Abstract
Objectives: To investigate eating episodes in a group of adolescents in their home-setting using wearable electromyography (EMG) and camera, and to evaluate the agreement between the two devices. Approach: Fifteen adolescents (15.5 +/- 1.3 years) had a smartphone-assisted wearable-EMG device attached to the jaw to assess chewing features over one evening. EMG outcomes included chewing pace, time, episode count, and mean power. An automated wearable-camera worn on the chest facing outwards recorded four images/minute. The agreement between the camera and the EMG device in detecting eating episodes was evaluated by calculating specificity, sensitivity, and accuracy. Main results: The features of eating episodes identified by EMG throughout the entire recording time were (mean (SD)); chewing pace 1.64 (0.20) Hz, time 10.5 (10.4) minutes, episodes count 56.8 (39.0), and power 32.1% (4.3). The EMG device identified 5.1 (1.8) eating episodes lasting 27:51 (16:14) minutes whereas the cameras indicated 2.4 (2.1) episodes totaling 14:49 (11:18) minutes, showing that the EMG-identified chewing episodes were not all detected by the camera. However, overall accuracy of eating episodes identified ranged from 0.8 to 0.92. Significance: The combination of wearable EMG and camera is a promising tool to investigate eating behaviors in research and clinical-settings.