Background Most respiratory diseases are associated with abnormal adventitious lung sounds. In contrast to auscultation, computerised lung sound analysis is objective, can be done continuously over an extended period and audio recordings can be stored. To date, only little is known about its application in young infants. Therefore, the aim of this study was to compare computerised wheeze detection with the auditory assessment of trained clinicians during the first months of life.
Methods Lung sounds were recorded in 120 sleeping infants on 144 test occasions by an automatic wheeze detection device (PulmoTrack®) at a median (interquartile range) postmenstrual age of 51(44.5–67.5) weeks. The records were blinded and evaluated retrospectively by three trained clinicians. If there was agreement in their assessment, these data were used to calculate optimal cut-off values for the automatically detected duration of inspiratory and expiratory wheezing related to the inspiratory or expiratory time, using ROC-analysis. Sensitivity, specificity and the inter-rater agreement were calculated.
Results Optimal cut-off values for automatically detected inspiratory and expiratory wheezing were 2% and 3%, respectively. The resulting sensitivity of inspiratory and expiratory wheezing were 83.3% and 84.6%, and the specificity 78% and 82.5%, respectively (Figure). The inter-rater agreement was moderate with a Fleiss’ Kappa of 0.59 for inspiratory wheezing and 0.54 for expiratory wheezing.
Conclusion Computerised lung sound analysis is feasible already during the first months of life and provides quantitative and noninvasive information about the extent of wheezing, whereas the assessment by trained clinicians was subjective and only moderate in inter-rater agreement.