Fighting malaria with drones

Malaria is a disease caused by a parasite. The parasite is spread to humans through the bites of infected mosquitoes.

Each year nearly 290 million people are infected with malaria, and more than 400,000 people die of the disease.

Mosquitoes fly in swarms and can reach millions of individuals after hatching. Fighting and eradicating them is a painful, lengthy, and expensive process. Therefore, most remedies were targeted to finding a cure, or a vaccine to fight the virus itself.

What if we can use non-traditional methods to eradicate the virus altogether before it reaches humans, and before the mosquitoes even hatch!

Vector control in fighting malaria is an idea that focuses on killing the mosquitoes that are carrying the virus before they transmit it to humans.

One method of disease control is to decrease the mosquito population by modifying larval habitat, known as larval source management (LSM). In malaria control, LSM is currently considered impractical in rural areas due to perceived difficulties in identifying target areas.

The increasing availability of high-resolution drone mapping, however, is creating solutions to take high-quality images from the sky and overcome the barrier in identifying larval habitats, quickly, and at a very large scale.

 Crowddroning by GLOBHE  local pilots worked hard in Malawi to capture critical drone data that is helping researchers and national authorities to map larvae habitats.

We are employing local pilots for the Maladrone project via the ‘crowddroning’ organisation GLOBHE. As our team consists of epidemiologists and entomologists with limited drone skills, our experiences so far indicate that combining local drone expertise with the needs of national malaria and vector control programs is the way to go. Liverpool School of Tropical Medicine

What are drone maps? How can we use them and how can they help us locate mosquitoes?

Drones take a large number of high-resolution photos over specific areas. According to the project, drones can be equipped with other sensors to gather other types of data, such as thermal imagery and LIDAR. In our case, only images are needed.

These captured images, however, need to be processed before starting to take information out of them. Below is a detailed explanation of the workflow.

An orthomosaic 2D map is like stitching all images together to create a large image that is geotagged and has both latitude and longitude information.

The resulting map is a high-resolution and recent representation of a geographic location, containing much more valuable information than an outdated low-resolution satellite image.

The ecology of preferred breeding grounds for mosquito oviposition varies both within and between species. The best sites are usually shallow and calm vegetated water are optimal sites for larval development

In the image below, kids are playing in what could very likely be malaria mosquito breeding sites, which is putting them at great risk of contracting malaria and ending their life.

Once the data is processed, identification of the different visual elements in the image begins. This includes classifying the different types of vegetation, water type, and soil, to identify likely mosquito breeding sites.

One way to do that is through manual work. This means that scientists identify the vegetation type from the image and draw a GIS map out of this. An example is provided in the map below. This could work for small datasets because it is time-consuming and could be prone to mistakes.

If we are trying to analyze images at a large scale, classifying the vegetation manually becomes a very challenging task. Therefore, we use computer algorithms to solve this issue for us. One solution is to use a classifier algorithm to do identify the vegetation type automatically, and classify them into different groups.

After classifying the type of vegetation in drone images, we can isolate the areas that are typical of mosquito breeding sites which mostly correspond to confined shallow waters with emerging aquatic vegetation. These locations are then placed on a map and communicated to local authorities to take actions

The map contains the exact location of the suspected water bodies and their size. This map is then delivered to the client, which could correspond to NGOs, UN organizations, or local authorities, who will use this information to plan a quick intervention program to eradicate the larvae before they become mosquitoes and spread malaria. By doing so, they gain momentum against the spread of the virus and save millions of dollars in indirect costs after the spread of the virus.

This new technology has proved to be very effective against Malaria and very scalable worldwide. It is not possible to adopt this workflow with satellite images because their resolutions are too low and therefore could not spot mosquito breeding sites as drones do.

The vegetation cover and the water body are continuously changing from month to month, and throughout the year and therefore regular inspection of the same lake areas should be undertaken and the workflow repeated to keep targeting the new larvae. Drone technology is highly robust and drones can be quickly and repetitively deployed to scan these areas, while the images are simultaneously fed to AI models for quick analysis and map delivery to clients.

Scale-up is very possible with  Crowddroning by GLOBHE  because it connects thousands of pilots that are capable to perform this work globally. The impact on the local communities is not only that this helps eradicate Malaria and saves tax money to be spent on other development projects but it also directly creates jobs for local drone pilots and helps the local economy.