New deep learning approach can diagnose malaria as well as human experts
This World Malaria Day (25 April), we explore new deep learning approach for malaria diagnosis developed by team from UCL Computer Science in collaboration with College of Medicine University of Ibadhan
Quicker, more accurate malaria diagnosis will enable faster delivery of clinical services to facilitate International Development Goals for both sub-Saharan Africa and other regions of the World affected by malaria.
Access to effective malaria diagnosis is a challenge faced by all developing countries where malaria is endemic. Human-microscopic examination of blood smears remains the ‘gold standard’ for malaria diagnosis and despite its major drawbacks, other non-microscopic methodologies have not been able to outperform it. Presumptive treatment for malaria –i.e. without microscopic confirmation of the disease – is wasteful of drugs; ineffective if the diagnosis was wrong; a drain on limited healthcare resources and fuels antimalarial resistance. This scenario has prompted Global Health organisations to emphasise the urgent need for tools to overcome the deficiencies of human-operated optical-microscopy malaria diagnosis and other non-microscopic tests.
An international cross-disciplinary team led by Professor Delmiro Fernandez-Reyes at UCL Computer Science, in equal collaboration with the College of Medicine of University of Ibadan (COMUI), Nigeria, have develop a deep learning system that achieves expert-level automated malaria diagnosis on routine blood films. The work, “Expert‐Level Automated Malaria Diagnosis on Routine Blood Films with Deep Neural Networks” is now published in the American Journal of Hematology (doi: 10.1002/ajh.25827).
The AI work is a significant step towards scalable automated systems capable of reliably diagnosing malaria in sub-Saharan West-Africa by using routine and ubiquitous blood films. The work shows that AI approaches are capable of tackling the challenges posed by human-operated light-microscopy currently used in the diagnosis of malaria. The diagnosis system harnesses AI approaches to support malaria care clinical pathways in large urban sub-Saharan settings such as the city of Ibadan, a 3.5 million inhabitants’ metropolis in Nigeria, where the study was conducted.
Dr. Petru Manescu from UCL Computer Science and first author of the work commented: “Visual inspection of malaria thick blood films strongly relies on the availability of trained personnel; it is time-consuming and subject to human error caused by fatigue and cognitive overload in busy clinical-microscopy services. Deep learning has the potential to overcome these challenges and increase the throughput of patient diagnosis in the clinical pathways and allowing pathologists to concentrate on higher level consultative tasks.”
Professor Biobele J. Brown from the College of Medicine of University of Ibadan and co-author of the study commented: “In Nigeria alone, malaria is one of the most common causes of death in children below five-years of age. Rapid, reliable and accurate diagnosis of the disease, giving way to prompt, adequate treatment, is a crucial challenge for fulfilling the International Development Goals and impacting on the sub-Saharan African region and other regions of the World affected by malaria.”
Professor Delmiro Fernandez-Reyes commented: “Our novel approach aims to translate AI research into a scalable and deployable automated system capable of reliably performing the tasks required for rapid malaria diagnosis on routine blood films. Such systems will have an impact on already-stretched clinical services in the region by enabling healthcare providers to divert human and financial resources to further improve healthcare provision to those more affected by malaria: pregnant women and children. The work provides a clear example of how AI, engineering and digital-technologies research aligned with clinical research can underpin successful and sustainable healthcare provision in the sub-Saharan region and therefore significantly improve Global Health.”
The work was supported by the College of Medicine of the University of Ibadan, Ibadan, Nigeria; the UK Medical Research Council (MC_U117585869); Department of Computer Science, Faculty of Engineering Sciences of University College London, United Kingdom and UK Engineering and Physical Sciences Research Council (EP/P028608/1).
UCL Computer Science is bringing together UCL’s wealth of cross discipline intellectual capital, to find innovative, workable solutions to Global Health problems. The College of Medicine of the University of Ibadan is one the strongest academic environment in the region providing research and training to professionals for the whole sub-Saharan Africa region and internationally. The partnership has resulted in the creation of a joint African Computational Sciences Centre for Health and Development (https://african-cschd.org) with a vision to tackle intractable global challenges of African developing countries by means of both inter- and multi-disciplinary research that leads to the development and deployment of innovative sustainable and scalable computational solutions to these problems.
Malaria is a mosquito-borne infectious disease affecting humans and other animals. It is caused by parasitic protozoans belonging to the Plasmodium species. Human malaria is caused by five Plasmodium species (P. falciparum or -lethal malaria-; P. vivax; P. ovale; P. malariaeand P. knowlesi).
Malaria is widespread in the tropical and subtropical regions including much of Sub-Saharan Africa, Asia, and Latin America. In 2017, there were an estimated 219 million cases of malaria worldwide resulting in around 435,000 deaths. Although current malaria control strategies have resulted in speculative estimates of a decrease on mortality rates, success on reducing the burden of disease Worldwide has remained largely elusive.
Life-threatening Plasmodium falciparum malaria is still a major cause of mortality in sub-Saharan Africa, and together with Tuberculosis and HIV, remains a primary Global Health Challenge in the region. Up to eighty-five percent of the cases worldwide occur in sub-Saharan Africa with about 90% mortality in the under five years–of-age group due to severe malaria syndromes. Large all-year-round lethal malaria morbidity and mortality burden has severely hindered the wellbeing and the socioeconomic development of the West sub-Saharan geopolitical region.