Diabetasty: Personalized Recipe Generation System for Type 2 Diabetes Management Using Large Language Models and Reinforcement Learning
IEEE Access · 2026 · 14 · 168458-168477
DOI: 10.1109/access.2026.3748524 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Nitya Rajan und 4 weiteren Autor:innen. Die Arbeit nennt 47 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Vellore Institute of Technology, Vellore, Tamil Nadu, India.
Themen: Diabetes mellitus
Abstract (Originalsprache)
While it is possible to design a dietary approach for managing Type 2 diabetes (T2D), this is an incredibly difficult process due to the heterogeneity of patients and the scarcity of effective, individualized interventions. The present paper proposes a new generative AI model that aims at generating delicious and healthy diabetes-appropriate meals for patients in India. Given a set of ingredients, the application automatically generates recipes that meet most patient’s dietary restrictions while providing, at the same time, creativity and versatility. Overall, the method proposed here provides an efficient, bottom-up, and scalable approach for massive numbers of patients with Type 2 diabetes and for healthcare professionals who have to attend them. The trial runs show that with low glycemic index recipes, it is possible to create meaningful and coherent recipes inspired by diabetes severity levels. The system employs Large Language Models (LLMs) integrated with dynamic health railguards to generate recipes that the model predicted to have the lowest GI values possible accompanied by flavors and practicality. These rail guards use individual user data, with key health parameters of blood glucose levels, HbA1c, BMI, etc. For each user, a more basic Diabetes Severity Score, or DiSSco, is computed based on the level of dietary restriction required. The personalized approach comes closer to health profiles and proposes personalized recipes keeping in view clinic nutrition therapy and patients’ choice. As this method combines AI techniques with medical knowledge and advice concerning T2D, it will enable patients responsible for their diets and, therefore, increase the efficiency of their self-management. Such an interprofessional approach aims at increasing the quality of life for T2D patients by supplying them with simple, specific, delicious and nutritionally balanced dishes.
Förderung
- Vellore Institute of Technology, Vellore, Tamil Nadu, India
Zitieren
APA
Rajan, N., Changalasetty, S., Prabakaran, N., Karthik, N., & Devi, B. A. (2026). Diabetasty: Personalized Recipe Generation System for Type 2 Diabetes Management Using Large Language Models and Reinforcement Learning. IEEE Access, 14, 168458-168477. https://doi.org/10.1109/access.2026.3748524
Vancouver
Rajan N, Changalasetty S, Prabakaran N, Karthik N, Devi BA. Diabetasty: Personalized Recipe Generation System for Type 2 Diabetes Management Using Large Language Models and Reinforcement Learning. IEEE Access. 2026;14:168458-168477. doi:10.1109/access.2026.3748524
BibTeX
@article{Rajan2026,
author = {Rajan, Nitya and Changalasetty, Sreelasya and Prabakaran, N. and Karthik, N. and Devi, B. Aruna},
title = {Diabetasty: Personalized Recipe Generation System for Type 2 Diabetes Management Using Large Language Models and Reinforcement Learning},
journal = {IEEE Access},
year = {2026},
volume = {14},
pages = {168458--168477},
doi = {10.1109/access.2026.3748524}
}
EUVIMED