Teleauscultation: an innovative initiative to categorize and analyse lung sounds

Abstract

Auscultation on the chest has long been considered a valuable part of the physical examination. The analysis of acoustic signals accompanying is non-invasively suitable to assess respiratory and cardiac function. However, the diagnosis based on auscultation maybe subjective and biased by human factors. We aimed at extending the auscultation technic by telemedicine and artificial intelligence tools to minimize the subjective factors present in the traditional evaluation of lung sounds.

Teleauscultation was performed on hospitalised or outpatient care children aged 3 months to 10 years with diagnosed acute bronchiolitis (n=9), asthma (n=5), obstructive bronchitis (n=4), chronic bronchitis (n=3) and other lung diseases (n=10) by using electronic stethoscope (3M Littmann Model 3200) controlled by self-developed smartphone application. The registered sound files were uploaded to a clinical cloud storage that can be accessed only by the participating paediatric pulmonologists for further evaluations. This web based system allows the annotation of the acoustic events such as wheezes, stridors, crackles and other disease-specific lung sounds.

Altogether 340 characteristic sounds were identified from the 192 recordings, with number of rhonchi, wheezes, stridors and other lung sound events of 140, 25, 40 and 135, respectively. This teleauscultation system is suitable for replaying and systematic re-evaluation of the sound files, which has importance in more detailed further analyses and in medical education. Applying intelligent learning algorithms to the archived and structured acoustic database has the promise to contribute to the more objective interpretation of lung sounds.

Grant: EFOP-3.6.1-16-2016-00008