RatioLogo
Back

The Family Eating Dynamics Monitoring Revolution

For preventive medicine, the dinner table has become a new frontier. While family influence on diet is well-known, capturing the "how" and "when" has been hindered by unreliable human memory.

A new study introduces a system designed to move past biased food diaries and into the realm of real-time, objective data.

Introducing MFED: A New Paradigm in Monitoring

MFED (Monitoring Family Eating Dynamics) is a system developed by a multi-disciplinary research team. It was deployed to track the eating habits of 74 participants, including 39 children, in their own Los Angeles homes over a 14-day period.

This matters because obesity and Type 2 diabetes are often the result of family-wide patterns, not individual failures.

The Core Technical Innovation: Thresholding + CNN

The system's breakthrough is a two-step algorithm that conserves resources.

  • Traditional Wearables: Constantly stream data, draining batteries.
  • The MFED Approach: Uses a "Thresholding + CNN" algorithm.
  • The Result: A 19% improvement in F1-score for detecting eating gestures using less than 20% of the computational resources.

Essentially, the smartwatch only activates heavy-duty processing when it detects a likely hand-to-mouth movement.

Key Findings from the Home Laboratory

The deployment yielded significant human behavioral data.

  • Ecological Momentary Assessments (EMAs): 14,413 digital prompts were sent.
  • Detected Eating Gestures: 637 alerts were triggered.
  • User Confirmation Rate: 72.8% were confirmed as true eating events (256 meals, 87 snacks).
  • Family Correlation: A 0.87 correlation in response rates suggested health tracking engagement is contagious within families.

Challenges & The Path Forward

The transition to home-based monitoring revealed specific "ghosts in the machine."

  • Common False Positives: Movements like fixing hair, smoking, or using a phone mimicked eating.
  • Limitations: Bluetooth beacons in common areas meant the system lost track when participants charged devices or left the house.
  • Next Steps: The team aims to refine sensors to distinguish between family members during group meals and differentiate between eating and other hand gestures.

Source: MFED: A System for Monitoring Family Eating Dynamics by Md Abu Sayeed Mondol, Brooke Bell, Meiyi Ma, et al. (arXiv:2007.05831v1, July 11, 2020).