Predicting Real-Life Eating Behaviours Using Single School Lunches in Adolescents

A smartphone application for semi-controlled collection of objective eating behavior data from multiple subjects

BACKGROUND & OBJECTIVE

The study of eating behavior has made significant progress towards understanding the association of specific eating behavioral patterns with medical problems, such as obesity and eating disorders. Smartphones have shown promise in monitoring and modifying unhealthy eating behavior patterns, often with the help of sensors for behavior data recording. However, when it comes to semi-controlled deployment settings, smartphone apps that facilitate eating behavior data collection are missing. To fill this gap, the present work introduces ASApp, one of the first smartphone apps to support researchers in the collection of heterogeneous objective (sensor-acquired) and subjective (self-reported) eating behavior data in an integrated manner from large-scale, naturalistic human subject research (HSR) studies.

METHODS

This work presents the overarching and deployment-specific requirements that have driven the design of ASApp, followed by the heterogeneous eating behavior dataset that is collected and the employed data collection protocol. The collected dataset combines objective and subjective behavior information, namely (a) dietary self-assessment information, (b) the food weight timeseries throughout an entire meal (using a portable weight scale connected wirelessly), (c) a photograph of the meal, and (d) a series of quantitative eating behavior indicators, mainly calculated from the food weight timeseries. The designed data collection protocol is quick, straightforward, robust and capable of satisfying the requirement of semi-controlled HSR deployment.

RESULTS

The implemented functionalities of ASApp for research assistants and study participants are presented in detail along with the corresponding user interfaces. ASApp has been successfully deployed for data collection in an in-house testing study and the SPLENDID study, i.e., a real-life semi-controlled HSR study conducted in the cafeteria of a Swedish high-school in the context of an EC-funded research project. The two deployment studies are described and the promising results from the evaluation of the app with respect to attractiveness, usability, and technical soundness are discussed. Access details for ASApp are also provided.

CONCLUSIONS

This work presents the requirement elucidation, design, implementation and evaluation of a novel smartphone application that supports researchers in the integrated collection of a concise yet rich set of heterogeneous eating behavior data for semi-controlled HSR.

Validation of a Deep Learning System for the Full Automation of Bite and Meal Duration Analysis of Experimental Meal Videos

Eating behavior can have an important effect on, and be correlated with, obesity and eating disorders. Eating behavior is usually estimated through self-reporting measures, despite their limitations in reliability, based on ease of collection and analysis. A better and widely used alternative is the objective analysis of eating during meals based on human annotations of in-meal behavioral events (e.g., bites). However, this methodology is time-consuming and often affected by human error, limiting its scalability and cost-effectiveness for large-scale research. To remedy the latter, a novel "Rapid Automatic Bite Detection" (RABiD) algorithm that extracts and processes skeletal features from videos was trained in a video meal dataset (59 individuals; 85 meals; three different foods) to automatically measure meal duration and bites. In these settings, RABiD achieved near perfect agreement between algorithmic and human annotations (Cohen's kappa κ = 0.894; F1-score: 0.948). Moreover, RABiD was used to analyze an independent eating behavior experiment (18 female participants; 45 meals; three different foods) and results showed excellent correlation between algorithmic and human annotations. The analyses revealed that, despite the changes in food (hash vs. meatballs), the total meal duration remained the same, while the number of bites were significantly reduced. Finally, a descriptive meal-progress analysis revealed that different types of food affect bite frequency, although overall bite patterns remain similar (the outcomes were the same for RABiD and manual). Subjects took bites more frequently at the beginning and the end of meals but were slower in-between. On a methodological level, RABiD offers a valid, fully automatic alternative to human meal-video annotations for the experimental analysis of human eating behavior, at a fraction of the cost and the required time, without any loss of information and data fidelity.