WHO expands use of artificial intelligence to support early tuberculosis detection

14:41

By: Azzat Manal

WHO expands use of artificial intelligence to support early tuberculosis detection

The World Health Organization is expanding its support for artificial intelligence tools designed to help detect tuberculosis at an earlier stage, as health systems continue to face difficulties identifying a significant share of people affected by the disease.

The initiative involves including computer-aided detection software capable of analysing digital chest X-rays in the WHO’s prequalification framework for medical technologies. The approach is intended to help health services identify people who may require further testing, particularly in settings where access to specialised diagnostic services is limited.

Tuberculosis remains one of the world’s major infectious disease challenges. WHO estimates indicate that around 10.7 million people developed tuberculosis in 2024, while approximately 8.3 million cases were officially diagnosed and notified. The disease was associated with about 1.23 million deaths during the same year, highlighting the continuing gap between the estimated burden and the cases identified by health systems.

AI-assisted screening tools work by examining chest X-ray images for patterns that may be associated with tuberculosis. Rather than providing a definitive diagnosis, the technology can serve as an initial screening mechanism, helping healthcare workers determine which individuals should undergo additional examinations.

The WHO has supported the use of computer-aided detection software for tuberculosis screening since 2021, particularly for people aged 15 and above. The technology can be especially useful in areas where radiologists and other trained specialists are in short supply, allowing large numbers of images to be assessed rapidly.

The potential benefit is particularly significant for health systems facing limited resources. Automated analysis can help expand screening programmes while reducing some of the workload placed on medical staff. It can also facilitate earlier referral for laboratory testing and other diagnostic procedures when the software identifies an image requiring further assessment.

However, an AI-generated screening result is not equivalent to a medical diagnosis. Individuals identified as potentially having tuberculosis still need appropriate confirmatory testing and clinical evaluation before treatment decisions are made. Maintaining this distinction is essential to prevent both missed cases and unnecessary treatment.

The WHO’s prequalification process could also provide governments and health authorities with an additional reference when assessing the quality, safety and performance of AI-based medical products. This may be particularly relevant in countries where regulatory capacity for advanced medical technologies remains limited.

Prequalification does not replace national regulatory systems. Decisions concerning the licensing, procurement and use of AI screening software remain the responsibility of individual countries and their health authorities, according to their own regulatory frameworks and public health requirements.

The wider adoption of AI in tuberculosis screening nevertheless presents technical and medical challenges. The performance of algorithms can vary depending on the quality of X-ray images, the equipment used and the characteristics of the populations on which the technology is deployed. Continuous evaluation is therefore important to ensure that tools remain reliable across different healthcare environments.

Effective implementation also requires a clear pathway connecting AI-assisted screening to conventional diagnosis and treatment. A person flagged by an algorithm must be able to access confirmatory tests, qualified medical assessment and appropriate care without unnecessary delays.

By incorporating AI-supported screening into its medical technology framework, the WHO is seeking to use digital tools to help close the gap between the number of people estimated to have tuberculosis and those who are actually detected by health systems. The technology is being positioned as an additional tool for public health services, while medical testing and professional judgment remain central to confirming the disease and guiding treatment.