Measuring the Consumption of Individual Solid and Liquid Bites Using a Table-Embedded Scale During Unrestricted...

Area of Science:

• Nutrition Science
• Biomedical Engineering
• Data Science

Background:

• The universal eating monitor (UEM) measures food consumption but requires restricted eating conditions.
• Laboratory studies using UEM are limited by the need to control food types and eating behaviors.
• Unrestricted eating environments present challenges for accurately measuring individual food bites.

Purpose of the Study:

• To develop and validate a novel algorithm for detecting and weighing individual food bites during unrestricted eating.
• To overcome the limitations of the UEM by enabling analysis of scale data without strict behavioral constraints.
• To assess the algorithm's performance in a real-world cafeteria setting.

Main Methods:

• An algorithm was designed to identify stable scale weight periods and analyze surrounding weight changes.
• The algorithm differentiates between single food bites, mass bites, and drink consumption based on weight change patterns.
• Performance was evaluated on 271 subjects, with 24,101 bites manually annotated from synchronized videos for ground truth.

Main Results:

• The algorithm achieved approximately 39% bite detection accuracy with a low false positive rate (1 FP per 10 actual bites).
• This represents a threefold increase in true detections and a 90% reduction in false positives compared to the UEM.
• No significant difference in average bite weight was found between weighable and unweighable bites.

Conclusions:

• The developed algorithm significantly enhances the accuracy and feasibility of using table scales for measuring food intake.
• This innovation allows for studies in non-laboratory settings with natural eating behaviors.
• The algorithm holds potential for broader applications in nutritional research and clinical settings.