Late Breaking Abstract - Performance of artificial intelligence in determining the intensity of abnormal breath sounds in asthma patients

Abstract

Background: Effective and reliable monitoring of asthma at home is a relevant factor that may reduce the need to consult a doctor in person.

Aim: We analyzed the possibility to determine intensities of pathological breath phenomena based on artificial intelligence (AI) analysis of sounds recorded during standard stethoscope auscultation.

Methods: The evaluation set comprising 1043 auscultation examination (9319 15-second recordings) was collected from 899 patients. Each examination was assigned to one of four groups: asthma with abnormal sounds (AA), asthma without abnormal sounds (AN), no-asthma with and without abnormal sounds (NA and NN, respectively). Presence of abnormal sounds was evaluated by a panel of 3 physicians that were blinded to the AI predictions. AI was trained on an independent set of 9847 recordings to determine intensity scores (indexes) of wheezes, rhonchi, fine and coarse crackles and their combinations, e.g. continuous phenomena (wheezes + rhonchi) and all phenomena. The pair-comparison of groups of examinations based on Area Under ROC-Curve (AUC) was used to evaluate the performance of each index in discrimination between groups.

Results: Best performance in separation between AA and AN groups was observed with Continuous Phenomena Index (AUC 0.94) while for NN and NA groups All Phenomena Index (AUC 0.91) showed the best performance. The AA group showed only a slightly higher prevalence of wheezes compared to the NA group.

Conclusions: The results showed high efficiency of the AI to discriminate between the asthma patients with and without abnormal sounds, thus this approach has great potential and can be used to monitor asthma symptoms at home.