Jeonbuk National University researchers develop an AI model for personalized blood glucose monitoring
The hybrid model integrates three components to address key challenges and was rigorously evaluated, paving way for accurate blood glucose prediction
image: BiT-MAML adopts meta-learning to address inter-patient variability and a hybrid architecture to capture both short-term and long-term patterns in BG levels. The proposed evaluation scheme shows a new approach for predicting BG levels in patients while accounting for inter-patient variability.
Credit: Professor Jaehyuk Cho from Jeonbuk National University, Korea
Type 1 diabetes (T1D) is an autoimmune condition in which the body’s own immune system attacks insulin-producing cells. As a result, patients with T1D must closely monitor their blood glucose (BG) levels and rely on insulin injections or pumps. Even small miscalculations or oversights can lead to unregulated blood sugar levels, leading to potentially life-threatening complications.
Continuous glucose monitoring (CGM) systems have emerged as a promising tool for predicting and forecasting BG levels. Over the past decade, researchers have explored artificial intelligence (AI) models for improving the prediction accuracy of CGM systems. However, differences in physiology between patients and poor adaptation for new users persist to challenge the widespread adoption of this technology in real-world settings. In addition, traditional models often focus on either short-term or long-term glucose patterns, but not both.
In an attempt to address these issues, a research team led by Professor Jaehyuk Cho from the Department of Software Engineering at Jeonbuk National University in South Korea, have developed an innovative model, named BiT-MAML, aimed at tackling inter-patient variability in BG prediction. Explaining further, Prof. Cho says, “BG dynamics are not uniform across all patients. The physiological patterns of an elderly patient are vastly different from those of a young adult.” Adding further, he says, “Our model demonstrates how this variability can be accounted for by developing more personalized models.” Their findings were published in Scientific Reports on August 20, 2025.
BiT-MAML (where “BiT-“ stands for Bidirectional LSTM-Transformer” and “MAML” stands for “Model-Agnostic Meta-Learning”) uses hybrid architecture combining two deep learning models: bidirectional long-short-term memory (Bi-LSTM) and Transformer. Bi-LSTM processes time-series BG data bidirectionally, precisely capturing short-term patterns. Simultaneously, the transformer, utilizing a multi-head attention approach, efficiently models long-term patterns, capturing complex day-to-day and lifestyle-based cyclical variations. During training, the researchers applied a meta-learning approach known as Model-Agnostic Meta-Learning (MAML) that helps the model quickly adapt to new and diverse patients using only a small amount of training data by learning from a wide range of patient examples.
To test model performance, the researchers adopted a Leave-One-Patient-Out Cross-Validation (LOPO-CV) scheme. “In simple terms, we train the AI on five patients, then test it on the sixth patient it has never seen before,” explains Prof. Cho. “This is effective for assessing the model’s ability to generalize to unseen patients.”
The model demonstrated significantly reduced prediction error compared to conventional models. Notably, the prediction error varied from an excellent 19.64 milligram/decilitre (mg/dL) for one patient to a challenging 30.57 mg/dL for another. While these results represent a clear improvement over the standard LSTM models, they also highlight the persistent difficulty of managing inter-patient variability in real-world settings. “Our study shows how AI-based BG prediction models should be evaluated to improve both trust and model performance,” concludes Prof. Cho. “Addressing this challenge will contribute to the development of effective CGM models that can serve diverse patients with T1D, from children to elderly.”
These findings attest to the fact that the development of effective personalized BG prediction requires the use of advanced AI models incorporating robust evaluation methods that can transparently report the full spectrum of performance.
***
Reference
DOI: 10.1038/s41598-025-13491-5
About Jeonbuk National University
Founded in 1947, Jeonbuk National University (JBNU) is a leading Korean flagship university. Located in Jeonju, a city where tradition lives on, the campus embodies an open academic community that harmonizes Korean heritage with a spirit of innovation. Declaring the “On AI Era,” JBNU is at the forefront of digital transformation through AI-driven education, research, and administration. JBNU leads the Physical AI Demonstration Project valued at around $1 billion and spearheads national innovation initiatives such as RISE (Regional Innovation for Startup and Education) and the Global University 30, advancing as a global hub of AI innovation.
Website: https://www.jbnu.ac.kr/en/index.do
About the Author
Jaehyuk Cho is a Professor in the Department of Software Engineering at Jeonbuk National University and Director of the Adaptive AI Laboratory. He also serves as CEO of Human AI Plus, a university start-up translating research into practical healthcare tools. His research focuses on medical AI, healthcare data analytics, and physical AI convergence, with particular emphasis on developing AI systems that are not only accurate but genuinely usable in real clinical environments. His group is committed to bridging the gap between AI research and patient-centered healthcare.