Benjamin Yin¹ and Marisabel Chang², ¹USA, ²California State Polytechnic University, USA

Diabetes affects homeless populations at rates similar to the general population (8%), but homeless individuals receive significantly less medical attention and face higher complication rates due to food insecurity and lifestyle instability. VitalityShield addresses this challenge through a Flutter-based mobile application that provides AI-powered food analysis and personalized diabetes prevention recommendations. The system integrates three core components: an OpenAI GPT-4 Vision food scanner for nutritional analysis, an AI recommendation service using GPT-4o-mini for personalized dietary suggestions, and interactive health analytics for progress tracking [1]. Key challenges included achieving accurate food recognition across varying image qualities and generating practical recommendations for populations with limited food access. Experimental results demonstrated 83.25% accuracy in nutritional analysis and moderate practicality scores (3.4/5) for recommendations. While limitations exist in accessibility and accuracy, VitalityShield offers significant advantages over traditional outreach methods by providing scalable, 24/7 diabetes prevention support that adapts to individual dietary patterns, potentially reducing diabetes risk in vulnerable homeless communities.

Diabetes Prevention, Homeless Populations, Artificial Intelligence, Computer Vision, Mobile Health Applications