FoodCoach: Fully Automated Diet Counseling.

Abstract

Unhealthy dietary habits are a major preventable risk factor for widespread non-communicable diseases (NCD). Diet counseling is effective in managing diet-related NCDs, but constrained by its manual nature and limited (clinical) resources. To address these challenges, we propose a fully automated diet counseling system FoodCoach. It monitors people's food purchases using digital receipts from loyalty cards and provides structured dietary recommendations. We introduce the FoodCoach system's recommender algorithm and architecture, along with evaluation results from a two-arm randomized controlled trial involving 61 participants. The trial results demonstrate the technical feasibility and potential for scalable, fully automated diet counseling, despite not showing a significant change in participants' food purchase healthiness. We further show how others can deploy and extend the FoodCoach system in their own context and provide all relevant component implementations. Our core research contributions are: 1) a novel dietary recommendation algorithm designed and implemented with clinical experts, and 2) a scalable system architecture that employs a knowledge graph for enhanced interoperability and applicability to diverse domains and data sources. From a practical perspective, FoodCoach can augment traditional diet counseling through automatic diet monitoring and evaluation modules. Additionally, it streamlines the counseling process, conserving clinical resources, and ultimately contributing to the reduction of NCD prevalence.