Researchers from several partner institutions of the German Center for Diabetes Research (DZD) have collaborated with international colleagues to develop a new approach for visualizing subtle tissue changes in the pancreas in type 2 diabetes. The results provide new insights into the development of type 2 diabetes. The study has now been published in Nature Communications .
More than 500 million people worldwide live with type 2 diabetes. They often suffer from serious complications. Nevertheless, it has been difficult to draw reliable conclusions about a person's glycemic status based on classic histopathological examinations. Many subtle morphological changes associated with impaired insulin secretion and beta cell dysfunction are barely visible to the naked eye.
Extensive data set
To close this diagnostic gap, the research team created an extensive data set from pancreatic tissue sections from living donors. The samples were contrasted using chromogenic and multiplex immunofluorescent staining and then captured in high resolution using gigapixel microscopy.
Deep learning models distinguish between tissue samples from people with and without type 2 diabetes
On this basis, the scientists trained deep learning models that could reliably distinguish between tissue samples from people with and without type 2 diabetes.
The models were able to accurately predict diabetes status and, for the first time, identify which tissue structures play a central role in the disease—including changes in the islets of Langerhans, α cells, neuronal axons, and the proximity of fat cell clusters to islet structures. Using explainable AI, the identified features were analyzed, quantified, and described as potential biomarkers.
This AI-supported evaluation provides new insights into early and previously difficult-to-detect changes in type 2 diabetes. The findings open up new perspectives for understanding the crucial processes that take place in the pancreas during the development of type 2 diabetes.
Original Publication:
Klein, Ziegler, Gerst, Morgenroth… Solimena & Wagner: Explainable AI-based analysis of human pancreas sections identifies traits of type 2 diabetes. Nat Commun 17, 1558 (2026). DOI:
Nature Communications
Data/statistical analysis
People
Explainable AI-based analysis of human pancreas sections identifies traits of type 2 diabetes.
9-Feb-2026