Wheezing Diagnosis Using a Smartphone

Overview

Abnormal respiratory sounds (wheezing and/or crackles) are diagnosis criteria of acute bronchiolitis. One third of these infants will develop recurrent episodes, leading to the diagnosis of infant asthma. Nowadays, no available treatment shortens the course of bronchiolitis or hastens the resolution of symptoms, thus, therapy is supportive. Our hypothesis is that the diagnosis of wheezing during bronchiolitis (\~60% of infants) will help to select infants who will benefit from anti-asthma therapy. In this setting the diagnosis of wheezing is crucial, and an objective tool for recognition of wheezing is of clinical value. The investigators developed a wheezing recognition algorithm from recorded respiratory sounds with a Smartphone placed near the mouth (Bokov P, Comput Biol Med, 2016). The objectives of the present cross sectional, observational study are 1/ to further validate our approach in a larger sample of infants (1 to 24 months) admitted to hospital for a respiratory complaint during the period of viral bronchiolitis, and 2/ to use gold standard diagnosis of wheezing by respiratory sound recording (Littmann) and subsequent analysis by two experienced pediatricians.

Infants (1 to 24 months old) are recruited in two emergency departments (Robert Debré; Antoine Béclère hospitals of Assistance publique - Hôpitaux de Paris) based on a respiratory complaint. Six characteristics are recorded (age, sex, SpO2, presence or absence of wheezing, other respiratory sound, initial diagnosis). Two recordings of respiratory sounds are obtained almost simultaneously: one with a Smartphone at the mouth (5 cm) and one with an electronic stetoscope (Littmann). Two expert pediatricians listen the recordings giving thee groups: with wheezing (agreement), without wheezing (agreement) and non agreement diagnosis. The recordings made with the Smartphone are subjected to the wheezing recognition algorithm as previously described. The sensitivity, specificity, PPV, NPP are then evaluated. The algorithm will further be improved if necessary using the true negative and true positive recordings (those with expert agreement).

Conditions