A mobile application developed by researchers at the Joondalup Health Campus and Princess Margaret Hospitals in Australia has used artificial intelligence to accurately diagnose common respiratory disorders in children.
According to a research paper published in Respiratory Research, cough sounds were recorded by a smartphone, then an analysis was performed using cough data and up to a five-symptom input derived from patient and parent-reported history – ailments included asthma, croup and pneumonia.
Cough audio streams were recorded on iPhone 6 smartphones in realistic hospital environments, where background noises included talking, running medical equipment, crying, footsteps and closing doors.
An automatic cough detector, previously trained from a dataset of manually selected cough and non-cough events, was then combined with 5-symptom input obtained from the patient’s history.
The AI platform was able to diagnose pneumonia with 87 percent accuracy and asthma with 97 percent accuracy – both surpassing the diagnosing standards of the World Health Organization.
The report also noted the algorithm’s diagnostic accuracy could be further improved with the additional input of clinical signs, for example respiratory rate or chest recessions.
In a blog post describing the app, study author Dr. Paul Porter suggested the advantage that the technology offers is the ability to identify common respiratory disorders without the need for clinical examination, X-rays or bronchodilator testing.
“This study shows how new technologies such as machine learning, mathematical modeling and clinical medicine can come together to successfully develop a novel way of assessing and diagnosing disease,” Porter wrote. “We are optimistic that the technology will add to, rather than replace, the stethoscope as a valuable diagnostic aid, to be used when a clinical examination is not possible.”
He noted the team is expanding their research to look at adult respiratory diseases and examine the use of machine learning technology to stratify disease severity.
The algorithm’s ability to accurately identify lower respiratory tract diseases (LRTDs) from pure upper respiratory disease could also be useful in telehealth and remote area medicine by guiding when to attend or escalate medical care, the report noted.
Machine learning and AI technology could have wide-ranging impacts on the healthcare community – from smartphone apps that can diagnose a child’s cough to AI-enabled hearing aids that can detect a fall and send out an alert.
The Livio AI hearing aid, which features a multi-core twin compressor and dual radio system, became available in Australia this month, along with a mobile application with geotagging capabilities and other features.