Respiratory illness is one of the major contributors to disease and disability worldwide. Assessing these conditions typically involves in-person evaluations at primary care or hospitals, which can be difficult to access in under-resourced areas, or during pandemics. This study explores a novel approach to assess pulmonary disease by analysing changes in cough acoustics. We posit that vocalization may provide insights into an underlying lung pathology. Utilizing computational fluid dynamics, this study investigates flow-induced changes to vocalization, cough, and lung-generated acoustics for diagnosing pulmonary pathology. We use pneumonia, bronchiectasis, and cavitary tuberculosis as exemplars. A computational fluid dynamics model was developed to simulate the acoustic properties of both healthy and diseased lungs, and validated using data from 22 vocal recordings of infected patients with distinct clinical diagnoses. The study employed two computational fluid dynamics methods: Large Eddy Simulation (LES) with the Ffowcs Williams and Hawkings model and the Realizable k − ϵ model with the Broadband Noise Source Model. Our models suggest differences in sound pressure levels across a frequency range of 0 kHz to 16 kHz, with healthy lungs exhibiting higher sound pressure levels compared to those affected by the other conditions. Comparative analysis between actual cough recordings and our model predictions indicate qualitative correlations, suggesting that our model could enhance the understanding of flow-induced acoustics in various lung pathologies. The results suggest potential for developing synthetic datasets to train artificial intelligence models aimed at telemedicine-based diagnosis, thereby improving access to initial respiratory assessments and triage. • Novel CFD approach to analysis of cough acoustics for pulmonary disease assessment • Studied Healthy lungs, pneumonia, bronchiectasis and cavitary tuberculosis • Method validated against 22 patient recordings with distinct clinical diagnoses • Model correlations with recordings advance understanding of respiratory acoustics • Potential to generate synthetic data to train AI model for telemedicine diagnoses
Makhanya et al. (Mon,) studied this question.