Babbs, Charles F., "Biomechanical models of cough sounds in pneumonia: mechanisms underlying sound-based diagnosis in low-resource settings" (2020). Weldon School of Biomedical Engineering Faculty Working Papers. Paper 23.
Date of this Version
Abeyratne, acoustic cough monitoring, analysis, bronchopneumonia, frequency, global health, harmonic motion, lobar pneumonia, oscillation, pediatrics, point-of-care.
Objective. This paper describes a theoretical study of physical mechanisms underlying the generation of cough sounds and the expected differences in such sounds caused by anatomically localized pneumonia. Methods. A fresh, first-principles, physics-based analysis is done, describing radial motion of bronchi after sudden decompression of intrathoracic pressure during the early and middle phases of a cough. The mechanical model of each bronchus is a spring-mass-damper system, in which the spring force comes from elastic properties of the bronchial wall, the mass term comes from the bronchial wall and any surrounding fluid or pus in pneumonia, and the damping term comes from the extracellular matrix within bronchial walls. Upon release of pressure built-up during the compressive phase of the cough, model bronchial walls undergo damped sinusoidal motion. Cough sound intensity is computed as the weighted average of the signed products of air density and the squares of local radial wall velocities, with weighting factors determined by the aggregate inner surface areas of bronchi at each segmental level. Established anatomic and physiologic data, coupled with classical scaling rules, are used to create adult and child sized models with or without pneumonia. Digital signal processing is done to separate the hidden pneumonia-related signal from raw cough sound data. Results. Numerical computations generate simulated cough sounds of realistic amplitudes, durations, and frequency content. Large airways generate loud sounds, and small airways generate much softer sounds. Cough sounds generated by medium and small airways that are surrounded by pneumonic fluid have lower frequency and longer duration. Simple low pass filtering separates the fainter, prolonged, pneumonia-related sounds from louder, earlier sounds generated in the trachea and main stem bronchi. The ratio of the root mean square (RMS) low-pass filtered sound pressure to the RMS raw sound pressure, plotted in the time domain, provides excellent discrimination of pneumonia models from normal ones. Involvement of as little as 10% of the total lung tissue with pneumonic fluid infiltrate yields a 4-fold difference in the RMS power ratio. Conclusions. This work demonstrates a heretofore unrecognized mechanism underlying cough-sound based recognition of pneumonia cases that is anatomically, physiologically, and physically realistic, strengthening the rationale for a low cost, easy-to-use, completely painless, and cell-phone-based diagnostic tool for childhood pneumonia in remote, low-resource settings.