ByteSight

ByteSight is an imaging field tool that leverages computer vision to empower communities to fight malaria.

Rebecca Rosenberg

Almost half of the world’s population is at risk for malaria: a devastating disease that disproportionately impacts pregnant women and children under five years old. Approximately 90% of the global burden of malaria occurs in sub-Saharan Africa (SSA). Despite over $2B US dollars spent annually on the malaria problem, there were still 228M cases in 2018, resulting in over 400,000 deaths. The majority of malaria aid money goes to diagnostics and treatments, but the most effective way to fight malaria is to prevent it in the first place: by controlling dangerous mosquito populations.

Due to their differences in behavior and biting patterns, it is crucial for control programs to understand the local makeup of mosquito species in order to target interventions most appropriately. Only 40 out of thousands of mosquito species are capable of transmitting malaria and surveillance programs monitor the density and distribution of these dangerous mosquitoes through vector surveillance. 

Vector control relies heavily on the rapid and accurate identification of malaria-carrying mosquitoes. Currently, field workers must visually identify species, sometimes taking as long as 20 minutes to process a single mosquito. Due to a lack of qualified entomologists, the accuracy of such identification is as low as 50-66%.

The target market for the ByteSight technology is vector surveillance programs in sub Saharan Africa. These organizations exist to acquire surveillance data for governments, and employ both minimally trained field workers, and highly trained entomologists. Due to the training required, entomologists in these areas are often in short supply, which leaves mosquito speciation to more minimally trained workers. These workers require more time per mosquito, and are prone to misidentification. 

In August of 2019, the ByteSight team travelled to Zambia and Uganda to meet with stakeholders and potential customers. The team visited nine mosquito surveillance sites and conducted over 65 interviews with stakeholders including field workers, control program managers, research directors, and ministry of health employees. Our prototypes are currently being field-tested for efficacy, usability, and adoptability in Macha, Zambia, and further studies are planned for Ghana, Kenya, Angola, and Cote d'Ivore.

Morphological identification of mosquitoes is the critical bottleneck to deploying strategic vector control interventions. Moreover, vector control, in the form of bed nets, indoor residual spraying, larviciding, and environment management, has been shown to significantly, if not completely, limit malaria transmission in an area. When malaria transmission is lowered, there is lower morbidity and mortality, as well as a reduction in the economic burden imposed by malaria.

Our solution creates tangible impact for multiple stakeholders. Vector surveillance programs, which will adopt and use the solution, benefit in terms of labour costs, scalability, and accuracy. Our solution doesn’t require a trained expert, enabling surveillance programs to employ local workers, while achieving higher performance. The reduced need for experts, as well as the estimated cost savings in employing lay workers and in increasing the identification throughput, will allow programs to scale up their number of sentinel sites, therefore increasing the granularity and representativeness of their data. The improvement in the quality and quantity of the entomological data will then empower those who manage intervention implementation. Decision makers can ensure interventions are properly tailored to the needs of the local community, and can adjust interventions if they are not working as expected. 

In comparison, current practice in many places does not allow for dynamic adjustment of interventions, as entomological data is often not available until months after it is relevant. In turn, we expect this to benefit everyone in the communities where the tool is used- by reducing malaria transmission. 

The ByteSight team is currently finalizing our MVP, and prototypes have been sent to our partner site in Macha, Zambia. In the next one year, our team plans to deploy our technology at additional partner sites in Cote d’Ivore, Ghana, and Camroon. With the usability and field accuracy data collected during these trials, the computer vision algorithm will be updated to include additional mosquito species that are relevant to various parts of sub-Saharan Africa. 

In three years, the ByteSight team plans to have a final design of the technology, which has been deployed well beyond partner sites.

The ByteSight team is initially measuring success of our technology based on the increase in speed and accuracy of mosquito identification at our partner field sites. 

As our technology becomes adopted, we will additionally be following the effect the system has on malaria outbreaks in local areas.

In the next year, we expect the challenges associated with COVID-19 to be our biggest barrier. In order to overcome these challenges, the ByteSight team has worked to build out partnerships in different parts of sub-saharan Africa, who will be able to beta test our technology even if our team is unable to travel.

In addition to the aftermath of COVID, we expect our biggest challenge in the next 3 years will be promoting buy in at the national level, such that we are able to create a comprehensive vector surveillance system that can be accessed at all levels. In order to promote this, we plan to work with many countries to begin to build out “social proof.” Additionally, we plan to build credibility through publications regarding the engineering innovations, scientific results, and public health effects.

Johns Hopkins Center for Bioengineering Innovation and Design

Macha Research Trust, Macha, Zambia

Bill and Melinda Gates

Ministry of Health in a country in SSA