Ultrasound and AI: A new frontier in child pneumonia diagnosis across Africa
-
Childhood pneumonia is a disease of global health concern, and the leading cause of child morbidity and mortality in sub-Saharan Africa.
Childhood pneumonia is a disease of global health concern, and the leading cause of child morbidity and mortality in sub-Saharan Africa. Simultaneously Antimicrobial Resistance (AMR) has been identified by the World Health Organisation (WHO) as one of the “top global public health and development threats” facing the globe. The combination of lung ultrasound technology and artificial intelligence (AI) could reduce paediatric pneumonia, while actively tackling the overuse of antibiotics in treating it. The IMCI-PLUS international research consortium hopes to improve early and accurate diagnosis of childhood pneumonia in primary care hospitals and clinics, where resources are limited and diagnostic tools are scarce.
In many current contexts, clinicians face challenges in deciding whether an antibiotic treatment is needed as they have to rely solely on clinical signs like cough and difficulty breathing in a specific patient. Respiratory infections such as pneumonia in children can often have viral causes which do not respond to, and thus do not require, antibiotic treatment in the same way that bacterial infections do. In addition to the burden from pneumonia itself, this creates another important issue: antibiotic overprescription.
Introducing the IMCI-PLUS Consortium: an international collaboration
In response to these challenges, the IMCI-PLUS consortium, will use the well-established Integrated Management of Childhood Illness (IMCI) framework to confront diagnostic and treatment challenges. As pointed out, to better and more appropriately treat children with pneumonia, improved diagnostics are the key. The consortium hopes to show that Point-of-care Lung Ultrasound (PLUS) can be a vital tool in the clinical arsenal to assist in managing cases of paediatric pneumonia.
“We have long faced two challenges, which are unclear diagnoses at the patient’s bedside, and the continued overprescription of antibiotics, despite the fact that most pneumonia in children is not bacterial,” said Dr Kristina Keitel, scientific lead of the IMCI-PLUS consortium, based at the Swiss Federal Technology Institute of Lausanne (EPFL).
The first randomized controlled trial led by the consortium has commenced in five countries, with over 10,000 lung ultrasounds expected to be administered. 3500 children diagnosed with pneumonia will also take part in the trial. IMCI-PLUS is supported by a network of institutions in Switzerland, The Netherlands, South Africa, Senegal, and Tanzania- with the latter three countries hosting the clinical trial itself.
Portable ultrasound and its promise
Keitel reiterates the importance of newer and more accessible diagnostic tools, like portable ultrasound machines. “The lung ultrasounds will also help healthcare workers more accurately distinguish bacterial and non-bacterial infections. This will help identify complications or other conditions, like tuberculosis, earlier.”
Dr Jacques du Toit, an IMCI-PLUS project leader based at the SAMRC/Wits University Agincourt Research Unit, explains that there is a global move in clinical care to develop better point-of-care triage tools. “Ultrasound is emerging as a very sensitive and specific tool. It is low-cost, radiation-free and doesn’t require consumables,” he said.
Using AI to overcome technical barriers
Technical barriers remain, as performing and interpreting ultrasounds requires specialized skills. The IMCI-PLUS team aims to overcome this hurdle using artificial intelligence (AI). AI models will be trained using existing lung ultrasound images, along with some anonymous health data about each child, like age, weight, HIV status, symptoms, whether they were given antibiotics and their recovery after treatment. The AI tools developed will support healthcare workers in interpreting ultrasounds, which will then help them with prescribing antibiotics or suggest another treatment regimen.
Context-specific AI for local health needs
“We believe that AI has the potential to bridge the skills gaps and help decentralize access to lung ultrasound in places where there are no skilled personnel to undertake them,” said du Toit. He adds that unique AI models will be trained to respond to specific contexts as each country will have different disease burdens, clinical profiles and ultrasound artefacts.
“For example, we may have children with advanced HIV or malnutrition. Models not exposed to these health concerns might misclassify children in an automated response. We thus need to train AI models on local data to be accurate and relevant.”
Accelerating the path from research to practice
IMCI-PLUS’s broader aim goes beyond improving diagnostics; the project wants to realize translational research in a much shorter time-period. This will be a challenge in the context of notoriously long lag times between research and real-world implementation. “We’re trying to run policy engagement and research in parallel, not sequentially, to avoid the 15-year delay we usually see from research to practice,” Keitel notes.
Cross-sector partnerships to anticipate challenges and future needs
What excites Keitel and du Toit is the intersection of clinical care, technology and multidisciplinary partnerships. The consortium also extends to include a host of researchers, policy makers and healthcare workers across countries.
The project is funded by the Global Health European and Developing Countries Clinical Trials Partnership (EDCTP3) research and innovation programme, a joint African-European initiative focusing on African health priorities. This EU funding initiative aims to deliver new solutions for reducing the burden of infectious diseases in sub-Saharan Africa - while also hoping to strengthen research capacities, preparation and responses to emerging infectious diseases in this region and across the world.
This partnership is unique in its commitment to equitable research leadership, with African institutions playing central roles in design and delivery. For a device to reach scale, it must be evidence-based and approved for the WHO’s essential devices list, which relies on strong, collaborative science.
With a multidisciplinary approach, the consortium brings together computational scientists, social scientists, epidemiologists, clinicians, and policy experts.
Africa-led research for local impact
“There is a real shift towards empowering local researchers to lead,” says du Toit. “It’s not just research on Africa; it’s research led from Africa. There’s now significant capacity and leadership in Eastern and Southern Africa to run high-quality research that responds to local needs.”