A novel algorithm to detect non-wear time from raw accelerometer data using deep convolutional neural networks.

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Signal Processing

Background:

• Current non-wear detection algorithms rely on fixed time intervals, leading to limitations in detecting both short and long non-wear episodes.
• These interval-based methods struggle to balance the prevention of false positives and false negatives, impacting data accuracy.

Purpose of the Study:

• To develop and evaluate a novel non-wear detection algorithm that eliminates the need for predefined time intervals.
• To improve the accuracy and efficiency of non-wear time detection in accelerometer data.

Main Methods:

• A deep convolutional neural network was trained to detect non-wear time by analyzing accelerometer data immediately before and after removal/replacement events.
• The algorithm focuses on identifying the transition points of accelerometer wear, rather than monitoring within extended intervals.

Main Results:

• The proposed algorithm achieved perfect precision, a recall of 0.9962, and an F1 score of 0.9981.
• It significantly outperformed several baseline and existing non-wear detection algorithms in evaluation.

Conclusions:

• The novel interval-free non-wear detection algorithm demonstrates superior performance compared to traditional methods.
• The algorithm is adaptable for use with both hip-worn and wrist-worn accelerometers with retraining.