App diagnoses children's illnesses, based on their cough
If your child has a persistent cough, it goes without saying that you should check to see if it's anything serious. Taking them to the doctor is the best bet, but an experimental new cough-analyzing smartphone app could help in situations where that isn't possible.
Led by Dr. Paul Porter and Assoc. Prof. Udantha Abeyratne, researchers at Australia's University of Queensland started by collecting a database of cough audio recordings. These were gathered from 1,437 hospitalized children aged 29 days to 12 years old, who had a variety of already conventionally-diagnosed respiratory ailments.
With help from machine learning algorithms not unlike those used to develop speech recognition systems, the scientists then used 852 of the recordings to train an app to recognize the distinct sounds associated with pneumonia, croup, asthma, bronchiolitis, and general lower respiratory tract disease. When that app was subsequently used to diagnose the other 585 children (based on their recordings), its accuracy ranged from 81 to 97 percent.
Once the app is further refined, the researchers hope that it could be used by parents located in remote areas lacking medical facilities, by doctors who are remotely consulting with patients via telehealth systems, or even as an adjunct diagnostic system by physicians examining children within their own offices.
"It can be difficult to differentiate between respiratory disorders in children, even for experienced doctors," says Porter. "This study demonstrates how new technology, mathematical concepts, machine learning and clinical medicine can be successfully combined to produce completely new diagnostic tests."
A paper on the app was recently published in the journal Respiratory Research.
And in related news, Japanese scientists have previously developed an application known as Respiratory Sounds Visualizer, that diagnoses respiratory disorders by analyzing the sound of patients breathing.
Source: Biomedical Central
Please keep comments to less than 150 words. No abusive material or spam will be published.