NIT Rourkela develops AI-driven model for improved diabetes management

New Delhi, Feb 26 – Researchers at the National Institute of Technology (NIT), Rourkela, have introduced a novel AI-driven model to improve blood sugar level predictions, aiming to enhance diabetes management for patients and healthcare professionals. This machine-learning approach is designed to deliver more accurate glucose forecasts, supporting personalized treatment decisions.

The study, published in the IEEE Journal of Biomedical and Health Informatics, marks a significant advancement in predictive healthcare. The research team is now preparing for clinical trials in hospitals to assess real-world applications.

Addressing India’s Growing Diabetes Crisis​

India faces a growing diabetes burden, with cases projected to reach 124.9 million by 2045, according to Mirza Khalid Baig, Assistant Professor of Biotechnology and Medical Engineering at NIT Rourkela. Effective diabetes management relies on continuous glucose monitoring to prevent severe fluctuations—hyperglycemia (high blood sugar) and hypoglycemia (low blood sugar)—which can lead to critical health complications.

Baig highlighted key challenges in diabetes care, including limited specialist availability, unequal access to healthcare, low medication adherence, and poor self-care practices. He emphasized that AI-driven digital health solutions offer a promising approach to improving diabetes management while reducing costs.

Overcoming Limitations of Traditional AI Models​

Deepjyoti Kalita, research scholar and co-author of the study, noted that while machine learning (ML) techniqueshave been widely applied in diabetes research, existing predictive AI models have notable drawbacks. Many function as a “black box”, making their predictions difficult for doctors and patients to interpret, thus reducing trust in their reliability.

Additionally, traditional forecasting models, such as statistical methods and basic neural networks, struggle to capture long-term glucose fluctuations and require intricate fine-tuning, making them less effective in real-world applications.

Advancing Blood Glucose Prediction with Deep Learning​

The NIT Rourkela research team developed an advanced deep learning model that learns from past blood sugar trends, ensuring more accurate predictions than conventional methods. This specialized AI model eliminates the need for manual adjustments by automatically processing glucose data and identifying significant patterns.

Baig explained that their core innovation lies in multi-head attention layers within a neural basis expansion network. This technique allows the model to focus on the most relevant data while filtering out unnecessary noise, leading to higher accuracy with minimal training data and computational resources.

Future Applications in Digital Healthcare​

The AI-driven model has wide-ranging applications in diabetes management. According to researchers, it could be integrated into:
Smart insulin pumps to automate insulin delivery.
Mobile health apps for real-time blood sugar monitoring.
Clinical settings to assist doctors in formulating personalized treatment plans.

By combining precision with efficiency, this AI innovation aims to become a practical tool within digital health solutions, empowering both patients and healthcare providers to manage diabetes more effectively.

With clinical trials on the horizon, this development could revolutionize diabetes care, offering more reliable, automated, and personalized glucose management solutions.