Research pick: Every breath you don’t take - "Acoustic analysis of chronic obstructive pulmonary disorder using transfer learning: a three-class problem"

Artificial intelligence (AI) might be used to improve the early detection of chronic obstructive pulmonary disease (COPD), according to research in the International Journal of Innovative Computing and Applications. COPD is a serious and ultimately terminal condition of the lungs.

COPD is a long-term, progressive lung condition common in smokers and those exposed to noxious volatile substances, although it can affect non-smokers too. It is an umbrella term enshrouds chronic bronchitis and emphysema, both of which cause narrowing and damage to the airways and lead to a persistent cough, excess mucus, shortness of breath, and frequent respiratory infections. The disease gradually reduces the lungs’ ability to move air in and out, and although incurable, early diagnosis allows for better management with medication, pulmonary rehabilitation and lifestyle changes.

The new approach to diagnosis uses machine-learning techniques to analyse digital recordings of lung sounds could help recognise a large number of COPD cases that remain undiagnosed worldwide.

The researchers trained algorithms to differentiate between the sounds of air being inhaled and exhaled by healthy individuals and by patients with confirmed COPD, and other conditions such as asthma, pneumonia, respiratory tract infection, and bronchiolitis. This multifarious training reflects what clinicians routinely face: patients with symptoms that overlap with other respiratory conditions and problems. The algorithmic approach would assist the auscultation approach commonly used by clinician, whereby they listen to the patient breathing using a stethoscope and interpret what they hear.

The algorithms thus-developed can identify the subtle acoustic cues linked to respiratory disease through their prior exposure to the large, diverse datasets. The system achieves 95 per cent accuracy, which make it a useful addition to the diagnostic approaches available, and could be used to triage at-risk patients where clinician numbers and specialist resources are limited. Given that COPD is a leading cause of death and disability, and often progresses unnoticed until lung function is severely impaired, the approach could improve outcomes and quality of life for many putative patients.

Amose, J., Manimegalai, P., Amritha, M. and George, S.T. (2025) ‘Acoustic analysis of chronic obstructive pulmonary disorder using transfer learning: a three-class problem’, Int. J. Innovative Computing and Applications, Vol. 15, No. 3, pp.135–144.