Methodological aspects of using artificial intelligence for morphological diagnosis of fibrosis, degeneration, and inflammatory lesions of the liver

Non-tumor liver diseases are widespread and remain difficult to diagnose. According to current data, the prevalence of non-alcoholic fatty liver disease among adults in Russia is approximately 25%. Morphological confirmation of fibrosis, fatty and ballooning degeneration, inflammatory infiltration, and necrosis of liver tissue depends on the specialist’s subjective opinion, which complicates standardization. For these reasons, developing objective and automated methods for analyzing morphological changes in the liver can significantly improve diagnostic reproducibility. This review examines current approaches to using artificial intelligence in the morphological diagnosis of non-tumor liver diseases, as well as the key applications for neural network algorithms, including classification and segmentation of histological images. Furthermore, the review assesses the effectiveness of available models for detecting key morphological patterns: fibrosis, ballooning and fatty degeneration, and inflammatory infiltration.

The review includes publications found in the Google Scholar and PubMed databases. The search covered the period from 2020 to 2025, with 22 publications included in the final analysis.

It was found that artificial intelligence models demonstrate high accuracy; however, this depends on sample size, inter-laboratory variability, morphological patterns, microscope magnification, and staining methods. More open data and standardized procedures are needed for future advancement in this field. Nevertheless, models are being developed even with small datasets, making the methodology available to the scientific community.