Accelerometer-based estimation of respiratory rate using principal component analysis and autocorrelation.

Area of Science:

• Biomedical Engineering
• Physiological Monitoring
• Wearable Technology

Background:

• Respiratory rate (RR) is a critical vital sign, yet its monitoring is often overlooked.
• Existing RR monitoring technologies are frequently expensive or impractical for continuous use.
• Tri-axial accelerometry offers a minimally invasive approach for continuous RR monitoring.

Purpose of the Study:

• To validate RR estimation using tri-axial accelerometry against a reference method.
• To assess the agreement of accelerometry-based RR monitoring across a broad range of breathing rates.
• To compare a novel PCA-autocorrelation method with a single-axis approach.

Main Methods:

• Twenty-five healthy participants underwent RR monitoring using an abdominal tri-axial accelerometer.
• Accelerometer data were processed via low-pass filtering, principal component analysis (PCA), and autocorrelation.
• Breathing patterns included slow, normal, and fast paces, with a flow meter serving as the reference standard.

Main Results:

• The PCA-autocorrelation method demonstrated excellent agreement with the reference, showing a bias of 0.0 bpm and limits of agreement of [-1.9; 1.9] bpm.
• 99% of estimated RRs fell within ±2 bpm of the reference values, with a strong Pearson correlation (r=0.99, p<0.001).
• In contrast, the single-axis method exhibited significantly poorer agreement (bias=3.7 bpm, LOA=[-14.9; 22.3] bpm, r=0.44).

Conclusions:

• The PCA-autocorrelation method using tri-axial accelerometry provides a highly accurate and reliable means for respiratory rate monitoring.
• This advanced signal processing technique significantly outperforms simpler single-axis methods.
• Tri-axial accelerometry presents a validated, practical solution for continuous and wide-range respiratory rate assessment.