University of Queensland smartphone app can diagnose respiratory diseases

The University of Queensland (UQ) has developed technology that can analyse a person’s cough and help diagnose respiratory disorders quickly and easily in patients who lack access to doctors.

Udantha Abeyratne, UQ biomedical engineer and associate professor, and his team, have developed diagnostic technology that uses smartphones to instantly identify common respiratory disorders like asthma, croup, pneumonia, lower respiratory tract disease, Chronic Obstructive Pulmonary Disease, and bronchiolitis.

A clinical study on childhood respiratory diseases found the technology to have an accuracy between 81 to 97 per cent.

“Coughs can be described as wet or dry, brassy or raspy, ringing or barking, they can whistle, whoop or wheeze, but experts cannot always agree on the description or how to use cough sounds for diagnosis,” Abeyratne said.

“Our approach is to introduce single processing, and machine classification and learning technologies to extract useful diagnostic characteristics from coughs, removing the subjective elements for characterising them.”

Dr Paul Porter, a paediatrician from Joondalup Health Campus in Western Australia, says it could be difficult to differentiate between respirator disorders in children, even for experienced doctors in modern hospital facilities. However, this new technology could provide a new solution going forward.

“This study demonstrates how new technology, mathematical concepts, machine learning, and clinical medicine can be successfully combined to produce completely new diagnostics tests using the expertise of several disciplines,” Porter said.

Abeyratne says the technology can lead to earlier diagnosis and better patient outcomes throughout the world.

To develop the technology, UQ researchers trained algorithms to recognise features of cough, which are characteristics of five different respiratory diseases. The technology also allowed users to report other noticeable symptoms, to ensure diagnosis are as accurate as possible.