Dr. Véronique Suttels performs a smartphone-connected ultrasound on a symptomatic Beninese patient - 2025 EPFL/Véronique Suttels - CC-BY-SA 4.0
EPFL and Lausanne University Hospital (CHUV) are part of an African-EU partnership awarded €10-million by the EU's Global Health EDCTP3 Joint Undertaking to roll-out an AI-driven application for more accessible and cost-effective tuberculosis (TB) diagnosis.
Today, tuberculosis is preventable and treatable, yet this infectious disease that attacks the lungs still causes 1.3 million deaths every year and is the second leading cause of death in sub-Saharan Africa. The lack of appropriate tests to diagnose and manage TB in primary healthcare, such as chest X-rays, greatly contributes to this burden.
To help address this, EPFL's Laboratory for Intelligent Global Health and Humanitarian Technologies (LiGHT) in the School of Computer and Communication Sciences in partnership with CHUV's Infectious Diseases Department, have developed a new, highly sensitive triage tool: ULTR-AI (Ultrasound-led TB recognition using AI).
A paper presented at a recent Conference in Vienna outlined that this revolutionary AI-powered lung ultrasound fulfills and exceeds the World Health Organization's (WHO) diagnostic accuracy requirements for a point-of-care TB triage test.
Portable, smart, effective
Taking the AI lead in the project, called TrUST, EPFL developed a novel algorithm compatible with portable ultrasound devices that can be plugged into a simple smartphone and automatically assess the ultrasound images for the presence of TB.
"The strength of this revolutionary tool lies not only in its accessibility and low cost, but that it meets the stringent bar set by the World Health Organization for triaging TB. It's not good enough to make a good algorithm," said Adjunct Professor Mary-Anne Hartley, Head of the LiGHT Laboratory. "We have to ensure it is acceptable and accessible for real-world implementation."
"One of the greatest barriers to deploying these new, portable ultrasounds in settings where they are most needed is that ultrasound interpretation requires a lot of training and specialized skills, which is often not accessible to frontline healthworkers," explained Professor Dr. Noémie Boillat-Blanco, a physician with CHUV's Department of Infectious Diseases and a collaborator in the project. "The algorithm increases image reading performance, addressing this major gap towards widespread implementation."
Boillat-Blanco and her colleague Dr Véronique Suttels led a large diagnostic cohort study in Benin using the algorithm in portable ultrasound devices. "When a patient presents with symptoms, the ultrasound algorithm can quickly detect the likelihood of tuberculosis. Another advantage is that, once tuberculosis has been ruled out, pathologies such as pneumonia or cardiovascular disease can be identified using this device," she explained.
An innovative solution based on AI and data sharing
Following the success of the diagnostic TrUST study, a global consortium of ten health and research institutions has been awarded €10-million over five years to further develop the algorithm into a user-friendly app, starting in Benin, Mali and South Africa. Suttels will join EPFL to take the overall scientific lead of the project (CAD LUS4TB) going forward, which will include 3000 adult patients to investigate the use of ULTR-AI in TB triage and management.
"This five-year project is an example of the interdisciplinary work of our LiGHT group which is made up of medical doctors, clinical trialists and data scientists to ensure the genuine co-development of AI tools and impact-driven research, and we are thrilled to be working with partners such as CHUV," said Hartley.
An accessible and widely-implementable device to combat tuberculosis
The CAD LUS4TB African EU partnership is based on interdisciplinary collaboration, with the sharing of clinical and ultrasound data enabling the AI's performance to be continually enhanced. The model will also be open access. It involves specialists from a wide range of fields, including infectiology, clinical research, medical diagnostics, data science, computer science, social sciences, economics and health policy.
"This project is a great first step in the more portable and affordable diagnosis of TB, improving access to care and reducing the costs associated with the late treatment of the disease. Future AI-Lung Ultrasound research should prioritize imaging modalities designed for end-users, exploring barriers to technology adoption, tackle workflow challenges, and-crucially from a clinical perspective-ensure high-quality, continuous Point of Care Ultrasound training," concluded Suttels.
