(P11-021-24) Eating Detection via Smartwatch Orientation Sensor Analysis and Generative Data Augmentation

Objectives: Smartwatches with movement sensors have emerged as promising tools for monitoring and improving individuals' health and lifestyle. This study focuses on applying wristwatch sensor (accelerometers and gyroscopes) data for detecting eating episodes, aiming to provide a non-intrusive, continuous monitoring of dietary behavior.

Methods: Twenty healthy participants were equipped with smartwatches, during four meals under semi-naturalistic settings. All meals were recorded on video and synchronized with sensors, and eating behavior was manually annotated by two reviewers. Raw sensor data were transformed to obtain a rectangular dataset. Rigorous machine learning methods were used to classify eating activity. We employed data augmentation techniques using Generative Pre-trained Transformer (GPT) and Generative Adversarial Network (GAN) to address challenges related to variations in eating behaviors.

Results: More than 130k observations were obtained and analyzed. The best performance was shown by tree-based models (single and ensemble) with a balanced accuracy of 0.75. While high sensitivity was readily achieved, specificity was the most difficult metric to improve. Accelerometer features were identified as the most important. Gyroscope features influenced about 2% of the model performance. GPT-based augmentation faithfully reproduced the data and improved model accuracy by 2% for each augmentation fold.

Conclusions: We demonstrated the efficacy of smartwatch sensors in monitoring dietary behaviors. The utilization of tree-based models, coupled with GPT-based augmentation, highlights a promising avenue for enhancing the accuracy of eating episode detection. This study underscores the potential of wearable technology in providing a non-intrusive and continuous approach to dietary monitoring, albeit with room for improvement in specificity metrics.

Funding Sources: DigitAl lifelong pRevEntion (DARE) PNC0000002