Analysis of respiratory kinematics: a method to characterize breaths from motion signals

Abstract

Breathing motion (respiratory kinematics) can be characterized by the interval and depth of each breath, and by magnitude-synchrony relationships between locations. Such characteristics and their breath-by-breath variability might be useful indicators of respiratory health. To enable breath-by-breath characterization of respiratory kinematics, we developed a method to detect breaths using motion sensors. In 34 volunteers who underwent maximal exercise testing, we used 8 motion sensors to record upper rib, lower rib and abdominal kinematics at 3 exercise stages (rest, lactate threshold and exhaustion). We recorded volumetric air flow signals using clinical exercise laboratory equipment and synchronized them with kinematic signals. Using instantaneous phase landmarks from the analytic representation of kinematic and flow signals, we identified individual breaths and derived respiratory rate (RR) signals at 1Hz. To evaluate the fidelity of kinematics-derived RR, we calculated bias, limits of agreement, and cross-correlation coefficients (CCC) relative to flow-derived RR. To identify coupling between kinematics and flow, we calculated the Shannon entropy of the relative frequency with which flow landmarks were distributed over the phase of the kinematic cycle. We found good agreement in the kinematics-derived and flow-derived RR signals [bias (95% limit of agreement) = 0.1 (± 7) breaths/minute; CCC median (IQR) = 0.80 (0.48 – 0.91)]. In individual signals, kinematics and flow were well-coupled (entropy 0.9-1.4 across sensors), but the relationship varied within (by exercise stage) and between individuals. The final result was that the flow landmarks did not consistently localize to any particular phase of the kinematic signals (entropy 2.2–3.0 across sensors). The Analysis of Respiratory Kinematics method can yield highly resolved respiratory rate signals by separating individual breaths. This method will facilitate characterization of clinically significant breathing motion patterns on a breath-by-breath basis. The relationship between respiratory kinematics and flow is much more complex than expected, varying between and within individuals.

Indirect Estimation of Breathing Rate Using Wearable Devices

“…Different signal processing methods, like fast Fourier transform (FFT) and wavelet transform [17,20], or methods based on Artificial Intelligence (AI) can be employed for the indirect estimation of BR starting from a plethora of physiological signals (e.g., ECG, PPG, ballistocardiogram, seismocardiogram, oscillometric trace, and Korotkoff sounds related signal [21,22]). Indeed, AI algorithms are more and more applied for different purposes and literature is replete with examples of this type.…”