Validation of an Off-the-Shelf, Diet-Induced NASH Mouse Model using Digital Whole Slide Scanning of Liver Tissue and Artificial Intelligence-Enabled, Quantitative Histopathological Analysis
Summary:
Non-alcoholic fatty liver disease (NAFLD) and non-alcoholic steatohepatitis (NASH) are common chronic liver disorders resulting in lipid accumulation in the liver, inflammation, hepatocyte damage, and fibrotic scarring, all of which impact organ function. With the lack of approved therapies for NASH patients, and considerable inter pathologist variability in the scoring and semi-quantitative assessment of the pathology used in their diagnoses, much focus and effort is being placed in the development of rodent models, the assessment of individual disease features in these models, and understanding the effect of therapeutics on the disease process. The study described in this paper is the result of a unique collaboration between several companies with combined goals of providing the NASH research community with a time-efficient, flexible, off-the-shelf diet-induced mouse NASH B6 Model (Taconic Biosciences, Inc.) that can be utilized in directed research studies for therapeutic drug development (Explora BioLabs). The results were digitally scanned using a 3DHISTECH PANNORAMIC™ series scanner (Epredia) and validated with imageDx NASH, a digital assay platform featuring machine and deep learning models to quantify steatosis, inflammation, hepatocellular ballooning, and fibrosis (Reveal Biosciences, Inc). AI-generated quantitative results were compared with qualitative NAS scores provided by a board-certified veterinary pathologist to evaluate grade and fibrosis stage. The AI-based digital analysis results of the treated and untreated diet-induced NASH B6 model correlated with the CRN scores recorded by the experienced veterinary pathologist. Importantly, the quantitative data also demonstrated measurement of more granular pathologic changes. These subtle – but likely significant – pathologic changes are not currently well-captured by the standard scoring system, and underscore the need for and the utility for sensitive and reproducible quantitative analysis.
