Project information
- Category: Federated Nonalcoholic Steatohepatitis Staging
- Data: Publically Available Datasets Published in Nature
- Project date: June, 2024
- Project URL: https://doi.org/10.1002/ima.23118
Classification of the Stages of Nonalcoholic Steatohepatitis via Federated General Visual Representation Learning
Nonalcoholic fatty liver disease (NAFLD), characterized by excess liver fat without alcohol consumption, requires accurate subtype classification, especially for nonalcoholic steatohepatitis (NASH), to guide patient management and treatment decisions. This study demonstrates the potential of deep learning, specifically the federated version of the Big Transfer (BiT) model, for precise NAFLD/NASH classification using histopathological imaging. The model achieved high AUC scores of 0.9299 and 0.9582 for "main-class" and "multi-class" scenarios, with subclass AUCs ranging from 0.9376 to 0.9867 for ballooning, inflammation, steatosis, and fibrosis. Corresponding accuracy results were also strong, with subclass ACC scores of up to 0.9203, highlighting the effectiveness of federated learning in this medical domain.