Paper ID:250655XP
Authos: Anand K. Gavai and Jos van Hillegersberg
Title: Personalized and Sustainable Smoothie Generation for Metabolic Disease Management
Publisher: Algorithm_Lab.
Conference: ICAITD 2025: The Second International Conference of AI new Technology and open Discussion
Location: Pleasanter Lounge Nakano, Tokyo Japan
Date: 1-4 june 2025
Editor: Kazuo Ohzeki (Algorithm_Lab. Professor Emeritus of Shibaura Institute of Technology, Professor of Emeritus of International Professional University of Technology in Tokyo)
Citation: Proceedings: ICAITD 2025
https://doi.org/10.63211/j.p.25.645341
pages: 83-85
Abstract:
While AI has shown promise in dietary recommendations, existing systems often
lack personalization and evidence-based validation for chronic disease management. This
study presents an innovative AI-driven dietary recommendation system that leverages
Retrieval Augmented Generation (RAG) to create customized smoothie recipes for people
with obesity and type 2 diabetes. The system uniquely combines the generative capabilities of large language models (LLaMA3) with real-time data integration through the USDA's
FoodData Central API, while incorporating the Dutch dietary guidelines (RIVM) and the
European Food Information Council's (EUFIC) seasonal food database. Our approach implements a robust multi-stage algorithmic validation pipeline: (1) automated nutritional compliance verification against RIVM guidelines for macronutrient balance and glycemic control, with real-time nutrient calculations via USDA's API, (2) ingredient compatibility and seasonal availability assessment using EUFIC's European seasonal food database, and (3) sustainability scoring based on local sourcing and environmental impact metrics. A custom validation algorithm ensures each generated recipe strictly adheres to RIVM's dietary guidelines, automatically flagging and filtering out non-compliant recipes based on predefined nutritional thresholds and constraints. Each recipe generation is accompanied by an explainable AI component that provides detailed reasoning for ingredientselection, highlighting specific health benefits and nutritional contributions of each component to the overall therapeutic goals.
The system's web application interface facilitates personalization through user inputs including age, BMI, dietary preferences, and health conditions, generating real-time recipe recommendations with detailed nutritional breakdowns. Each recommendation includes an interactive explanation module that helps users understand why specific ingredients were chosen, their nutritional impact, and how they align with the user's health goals. For instance, the system might explain how a particular berry variety was selected for its low glycemic index and high antioxidant content, or why certain ingredient combinations enhance nutrient absorption. The RAG architecture, coupled with real-time USDA data access and algorithmic validation, ensures that every generated recipe meets strict nutritional and therapeutic requirements.
This novel integration of explainable AI-driven personalization with automated guideline compliance verification and multiple evidence-based sources (RIVM guidelines, USDA API, and EUFIC seasonality information) represents a significant advance in dietary management tools for chronic diseases, offering healthcare providers and patients a practical, scientifically-grounded, and transparent approach to nutritional therapy
Keywords: Keywords: RAG, Personalized Nutrition, Obesity & Type 2 Diabetes, Health Literacy, Food Sustainability
Category: Poster Abstract
Review process: Committee review
Publication date: July 2nd 2025
First received date: Oct. 30th 2024
Copyright :Author(Full), AlgorithmLab.(First-in-the-world publishing rights granted by the author as the Prceedings of ICAITD 2025)
Licence:Viewer can download and view this review paper, but cannot secondary distribute (redistribute) it. (It is not Creative Commons License, nor MIT Licence) In other words, "Do not distribute" is the License
[] reads