Work

AI-Based tool for dietary data collection: design and technical validation. The ALEVINT Project

Year: 2026

Type: article

Abstract: Background Dietary data collection and coding are complex, time-consuming tasks requiring trained personnel, limiting their use in public health and clinical settings. This affects dietary exposure estimation and its application in health outcome analysis. Within the ALEVINT project, we developed an AI-based tool to collect, codify, and evaluate dietary exposure through an efficient, flexible, and standardized approach aimed at improving data quality and reducing measurement bias. Methods An AI-assisted dietary intake system was developed with three modules: 1. conversational data collection, 2. transcription processing and dietary coding, 3. linkage with food composition databases. Data are collected via a Telegram chatbot driven by a large language model (ChatGPT) accessed through an API. The system was implemented in R on an R-Shiny server and integrated into the ANĺMATE app, an mHealth platform for clinical trials connected with REDCap. The system can be adapted to different objectives, languages, and cultural contexts through prompt-based configuration without additional programming. Results A functional “virtual dietitian” for 24-h dietary recalls (extendable to FFQs) was developed, enabling automated, standardized data collection with direct coding into foods and nutrients. Preliminary technical validation in a small user group showed flexibility in capturing dishes, ingredients, portion sizes, brands, timing, and habits, with good cultural and linguistic adaptability. Initial findings indicate reduced workload, improved standardization, and good acceptability, with adequate consistency and reproducibility. Conclusions The ALEVINT chatbot is an innovative AI-based tool for dietary data collection and processing. It has the potential to improve efficiency, speed, and standardization in public health nutrition research, while reducing costs. Initial results suggest technical feasibility and it is currently undergoing real-world validation against trained dietitians. Key messages • AI-powered “virtual dietitian” enables automated, standardized collection, coding, and evaluation of dietary data, reducing workload and improving data quality in nutrition research. • The ALEVINT AI chatbot integrates large language model (LLM) systems to streamline dietary data collection and coding, with high flexibility and scalability.

Source: European Journal of Public Health

DOI: 10.1093/eurpub/ckag169.1213