UCL Computer Science


AI tool enhances early detection of severe malarial anaemia in children 

27 June 2024

UCL and University of Ibadan researchers have created an AI-driven model to detect severe malarial anaemia early, aiming to reduce child mortality in sub-Saharan Africa. 

vials containing blood for analysis in a lab

Severe malaria anaemia (SMA) is a life-threatening condition predominantly affecting children in sub-Saharan Africa requiring urgent blood transfusion and anti-malarial treatment.  

Traditional methods for diagnosing SMA detect this severe presentation when is already established and rely on labour-intensive and variable microscope assessments. In low-resource settings, this hinders early referral to blood transfusion-capable healthcare centres. 

UCL Computer Science and College of Medicine of the University of Ibadan researchers have developed a deep learning model, Multiple Instance Learning Framework to Identify SMA (MILISMA), that analyses routine peripheral blood films to identify specific changes in red blood cells characteristic of SMA.  

The approach provides an accurate, scalable and consistent diagnostic tool, potentially transforming early detection of children progressing towards the severe anaemia life-threatening clinical presentation.  

Commenting on the research, Professor Delmiro Fernandez-Reyes (UCL Computer Science) said: "Severe malaria anaemia is a major cause of mortality in children under five in sub-Saharan Africa. MILISMA's ability to accurately detect red blood cell changes at scale will greatly enhance our ability to early detect those children progressing towards this severe malaria syndrome. This in turn will facilitate early clinical management thereby reducing the mortality rates, particularly in those low-resource healthcare settings with no access to blood transfusion.”". 

Dr Petru Manescu (UCL Computer Science) added: "Our deep learning model, MILISMA, not only improves the accuracy of SMA diagnosis but also provides a valuable tool for understanding the underlying pathology of the disease. This innovation could pave the way for studies that explore effective treatments and interventions." 

Professor Biobele J. Brown (College of Medicine of the University of Ibadan) said: “Our model bridges the gap between technology and healthcare, ensuring that even in resource-limited settings, we can offer accurate and timely diagnoses to those in need."