A feasibility study to assess Mediterranean Diet adherence using an AI-powered system

Area of Science:

• Nutrition Science
• Artificial Intelligence
• Digital Health

Background:

• The Mediterranean diet (MD) is linked to reduced non-communicable disease risk and obesity prevention.
• Accurate dietary assessment methods are crucial for monitoring MD adherence.
• Existing methods can be time-consuming and prone to recall bias.

Purpose of the Study:

• To develop and evaluate an AI-powered smartphone application for automated Mediterranean diet adherence assessment.
• To compare the AI system's MD adherence scoring with that of an expert dietitian.
• To assess the feasibility and user acceptance of the AI dietary assessment system.

Main Methods:

• Development of an AI system to identify food items and estimate serving sizes from meal photographs.
• Integration of the AI system into a smartphone application for automatic MD adherence scoring and feedback generation.
• Validation of the system by comparing AI-calculated scores with dietitian assessments and conducting a feasibility study with 24 participants.

Main Results:

• The AI system's MD adherence scores showed a statistically insignificant mean difference of 3.5% compared to expert dietitian assessments.
• The image recognition component achieved 61.8% mean Average Precision on a test set and 57.3% on feasibility study images.
• The feasibility study yielded positive feedback from participants and dietitians regarding the system's performance and usability.

Conclusions:

• The AI-powered smartphone application provides a feasible and accurate tool for assessing Mediterranean diet adherence.
• This technology has the potential to enhance dietary monitoring and support public health initiatives.
• Automated dietary assessment using AI can offer a convenient and objective alternative to traditional methods.