Spectrograms of a TB cough and a non-TB cough. Credit: Image: Stellenbosch University
A few coughs, an app and 30 seconds could soon be all that healthcare workers need to determine whether someone needs further, expensive yet critical tuberculosis (TB) testing.
Work on Cough Audio Triage for TB (CAGE-TB) is progressing well, says the South African team leading its development. Artificial intelligence (AI) driven algorithms integrated into the app ‘listen’ for certain audio signatures in coughs that reflect distortions in the lower parts of the lungs, an area typically damaged in those who have TB in much of Africa.. The telltale tone of a TB cough can be picked up by common cell phone microphones, but not by human ears, the researchers discovered.
CAGE-TB project coordinator, Daphne Naidoo says that a database of cough audio recordings is being collected from people who are TB positive; healthy; or have other respiratory conditions, but are suspected of having TB. These are used to develop statistical classifiers that require short-term spectral information to automatically distinguish between TB and non-TB coughs.
According to Thomas Niesler of the Digital Signal Processing Group in the Faculty of Electrical and Electronic Engineering at Stellenbosch University, South Africa, the human ear cannot tell frequencies apart when they are too close, but the algorithm used for CAGE-TB can do so. It uses this information in its decisions.
“This appears to indicate that the algorithm is using information that the human auditory system cannot detect,” he notes.
Naidoo believes the app will transform the process of managing potential TB patients. Many people visiting such health facilities could benefit from TB testing, but are not tested because they do not yet show symptoms, while others receive unnecessary testing.
Whenever the app identifies a possible case, a person would be referred for testing before medical staff can decide whether further TB treatment is needed, explains lead investigator, Grant Theron, of the Clinical Mycobacteriology and Epidemiology (CLIME) Group in the Division of Molecular Biology and Human Genetics at Stellenbosch University, South Africa.
Cost-effective, objective, quick and easy methods that can easily be used at clinics and that do not need biological sputum samples are the ‘holy grail’ of TB testing, Theron adds, and would markedly curb the spread of TB. The app could be especially helpful in screening for early stage TB patients who do not yet have symptoms, but can be spreading the disease.
The concept of a distinguishable and measurable TB cough ‘audio signature’ was first proposed in 2018 in Physiological Measurement after initial testing by digital signal processing engineers and TB specialists from Stellenbosch University and the University of Cape Town. It followed ventures in Australia showing proof of concept for childhood pneumonia.
At the time, the researchers said that although several effective diagnostic methods existed, clinics in lower income countries are often not able to rigorously clinically examine people to determine further testing, nor afford expensive tests and employ the trained experts needed to interpret results. In these situations, symptoms including cough are commonly used to identify patients for testing.
“However, self-reported cough has suboptimal sensitivity and specificity, which may be improved by digital detection,” they wrote.
Research into CAGE-TB started in earnest in 2021. The project, which recently received a funding boost of R20 million from the European and Developing Countries Clinical Trials Partnership (EDCTP), also involves the Amsterdam Institute for Global Health and Development (Netherlands), Makerere University (Uganda) and Gottingen University (Germany). Screening with CAGE-TB requires only a few forced coughs, which most people can easily produce, rather than having to wait for a bout of coughing or for a patient to produce sputum from within their lungs that can be tested, says Theron.
“The app is an exciting opportunity to transform TB diagnosis at scale, ensuring more people are tested, testing itself is done more efficiently, and TB is diagnosed earlier, stopping transmission in its tracks..
