Automatic lung sound analysis accuracy: a validation study - LUNGPASS

Automatic lung sound analysis accuracy: a validation study

Background: LungPass is an Automatic Lung Sound Analysis app-enabled device, including an electronic wireless stethoscope and a mobile application developed using neural networks training methodology. Such stethoscope could have crucial practical applications, such as the early identification of acute and exacerbations of chronic airway diseases. LungPass has achieved high accuracy in the identification of the main lung sounds in a training cohort. We present the accuracy characteristics of LungPass, evaluated in an independent, validation cohort.

AccuracySpecificitySensitivityPositive predictive valueNegative predictive value
Normal breathing89.1%91.3%82.5%75.9%94.0%

Methods: The accuracy characteristics of the LungPass were evaluated in a validation cohort of 110 patients with different lung diseases (74 adults and 36 children). Gold standard was defined as the judgement of a panel of expert pulmonologists about the recorded sounds.

Results: Our analysis was based on 320 sound recordings, including 80 sounds that were judged to be artifacts, in advance. The overall accuracy of the device in classifying lung sounds, in comparison with the gold standard was 81.6% [95% CI: 73.1%-90.1%]. The accuracy characteristics LungPass for identifying specific lung sounds are summarized in table 1. The auscultation was safe. Three patients developed hyperventilation but no other local or systemic adverse events were observed.

Conclusion: The LungPass platform is accurate. In the future, it could be used to complement clinical examination, or for patients’ telemonitoring.


Cite this article as: European Respiratory Journal 2021; 58: Suppl. 65, PA3456.

This abstract was presented at the 2021 ERS International Congress, in session “Prediction of exacerbations in patients with COPD”.

This is an ERS International Congress abstract. No full-text version is available. Further material to accompany this abstract may be available at (ERS member access only).

Copyright ©the authors 2021


contact us

A member of our team will respond within 48 hours.