AI powered dietary proportion assessment for improving accuracy and practicality of the balanced meal plate model

The 2:1:1 dietary proportion model, promoting balanced eating, faces challenges in practical application due to difficulty in visually estimating portion sizes. To address this, we developed an artificial intelligence (AI)-based application to assist in the accurate assessment of 2:1:1 dietary proportion. This study demonstrated the accuracy of the AI application compared to estimates made by nutrition and dietetics students (ND) and registered dietitians (RD), while also assessing user satisfaction and attitudes. The AI application was trained using images of three popular Thai dishes: Hainanese Chicken Rice, Shrimp Paste Fried Rice, and Egg Noodle, each prepared in three portion variations. The AI system demonstrated significantly lower mean absolute error (MAE) than both ND and RD groups in estimating proportions for Hainanese Chicken Rice and Shrimp Paste Fried Rice (p < 0.05), indicating superior accuracy. User satisfaction surveys revealed that 61% of participants rated their overall experience with the application as moderate, with suggestions for improving accuracy. Over half agreed that the AI tool shows potential as a practical resource for nutrition education and dietary assessments. The AI holds promise as a tool for promoting adherence to the 2:1:1 model and facilitating healthy lifestyle changes.