AI plus digital pathology could help reveal how MASH progresses, speed promising treatments
Pairing AI with digital pathology could diagnose liver patients more accurately and speed promising treatments from the lab to patients more quickly, according to the director of the Stravitz-Sanyal Institute for Liver Disease and Metabolic Health.
Arun Sanyal, M.D., was part of an international team of researchers investigating a more detailed way to study metabolic dysfunction-associated steatohepatitis (MASH), with its characteristic scarring. Their goal is to help scientists better understand how MASH starts and progresses, and to make it easier to evaluate promising treatments.
MASH happens when fat builds up in the liver and the liver becomes inflamed and damaged. Eventually, without treatment, this damage can cause fibrosis, the tough scar tissue that interferes with liver functions. Doctors and scientists want to catch these changes earlier and understand what drives them. Only two drugs are approved in the United States to treat MASH, which affects an estimated 15 million Americans.
In a study published by Nature Communications, the team described how they used AI and digital pathology, to look at mouse liver tissue in a more “map it out” way than standard methods allow. Instead of just giving a general score, the AI divided the liver into different zones and identified where key signs of MASH appear. It showed when different features are near each other, which may help explain how the disease progresses.
“By pairing AI with digital pathology, we can read liver tissue with much more detail and speed than ever before, so we better understand how MASH develops, how therapies change what’s happening in specific tissue regions, and, ultimately, move the most promising treatments from the lab to patients faster,” Sanyal said.
To build a clear picture of how the mouse models develop MASH and fibrosis, the researchers used several methods. They checked blood chemistry results, reviewed traditional tissue staining under a microscope, and also performed tests that reflected changes in genes, lipids, and metabolites. Together, these tools helped confirm that the models were showing biologically relevant signs of MASH.
The AI found early events that help explain how inflammation and metabolism influence each other as the disease grows worse. The researchers also noticed specific characteristics of granulomas, which are immune cell clusters that can form in chronic inflammation, and found links between how those granulomas look and how much liver scarring develops. That suggests the immune response in the liver may be connected to scarring in a measurable way.
The team then tested the approach in treatment studies using drugs that have real clinical relevance, such as semaglutide and resmetirom, the only two drugs approved by the federal Food and Drug Administration to treat MASH. The AI showed that the drugs had different effects on fat buildup types in the liver and on fibrosis that appeared in the same areas as other MASH features. The AI revealed more than just whether the liver damage improved. It helped reveal what kind of improvement happened and where it happened in the tissue.
The team then used diet-based mouse models, following a diet tailored to push the liver toward MASH and scarring. One of the models used a diet high in fat, fructose and cholesterol to build up stress in the liver over time, leading to fat buildup, inflammation and gradual scarring.
The other diet affected how the liver handles and exports fats and affects cell energy and cell membranes. It causes liver injury and fibrosis much faster, often within weeks. Given that the two models work differently, these two models help scientists study both the slower, metabolism related path and the quicker injury and scarring path that can happen in people.
The researchers said that by using AI to analyze liver tissue by zones and by spatial relationships, scientists may be able to pick the most “fit for purpose” mouse models for testing specific therapies. That could lead to stronger studies and better odds that treatments that look promising in animals will be more likely to help people.