2023-06-21 13:35:08
Asthma, bronchiolitis, pneumonia, three pathologies that can affect the respiratory functions of children. At the time of the medical consultation, it is important to be able to quickly differentiate between these three respiratory diseases. Swiss researchers have just developed an algorithm which makes it possible to distinguish these three pathologies, simply using an electronic stethoscope. Explanations.
Asthma, bronchiolitis, pneumonia, three childhood respiratory diseases
The bronchiolitis affects approximately 30% of children under two years of age each winter, causing winter epidemics that severely restrict the functioning of pediatric wards. There pneumonia of the child can be viral or bacterial. Pneumococcal pneumonia can be prevented by vaccinating infants. Finally, theasthma affects a growing number of children in France and causes approximately 30,000 hospitalizations of children under 15 every year.
These three pathologies have in common to affect respiratory function. If the respiratory signs are easily identifiable using a stethoscope (conventional or electronic), the cause of the symptoms and therefore the distinction between these three pathologies remains more delicate. The search for the cause often leads to additional examinations, in particular imaging. The rapid identification of the exact cause of the symptoms is essential to put in place an appropriate therapeutic strategy. Indeed, these three pathologies, although all three respiratory, do not require the same treatments.
Read also – Childhood asthma, the nasal microbiota in question?
Artificial intelligence to interpret breath sounds
In this context, Swiss researchers have used new technologies of artificial intelligence to develop an algorithm capable of differentiating these three respiratory pathologies of the child with an electronic stethoscope. The human ear has limited ability to distinguish the breath sounds associated with these three diseases. The researchers used with convolutional neural networks the technology of the deep learning, an automatic learning technique, which makes it possible to finely differentiate between very similar sounds. This technique has already proven itself in image recognition, for example in the interpretation of imaging data. This time, the researchers applied it to auditory signals.
To test and validate their algorithm, the researchers formed an observational cohort with 572 children and adolescents under the age of 16, from 5 different countries, of whom 71% presented abnormal respiratory signs (pneumonia, asthma or bronchiolitis) and 29% a normal respiratory function. The breathing of all participants was recorded over an average duration of 28.4 seconds at 8 different anatomical sites, with an electronic stethoscope.
Read also – Bronchiolitis: a vaccine from 2023?
A new diagnostic tool for children with respiratory problems
The Deep Breath algorithm has been shown to be able to effectively distinguish:
Children without respiratory problems Children with respiratory disease; Children with bronchiolitis; Children with asthma (wheezing); Children with pneumonia.
The diagnoses were then confirmed by a consultation with a specialized pediatrician.
THE diagnostic performance of the algorithm remained interesting even when reducing the number of anatomical sites (from 8 to 4) and/or the duration of the recordings (from 28 seconds on average to 5-10 seconds). Such a tool could make it possible to meet diagnostic needs in specific contexts, for example in the event of difficulty in interpreting a clinical examination or even when a consultation with a medical specialist is impossible. It could also allow faster orientation and diagnosis during winter epidemics.
Read also – Seasonal flu: towards a new treatment?
Estelle B., Doctor of Pharmacy
Sources
– DeepBreath-automated detection of respiratory pathology from lung auscultation in 572 pediatric outpatients across 5 countries. www.nature.com. Consulté le 21 juin 2023.
1687366321
#Respiratory #Diseases #Diagnosis #Health #Net