From the Journals

AI Model Finds Hidden Risk Signals in CGM Data

GluFormer accurately identified patients likely to experience clinically meaningful rises in HbA1c over a 2-year period.

Share

A study published in Nature reveals that the generative AI model GluFormer, trained on over 10 million continuous glucose monitoring (CGM) measurements, can significantly enhance prognostic analysis of glucose data compared to traditional methods like HbA1c. Developed by Guy Lutsker and colleagues at the Weizmann Institute, GluFormer uses advanced self-supervised learning to predict glucose dynamics across diverse populations. It notably improves the early identification of prediabetes and long-term risks of diabetes and cardiovascular diseases by analyzing complex glucose patterns.

Original Source(s)

AI Model Finds Hidden Risk Signals in CGM Data

GluFormer accurately identified patients likely to experience clinically meaningful rises in HbA1c over a 2-year period.

  • by Alun Evans

  • March 5, 2026

  • 3 min

Related Content

Diabetes Risk Assessment: A New Approach

Blood-based epigenetic markers show strong prognostic potential for Type 2 diabetes

AI's Expanding Role in Diabetes Care

Machine learning models applied to inpatient electronic health record data have predicted insulin needs more accurately than traditional guidelines.

Stratifying Prediabetes by DNA

Blood-based methylation profiles reproduced high-risk clusters without complex metabolic testing.