Researchers from Stellenbosch University (SU) are playing a leading role in a global trial investigating the use of artificial intelligence (AI) to improve the diagnosis of tuberculosis (TB).
The project aims to develop and test an algorithm that will enable health practitioners at primary care facilities to detect suspected TB cases by means of a portable ultrasound device and a smartphone.
“TB is still the world’s deadliest infectious disease, but its diagnosis falls far short,” explains Prof. Grant Theron, professor of clinical mycobacteriology and epidemiology at SU and coordinator of the trial. “A big challenge is that we often test the wrong people at the wrong time. Many patients undergo unnecessary tests, while others who urgently need them never undergo proper screening tests. There is an urgent need for accessible, affordable and expandable diagnostic tools for TB screening.”
Global partnership
The project, called ‘Computer-aided diagnosis with lung ultrasound for community-based pulmonary tuberculosis screening in Benin, Mali and South Africa’ (CAD LUS4TB), involves a consortium of ten health and research institutions across Africa and Europe. The European Union’s Global Health EDCTP3 Joint Undertakings provided funding of €10 million (more than R200 million) for the project.
The study will involve 3 000 adult patients to investigate ultrasound-guided TB recognition using AI in TB screening and management. The aim is to improve access to TB screening that can rule out TB in symptomatic adult patients at the primary health care level.
“Point of care lung ultrasound uses portable imaging devices that can detect abnormalities in the body, including those characteristic of TB,” Theron explains. “This technology was previously limited because specialized expertise was needed to interpret images. However, AI now offers unprecedented opportunities to automate image classification, allowing health practitioners with the minimum training to quickly and easily determine which patients need further testing. CAD LUS4TB is therefore a much-needed, sample-free diagnostic test that can be used in the fight against TB.”
The AI-powered solution
SU will also develop and validate the machine learning algorithms in collaboration with European partners and with the involvement of Prof Thomas Niesler’s Digital Signal Processing Group in SU’s Faculty of Engineering. The new algorithm will be developed by researchers to be compatible with portable ultrasound devices that can be connected to smartphones. The technology will automatically assess ultrasound images for TB indicators and will be packaged into a user-friendly mobile application for widespread deployment.
The project will start on 1 September 2025 under the co-leadership of Dr Veronique Suttels from the Laboratory for Intelligent Global Health and Humanitarian Technologies at the Swiss Federal Institute of Technology of Lausanne, as well as Prof Ablo Prudence Wachinou from the National Teaching Centre for Pneumology and Tuberculosis in Benin.
The CAD LUS4TB consortium focuses on generating population-specific evidence and advocates for the integration of computer-aided diagnosis (CAD) using AI to support the implementation of lung ultrasound in healthcare policies.
