University of Washington Medicine doctors along with researchers across the globe have created technology that screens respiratory illness like tuberculosis through an app on a smartphone.
If study results are replicated, a theoretical smartphone app could expand testing in areas of the globe where it is not available.
This initial study focuses on TB, finding coughs caused by the bacterial disease have a distinct signature that can be discerned through audio. The small differences in the coughs of tuberculosis and other diseases are not discernible by the human ear. But algorithm technology used in the study can detect these differences and correctly categorize a TB cough 82% of the time, according to the study.
“Working with smartphones and machine learning allowed us to look at cough frequency rather than sound. It is not possible to differentiate TB coughs by just hearing with the human ear,” said lead author Manuja Sharma, a UW doctoral student in electrical and computer engineering.
The study’s participants were sourced and studied in Kenya, where TB is common. Researchers collected over 33,000 coughs from 149 TB patients and from 46 patients who presented with other respiratory illnesses.
The audio of their coughs were fed through a machine-learning AI model and sorted accordingly.
Though rates of TB remain low in the United States, it is the leading cause of infectious diseases worldwide behind COVID-19. According to the World Health Organization, 10.6 million people contracted TB in 2022 and 1.3 million people died from the disease that year.
The disease is especially prevalent in the global south where testing for TB and other diseases may be more difficult to obtain. A conventional diagnosis of TB requires a sophisticated clinical laboratory with access to supplies and reliable electricity.
“Something that is cellphone-based where there is not someone needed to administer the test could have a major impact on low resource places,” said Dr. David Horne, UW professor of medicine and an author on the paper. “It also could allow physicians to go ahead and treat the suspected TB while you are waiting for results of the regular test, which can take a week or two to take back.”
Though the UW study needs to be replicated before an app can be developed, Horne hopes this technology can be replicated to diagnose other respiratory illnesses like COVID or the flu. It may even aid in the diagnosis of certain respiratory conditions like asthma.
“You could imagine the applications of this technology could be broad and go far beyond tuberculosis,” he said.