Stellenbosch University (SU) researchers are leading a €10 million international trial exploring how artificial intelligence can improve tuberculosis (TB) diagnosis using handheld ultrasound and smartphones.

The project – – aims to create an AI-driven diagnostic tool that helps frontline healthcare workers detect likely TB cases quickly and accurately, especially in under-resourced settings. It brings together 10 health and research institutions from Africa and Europe and is funded by the EU’s Global Health EDCTP3 Joint Undertakings.

“TB is the world’s deadliest infectious disease but remains severely underdiagnosed,” said Prof Grant Theron, SU trial coordinator. “We often test the wrong people at the wrong time. This project tackles the urgent need for scalable, accessible screening tools.”

The trial will involve 3,000 adult patients across South Africa, Benin, and Mali. Using point-of-care lung ultrasound – once limited by the need for trained radiologists – AI will now allow minimally trained workers to identify possible TB indicators without needing lab specimens.

Theron explained that AI can now automate image classification, offering a specimen-free triage tool that improves decision-making at the primary care level.

SU’s Faculty of Engineering, led by Prof Thomas Niesler’s team, will help design the machine learning algorithms, which will be embedded in a smartphone app compatible with portable ultrasound devices.

The study officially launches on , under co-leadership from Dr Veronique Suttels of the Swiss Federal Technology Institute of Lausanne and Prof Ablo Prudence Wachinou from Benin’s National Teaching Centre for Pneumology & Tuberculosis.

The CAD LUS4TB initiative also seeks to generate policy-driven evidence supporting AI-assisted lung ultrasound in TB management, with the long-term goal of integrating this technology into national healthcare systems.