Watch: AI and blood test could change the way TB is diagnosed

A South African-led research project is combining AI and a simple finger-prick blood test to help detect a life-threatening infectious disease more quickly.

8 hours ago

Content supplied 2 minutes read

Photo for illustration purposes. Photo: Mufid Majnun/Unsplash

A new international research project aims to make TB testing quicker and easier for patients by combining AI technology with a simple finger-prick blood test.

The project, co-ordinated by Stellenbosch University (SU), will develop and test a new way of detecting TB that could reduce the need for patients to provide sputum samples (mucus coughed up from the lungs).

If successful, the technology could help healthcare workers identify TB faster, allowing patients to start treatment sooner and reducing the risk of the disease spreading.

One of the major challenges in the fight against the disease is effective detection. Of the estimated 10.7 million new cases of TB each year, around 2.5 million people remain undiagnosed. This is partly because current diagnostic tools are often too expensive, laboratory-dependent or difficult to deploy at the point of care.

To improve access to timely TB diagnosis, the new international research project, AddiCAD, was officially launched in May 2026 with R46m in funding from the Global Health European and Developing Countries Clinical Trials Partnership 3.

It brings together six partners from Africa and Europe with complementary expertise in clinical research, diagnostics, artificial intelligence, data science and implementation.

Combining AI imaging with biomarker science

The novel approach combines two promising technologies in a single diagnostic model – CAD4TB, an AI system that analyses digital chest X-rays for signs of TB, and a biomarker test measuring the body’s immune response to infection. By integrating these data sources, AddiCAD aims to provide more accurate results than either method can achieve on its own.

The project builds on preliminary findings showing that AddiCAD achieved a 20% improvement in specificity compared to CAD4TB alone, while still finding most people who have TB. This means the combined approach could help reduce false-positive results while still identifying people who are likely to have TB – an important step towards more efficient and reliable diagnosis in high-burden settings.

Designed for use where rapid diagnosis is needed most

The innovation has the potential to transform TB screening and diagnosis in areas with limited healthcare resources. Rather than relying solely on sputum samples – which can be difficult to obtain and process – healthcare workers could use AddiCAD to rapidly identify people most likely to have TB and ensure they receive testing and treatment without delay.

Now that the project is underway, the partners will develop a new biosensor and a mobile app, before testing them in a clinical study involving about 1 000 adults with suspected TB in South Africa, Namibia and The Gambia. The team will also work closely with healthcare providers, patient representatives, regulators and commercial partners to support future use in healthcare systems. If the initial findings are validated, implementation of AddiCAD may enable life-saving treatment to many additional TB patients.

“For many people, a timely TB diagnosis can prevent negative consequences like transmission, lung damage or death. Yet far too many diagnoses are delayed or missed,” says Prof Stephanus Malherbe, SU associate professor in immunology and AddiCAD project co-ordinator.

“What excites us about AddiCAD is its potential to bring together cutting-edge science and real-world usability in a way that could make accurate diagnosis more accessible where it is needed most.”