Measuring Caloric Intake at the Population Level (NOTION): Protocol for an Experimental Study.

Area of Science:

• Biomedical Engineering
• Machine Learning
• Nutritional Science

Background:

• Traditional dietary monitoring methods (questionnaires, diaries) have significant inaccuracies.
• Wearable devices offer potential but struggle with massive datasets and limited analytical accuracy.
• Accurate caloric intake monitoring is crucial for individual and public health.

Purpose of the Study:

• To develop a precise and stable algorithm for estimating caloric intake using machine learning.
• To overcome limitations of current wearable device analytics for dietary monitoring.

Main Methods:

• Recruited 20 healthy Italian adults (18-66 years) for 2 months of data collection.
• Utilized wearable devices to capture eating activity data, linked with anthropometric and demographic information.
• Employed advanced machine learning to analyze massive data flows and predict caloric intake, validated by calorimetric assessments.

Main Results:

• Expect to develop a prototype algorithm for real-time caloric intake estimation.
• The algorithm aims to recognize food type and quantity (number of bites) from movement data.
• High accuracy is anticipated in matching bite-related movements to corresponding caloric intake.

Conclusions:

• An automated caloric intake calculation method independent of proprietary device algorithms has significant potential.
• Applications include clinical nutrition (e.g., cardiovascular health, dietary control) and public health surveillance.
• This approach can serve as a low-cost tool for monitoring population eating habits.