Related Experiment Video
Updated: Jul 29, 2025
08:58
Optimized Analysis of In Vivo and In Vitro Hepatic Steatosis
Published on: March 11, 2017
16.0K
Deep Learning-Based Image Analysis of Liver Steatosis in Mouse Models
Laura Mairinoja1, Hanna Heikelä1, Sami Blom2
1Research Centre for Integrative Physiology and Pharmacology, Institute of Biomedicine and Turku Center for Disease Modeling, University of Turku, Turku, Finland.
The American Journal of Pathology
|May 26, 2023
Summary
A new deep learning model accurately quantifies liver steatosis in mice using whole slide images. This AI tool aids in studying nonalcoholic fatty liver disease and evaluating drug efficacy in preclinical research.
Area of Science:
- Pathology
- Artificial Intelligence
- Preclinical Research
Background:
- Nonalcoholic fatty liver disease (NAFLD) is a growing global health concern, often linked with obesity.
- Efficient methods for studying NAFLD and assessing drug efficacy in preclinical models are crucial.
- Current quantification methods for liver steatosis can be labor-intensive and subjective.
Purpose of the Study:
- To develop and validate a deep neural network (DNN)-based model for quantifying microvesicular and macrovesicular steatosis.
- To utilize hematoxylin-eosin-stained whole slide images for automated liver steatosis analysis.
- To establish a reliable tool for large-scale preclinical studies of NAFLD.
Main Methods:
- A DNN model was developed using Aiforia Create for analyzing 101 whole slide images from mouse models.
- The algorithm was trained to identify liver parenchyma, exclude artifacts, differentiate steatosis types, and quantify affected areas.
- Image analysis results were compared with expert pathologist evaluations and ex vivo EchoMRI measurements.
Main Results:
- The DNN model accurately quantified microvesicular and macrovesicular steatosis.
- Image analysis results showed strong correlation with expert pathologist assessments.
- The model's quantification correlated well with liver fat content measured by EchoMRI and total liver triglycerides.
Conclusions:
- The developed deep learning model offers a novel and reliable tool for quantifying liver steatosis in mouse models.
- This AI-driven approach can significantly facilitate research in preclinical NAFLD studies.
- The model enables efficient and accurate assessment of steatosis in large cohorts, aiding drug development.
