon tetrachloride (CCl), and a surgical method, bile duct ligation (BDL). The results obtained by the new technique were compared with that using Ishak fibrosis scores and two currently used quantitative methods for determining liver fibrosis: the collagen proportionate area (CPA) and measurement of hydroxyproline (HYP) content. We show that 11 shared morphological parameters faithfully recapitulate Ishak fibrosis scores in the models, with high area under the receiver operating characteristic (ROC) curve (AUC) performance. The AUC values of 11 shared parameters were greater than that of the CPA (TAA: 0.758-0.922 vs 0.752-0.908; BDL: 0.874-0.989 vs 0.678-0.966) in the TAA mice and BDL rat models and similar to that of the CPA in the TAA rat and CCl mouse models. Similarly, based on the trends in these parameters at different time points, 9, 10, 7, and 2 model-specific parameters were selected for the TAA rats, TAA mice, CCl mice, and BDL rats, respectively. These parameters identified differences among the time points in the four models, with high AUC accuracy, and the corresponding AUC values of these parameters were greater compared with those of the CPA in the TAA rat and mouse models (rats: 0.769-0.894 vs 0.64-0.799; mice: 0.87-0.93 vs 0.739-0.836) and similar to those of the CPA in the CCl mouse and BDL rat models. Similarly, the AUC values of 11 shared parameters and model-specific parameters were greater than those of HYP in the TAA rats, TAA mice, and CCl mouse models and were similar to those of HYP in the BDL rat models. The automated evaluation system, combined with 11 shared parameters and model-specific parameters, could specifically, accurately, and quantitatively stage liver fibrosis in animal models.
Background: Many clinical trials with potential drug treatment options for non-alcoholic fatty liver disease (NAFLD) are focused on patients with non-alcoholic steatohepatitis (NASH) stages 2 and 3 fibrosis. As the histological features differentiating stage 1 (F1) from stage 2 (F2) NASH fibrosis are subtle, some patients may be wrongly staged by the in-house pathologist and miss the opportunity for enrollment into clinical trials. We hypothesized that our refined artificial intelligence (AI)-based algorithm (qFibrosis) can identify these subtle differences and serve as an assistive tool for in-house pathologists. Methods: Liver tissue from 160 adult patients with biopsy-proven NASH from Singapore General Hospital (SGH) and Peking University People’s Hospital (PKUH) were used. A consensus read by two expert hepatopathologists was organized. The refined qFibrosis algorithm incorporated the creation of a periportal region that allowed for the increased detection of periportal fibrosis. Consequently, an additional 28 periportal parameters were added, and 28 pre-existing perisinusoidal parameters had altered definitions. Results: Twenty-eight parameters (20 periportal and 8 perisinusoidal) were significantly different between the F1 and F2 cases that prompted a change of stage after a careful consensus read. The discriminatory ability of these parameters was further demonstrated in a comparison between the true F1 and true F2 cases as 26 out of the 28 parameters showed significant differences. These 26 parameters constitute a novel sub-algorithm that could accurately stratify F1 and F2 cases. Conclusion: The refined qFibrosis algorithm incorporated 26 novel parameters that showed a good discriminatory ability for NASH fibrosis stage 1 and 2 cases, representing an invaluable assistive tool for in-house pathologists when screening patients for NASH clinical trials.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.