Estimation of respiration rate and sleeping position using a wearable accelerometer

Area of Science:

• Biomedical Engineering
• Sleep Medicine
• Wearable Technology

Background:

• Remote patient monitoring for sleep disorders is crucial.
• Wearable inertial sensors offer a non-invasive, comfortable, and cost-effective solution.
• Accurate estimation of respiration rate and sleeping position is vital for diagnosing sleep-related conditions.

Purpose of the Study:

• To develop and evaluate novel methods for estimating respiration rate and sleeping position using accelerometer data.
• To assess the performance of these algorithms across different sensor locations (chest and abdomen).
• To compare accelerometer-derived respiration rates with benchmark data from polysomnography (PSG) across various sleeping positions.

Main Methods:

• Eleven participants wore tri-axial accelerometers on their chest and upper abdomen during clinical polysomnography (PSG).
• Sleeping position was classified using logistic regression with features from filtered acceleration and orientation data.
• Respiration rate was estimated using an adaptive peak detection algorithm on 30-second epochs, combining filtered acceleration and orientation data.
• Accelerometer-derived respiration rates and positions were compared against PSG benchmark data.

Main Results:

• Mean absolute error (MAE) for respiration rate estimation was below 2 bpm for 83.6% of all analyzed windows.
• Respiration rate MAE did not significantly differ between chest and abdomen sensor locations (p=0.52).
• Reduced MAE was observed for side sleeping positions (1.58±0.54 bpm) compared to supine (2.43±0.95 bpm) (p<0.01).
• The position classifier achieved a ROC AUC of 0.87 for distinguishing supine from side positions and 0.94 for left versus right side positions.

Conclusions:

• Novel methods utilizing wearable accelerometers can accurately estimate respiration rate and sleeping position during sleep.
• These methods provide a low-cost, comfortable, and effective solution for remote, long-term monitoring.
• The findings support the potential for in-home sleep monitoring applications, improving patient management and diagnosis.