CAD LUS4TB

The challenge

Tuberculosis is one of the world’s most common infectious diseases, infecting an estimated one in four of the world’s population. Only a proportion of those infected developactive TB disease, which can be fatal if not treated. However, it is challenging to identify these patients because the symptoms of TB are non-specific, and diagnostic tests require specialist equipment (such as chest X-rays) or expensive molecular tests that are not available at many primary care facilities.

As a result, many patients with TB are not identified, so they do not get the right treatment, leaving the disease with more opportunities to spread. To enhance detection rates, there is interest in TB triage (rapid, easy-to-use assessments of the likelihood of TB disease), enabling diagnostic follow-up to focus on those most likely to have active TB.

Point-of-care lung ultrasound has emerged as a promising approach for TB triage. Point-of-care ultrasound is beginning to be used in primary care centres in sub-Saharan Africa, and a study in Benin has shown that the technology can discriminate between TB and other lung infections. However, while health workers were comfortable using the technology and acquiring images, they were less confident in interpreting them. 

The project

Building on these promising findings, the CAD LUS4TB project is refining a digital tool for automatically categorising lung ultrasound images and assessing its suitability for TB triage.

Around 50,000 expert-labelled images from 500 patients from Benin will be analysed to fine-tune the diagnostic algorithm. At seven sites in Benin, Mali, and South Africa, the use of this algorithm will be assessed in a stepped-wedge clinical trial, in which introduction is phased over time and sites scheduled for later-stage introduction act as controls. 

The diagnostic performance of the lung ultrasound algorithm will be compared with gold-standard methods, such as culture and molecular diagnostic tests. The project will also compare lung ultrasound’s performance with a WHO-recommended diagnostic strategy that incorporates chest X-rays.

The project will also investigate ways to simplify image collection and analysis for non-expert use. In addition, it will use the trial to investigate whether lung ultrasound images can help identify the likely causes of respiratory symptoms not caused by TB. 

To support routine implementation, the project will develop a mobile phone-based app and conduct consultations to assess issues such as health worker perceptions, workflow integration, and patient acceptability. 

The project builds on the work of another Global Health EDCTP3-funded project, IMCI-Plus, which is evaluating the use of point-of-care ultrasound to determine the need for antibiotic use in children with respiratory tract infections.

Impact

The CAD LUS4TB project will assess the potential of an emerging new technology for triaging potential TB cases. It will:

• Generate high-quality evidence on the ability of automated lung ultrasound image analysis to rule out TB infection.
• Provide a way for expensive molecular testing to be focused on those most likely to have TB.
• Generate insights into the factors that might affect the implementation of the new approach in routine practice.

If the value of lung ultrasound for TB triage is confirmed, it could be rapidly adopted to refine the TB diagnosis pathway, reducing the very large number of patients who currently do not receive a diagnosis and appropriate treatment.