Ninety per cent of malaria-related deaths occur in Africa, in part due to a lack of high-quality diagnosis in remote communities. Amref Health Africa is partnering with a social enterprise startup to transform smartphones into affordable, easy-to-use, reliable tools to diagnose malaria. If the pilot proves successful, it has the potential to reduce malaria-related deaths by 1.5% through better diagnostics alone.
Project timeline: April 2017 to March 2018
Funder: This project is supported by Grand Challenges Canada.
Situation – Children are dying of malaria every minute
Every minute a child somewhere in the world dies of malaria; and 90% of these deaths occur in Africa. Currently there is a big gap in the availability of high-quality malaria diagnosis, especially at the community level in African countries.
In partnership with social enterprise startup Mobile Malaria Labs (MOMALA), we will test how to help close this gap. Early, accurate diagnosis and treatment give the best chance of survival. This project will turn smartphones into tools that quickly diagnose malaria using a proprietary algorithm and microscopic images of blood films to diagnose malaria with an accuracy of a World Health Organization-certified level one microscopist (lab technician that specializes in confirming the presence of malaria in blood) and compare the results to methods that take longer to perform.
Action – Introducing a malaria diagnostic method and a smartphone app
To improve malaria diagnosis:
- 300 suspected malaria cases will be tested in different levels of the health system (county, sub-county, hospital and health centre) in two high-transmission zones in Kenya
- We will train Kenya Ministry of Health lab staff on how to use the app. These staff will use standard microscopic examination to examine blood slides, and then re-examine the slides using the app. All slides will then be re-examined by a World Health Organization-certified level one microscopist
- The objectives of the pilot are to test how the app compares to standard microscopic examination; whether time spent by lab technicians on diagnosing malaria decreases; and whether the level of accuracy in diagnosis throughout the day is improved (addressing the potential decline in accuracy due to staff fatigue).
“MOMALA is very strong in machine learning, image recognition and mobile app development, but is not the expert in malaria and the African health market. We decided to partner with Amref Health Africa because it has exactly this expertise.” Bram den Teuling, Data Scientist at MOMALA
Results – Saving lives through improved malaria diagnosis
- 1.5% decrease in the number of deaths caused by malaria
- 2% reduction in malaria prevalence rate (the number of people in the population who have malaria at a given time)
- 997 people, including 235 children under the age of four and 272 children between ages 5 and 14, reached with the smartphone malaria diagnostic app
Ethical and Scientific Approval Obtained: Studies using human subjects require approval from a scientific and ethics review committee to ensure the study design is scientifically sound and meets ethical standards. The scientific protocol for this project—which includes elements such as the reporting tools for data collection and consent forms for participants—has received approval from the Amref Health Africa Ethics and Scientific Review Committee. Our partner, MOMALA, has received additional approval from the Dutch Medical Ethical Committee.
Government Approvals Obtained: County governments in which the study sites are located have given written approval and are co-investigators in the study. Kenya’s National Malaria Control Programme, run by the Kenya Ministry of Health, has been informed of, and agrees with, the study aims and objectives.
Study Sites Selected: We have selected sites to participate in the study based on the number of staff and the number and types of microscopes at each site: one site in western Kenya and one near Mombasa will carry out the study.
Machine Learning Algorithm and Dataset Being Developed and Tested: The MOMALA app is a state-of-the-art innovation and significant parts of the technical development have never been done before, so further app development is underway. The app uses machine learning, in which an algorithm is trained using a training dataset, which the algorithm uses in order to “learn” how to make predictions and diagnoses; then another dataset is used to test the algorithm for sufficient accuracy and precision. The project team has photographed blood slides containing the malaria-causing parasite to build this test dataset. The project team will test the machine learning algorithm, and through deep learning, it will improve its predictions and diagnoses until it is on par with the highest WHO standard.