mORAL: An mHealth Model for Inferring Oral Hygiene Behaviors in-the-wild Using Wrist-worn Inertial Sensors.

Area of Science:

• Biomedical Engineering
• Digital Health
• Wearable Technology

Background:

• Passive monitoring of oral hygiene behaviors is crucial for public health.
• Existing methods often require specialized or instrumented devices, limiting widespread adoption.
• Reliable detection of infrequent daily behaviors like toothbrushing from sensor data remains a challenge.

Purpose of the Study:

• To develop and validate a model for passive detection of oral hygiene behaviors (brushing, flossing) using wrist-worn inertial sensors.
• To enable detection without instrumented toothbrushes, ensuring applicability with manual toothbrushes.
• To improve the performance of detecting rare daily events from sensor data.

Main Methods:

• Development of the mORAL model utilizing wrist-worn inertial sensor data.
• Event-based candidate window identification strategy, contrasting with fixed-length timeblocks.
• Training and testing on 2,797 hours of data from 25 participants over 192 days, with video annotations for ground truth.
• Evaluation of brushing detection performance, including recall and false positive rate.

Main Results:

• The mORAL brushing model achieved 100% median recall.
• A low false positive rate of one event in every nine days of sensor wear was recorded.
• The average error in estimating the start/end times of detected brushing events was 4.1% of the actual event duration.

Conclusions:

• Passive detection of oral hygiene behaviors from wrist-worn sensors is feasible and accurate.
• The mORAL model demonstrates high performance in detecting toothbrushing, even with manual toothbrushes.
• Event-based detection strategies significantly enhance the performance for rare daily activities.