Background
Sjögren’s disease is a chronic autoimmune condition that mainly damages glands that produce moisture, leading to common symptoms like dry eyes and dry mouth. However, in about 40–50% of people, the disease can also affect other important organs such as the lungs, kidneys, blood, or brain. Detecting which patients are more likely to develop organ involvement early on is important because it allows doctors to monitor these patients more closely and start treatment before serious complications occur.
Study Design
Researchers studied 221 people with Sjögren’s disease who had lip biopsies (small tissue samples taken from the inside of the lower lip) when they first came to the hospital. The study compared two groups:
- Organ involvement group: Patients whose disease had already affected at least one vital organ (lung, kidney, blood, or brain).
- Gland-only group: Patients whose disease was limited to gland problems like dry mouth and dry eyes.
The research team used whole-slide digital images of the biopsy samples and combined them with information on whether each patient had organ involvement.
The AI Model
The team developed an artificial intelligence (AI) system designed to “look” at the biopsy images in great detail—far beyond what the human eye can see. This was done using advanced image analysis techniques called pattern recognition and multi-instance learning. In simple terms, the AI learned to recognize visual patterns in the tissue, such as how many immune cells were present, whether there was scarring (fibrosis), and how much of the normal tissue structure had been damaged. The model then used this knowledge to predict whether a patient was likely to have organ involvement. The predictions were tested on two separate sets of patient data to check how well the model worked.
Key Findings
The AI model showed high accuracy in identifying patients with organ involvement in both the original group of patients and in a separate validation group. It was able to pick up microscopic patterns in the biopsy tissue—like dense clusters of immune cells, damaged gland structures, and tissue scarring—that were linked to a higher risk of organ involvement.
Clinical Significance
This study shows that AI can be used to gain valuable information from a routine biopsy that doctors might not normally detect. By combining AI analysis with existing diagnostic methods such as lip biopsy, doctors could better identify patients who need closer monitoring and more personalized treatment plans. This could lead to faster intervention, fewer missed cases of organ involvement, and better long-term health outcomes.
Future Directions
The authors suggest testing the AI tool in larger, more diverse groups of patients should be the next step. They also propose combining the image-based findings with other patient information, such as lab tests and symptoms, to improve accuracy. Future research should also explore whether AI models can predict organ involvement before it develops, which could open the door for preventive care.
Reference:
Ren, Y., Xia, W., Wu, J. et al. Artificial intelligence-based prediction of organ involvement in Sjogren’s syndrome using labial gland biopsy whole-slide images. Clin Rheumatol 44, 2919–2927 (2025). https://doi.org/10.1007/s10067-025-07518-5
