Researchers Develop Novel AI Model for Diabetes Management

Strategic Insights -

A team of researchers from North Dakota State University has made significant strides in developing a novel artificial intelligence (AI) model for enhancing diabetes management with privacy-preserving edge AI. According to the study, the increasing prevalence of diabetes necessitates innovative glucose prediction methods that prioritize patient privacy. The researchers introduced a clustering-based federated deep learning (Clu-FDL) model to address the challenges of patient variability and optimizing federated learning for glucose prediction. The Clu-FDL approach achieves high precision, recall, and F1 scores, along with low Root Mean Square Error (RMSE) values, making it a promising solution for diabetes management.

Key Takeaways:

  • The Clu-FDL model is a novel AI approach that prioritizes patient privacy for diabetes management.
  • The model uses federated learning to enhance prediction accuracy by clustering patients based on carbohydrate intake patterns.
  • The study evaluates the performance of local patients who contribute to training the cluster and global (non-cluster) models.
  • The Clu-FDL approach achieves high precision (0.93), recall (0.96), and F1 scores (0.95), along with low Root Mean Square Error (RMSE) values (11.08 ± 1.77 mg/dL).
  • The model exhibits greater stability for new patients with different data durations, with smaller RMSE and higher precision, recall, and F1 scores compared to non-clustering models.
  • SimpleRNN and GRU models are most effective for new patients with 9 and 6 days of data, respectively.
  • The Clu-FDL approach empowers patients to manage their diabetes effectively with a privacy-preserving, clustering-based personalized approach.

Statistics:

  • The Clu-FDL approach achieves high precision (0.93), recall (0.96), and F1 scores (0.95).
  • The model exhibits low Root Mean Square Error (RMSE) values of 11.08 ± 1.77 mg/dL.
  • The study evaluates the performance of local patients who contribute to training the cluster and global (non-cluster) models.
  • The Clu-FDL approach achieves greater stability for new patients with different data durations, with smaller RMSE and higher precision, recall, and F1 scores compared to non-clustering models.
  • The model is most effective for new patients with 9 and 6 days of data using SimpleRNN and GRU models, respectively.

Sources:

  • A clustering-based federated deep learning approach for enhancing diabetes management with privacy-preserving edge artificial intelligence. Healthcare Analytics, 2025,7():100392.
  • NewsRx. Research Reports from North Dakota State University Provide New Insights into Artificial Intelligence (A clustering-based federated deep learning approach for enhancing diabetes management with privacy-preserving edge artificial intelligence). Diabetes Week. June 30, 2025; p 194.