Developing new approaches to analysing medical images in low-resource settings

Oxford University medical science researchers and medical staff of the department of Radiology at University College Hospital, Ibadan, Nigeria, used new approaches to analyse brain scans of stroke patients taken with lower quality technologies, offering potential for research and clinic practice in low-resource settings.

Researchers at a workstation at University College HospitalResearch workstation at University College Hospital, Ibadan, Nigeria.
The burden of stroke is increasing exponentially in Africa, due to the ageing of the population, lifestyle changes and increasing cardiovascular risk (e.g. high blood pressure and diabetes). However, data from patients of African origin is lacking in disease research, and diagnosis and treatment often relies on low-quality brain images.

Whilst high-quality Magnetic Resonance Imaging (MRI) scans are most effective at identifying stroke affected tissues and lesions for research and clinical purposes, most contexts in Africa only have access to Computed Tomography (CT) and low-field Magnetic Resonance Imaging. Such technologies are usually less effective and rely on skilled radiologists to interpret the images.

But biomedical researchers in the Wellcome Centre for Integrative Neuroimaging at Oxford University, have applied machine learning techniques to such lower quality images, to test whether these can reliably analyse stroke patient brain scans. Funding from the Knowledge Exchange Seed Fund allowed them to test and share the approach with colleagues from the University College Hospital, Ibadan, Nigeria in summer 2022.

“We tested the algorithms on 40 patient scans from the Stroke Investigative Research and Educational Network (SIREN) project, the largest stroke database in sub-Saharan Africa, which had been generated by CT or low-resolution MRI scanners.” says Dr Ludovica Griffanti. “We were delighted to find that, with some modifications, the approaches worked promisingly on this small selection of scans.”

The team shared the approach with the Nigerian clinicians and trained them to use the software and hardware provided by the grant. The Nigerian partners have subsequently trained other colleagues at Ibadan in the approach and are using it to analyse a wider selection of scans to verify the findings.

Whilst in Oxford, Dr Godwin Ogbole also developed a lecture for the Oxford Clinical Neuroimaging course on using brain imaging for diagnostic purposes in low-resource contexts. Dr Ayilara shared preliminary results of the project at an oral presentation to the Association of Radiologists in Nigeria (ARIN) conference in October 2022, winning an award for best presentation.

“This was a really fruitful knowledge exchange with our partners in Nigeria,” says Dr Griffanti. “The pilot showed that our approach has real potential to effectively analyse brain images from low-resource contexts – which could add to our understanding of stroke in patients of African origin. And these techniques could ultimately be helpful in contexts where a radiologist is not available, or teams are working with limited equipment and technology. The methods developed also have the potential to create products that would be useful in the UK health system and in other countries.”

“As a researcher, I’m very aware of the privilege of working with the best imaging technology, and of the gap between what we have here at the University, and what is available in many hospitals. Approaches that help bridge the gap between research and the clinic have huge potential, and we are looking for further funding to develop this work over the next year.”

“The next steps will be to further test the approach and extend it to the broader SIREN dataset covering over 3000 patients. Ultimately, we hope to enable the development and evaluation of machine learning approaches to establish automated diagnostic standards for stroke and other neurological conditions. This has the potential to enhance global research and clinical decision making in stroke care and influence interventions and outcomes, particularly for some of the most vulnerable populations.”

Dr Ludovica Griffanti is an Alzheimer’s Association research fellow at the Wellcome Centre for Integrative Neuroimaging, Department of Psychiatry, University of Oxford.
Dr Godwin Ogbole is Associate Professor of Radiology at University of Ibadan and Consultant Neuroradiologist at the University College Hospital, Ibadan.
Dr Segun Ayilara is a senior registrar at the University College Hospital, Ibadan
Oxford University collaborators on the project: Dr Pedro Diniz, Dr Taylor Hanayik, Dr Nele Demeyere, Prof Mark Jenkinson, Prof Sarah Pendlebury.

Funders: KE Seed Fund

Dr Ogbole’s six-month visit to the Nuffield Department of Clinical Neurosciences was funded by the Africa Oxford Initiative.

Dr Griffanti is supported by an Alzheimer’s Association Research Fellowship and by the National Institute for Health and Care Research (NIHR) Oxford Health Biomedical Research Centre (BRC).

The software development was supported by the John Fell Fund, the NIHR Oxford BRC, and the Wellcome Centre for Integrative Neuroimaging.