Behavioral interventions for weight loss include self-monitoring and self-report of diet. However, few people adhere to self-monitoring because they believe it will be a burden. For those who do adhere, biased reporting leads to poor accuracy. Self-report measures have helped us understand contextual factors of problematic eating behaviors, but we cannot validate such relationships to dietary intake and weight because of unreliable reporting. Therefore, we need an objective way to validate self-report measures. Automated self-monitoring, however, can lead to high eating false alarm detection rates (calling something eating when it is not eating). It also requires participants to push buttons during the start and end of a meal. Passive and unobtrusive ways to capture images of food intake would improve accuracy of detection, avoid the need for a person to remember what they ate, and limit bias based on what participants believe to be socially desirable. Such methods could also ease the self-monitoring burden, decrease errors associated with self- report measures, and lay the foundation for understanding when and how problematic eating behaviors occur. Image capture using wearable cameras may help us better understand obesity and its context (that is, the situation in which eating occurs). However, privacy and ethical concerns of bystanders whose images are taken are a significant barrier. Currently there is no privacy-preserving camera that participants are both willing to wear and that provides meaningful information on food intake and context associated with problematic eating behaviors. Several methods exist, known as obfuscation, that can filter unnecessary information in the scene. However, we do not know which method is most acceptable to a person wearing the device in everyday life that would encourage greatest wear-time. Our project aims to determine which method is best for preserving privacy while providing enough information to understand eating behaviors and their context. To do this, we will observe participants in their everyday life, with special attention to eating behaviors. We will use these new image capture techniques to help understand eating behaviors associated with obesity. First, we will select the best obfuscation method by testing the most well-known obfuscation methods in a cross-over trial to identify which method has the greatest participant acceptability and feasibility (compared with no obfuscation). We will test a novel wearable camera with an infrared sensor (allows us to determine objects that are near vs. far in the camera) with 3 obfuscation methods in real-world settings, including a cartooning gaming?based method. We will then select the obfuscation method that increases wear time and design an eating algorithm around it. Using this algorithm we will assess our ability to capture behaviorally meaningful context from the obfuscated image, such as whether people are eating alone or not, at home, using screen time, and eating prepared or cooked food. This will improve current research practices of evaluating dietary intake and pave the way for personalized interventions in behavioral medicine.
Obesity treatment would benefit greatly from an accurate understanding of problematic eating behaviors in real-world settings. Wearable video cameras provide significant utility in understanding eating behaviors, yet pose a major privacy concern in the real-world, significantly impacting longitudinal wear time. We are testing a privacy-conscious wearable device that individuals with obesity will wear and that can accurately detect eating, which will advance our understanding of behavioral phenotypes surrounding obesity, laying the foundation for future interventions to change problematic eating behaviors.
