Data Science Talent Logo
Call Now

How AI Is Driving the Eradication of Malaria By Arnon Houri-Yafin

Arnon Houri-Yafin is the Founder and CEO of Zzapp Malaria, a developer of software tools for malaria elimination. Arnon’s prior roles include Director of Research at Sight Diagnostics and Lecturer in Statistics at The Hebrew University of Jerusalem. His specialisms include data analysis, statistics and machine learning.
In this interview, Arnon discusses the breakthroughs Zzapp Malaria has made in the fight against this deadly disease. He explains how AI and data modelling have played crucial roles in developing Zzapp’s strategies:

So, if we start with your personal journey Arnon, could you tell us how you ended up working in both AI and in the field of malaria eradication?I have two answers. One is a personal story, where my friend took me to see malaria first-hand and then another one is poverty.

Back when I had to decide what to study in university, I chose economics. My goal was to become an economist in the context of developing countries, and specifically African countries. I believed that reducing poverty was the key to significantly improving the wellbeing of whole communities.

Just before I completed my MA, a friend called and told me that he was starting a company (Sight Diagnostics) dealing with malaria. I knew at the time that malaria was a huge problem in terms of public health, and that it’s one of the primary reasons for poverty. The disease prevents people from going to work, either because they are sick or they attend to sick family members, which reduces productivity. Malaria is also one of the primary reasons why children in Africa miss school, which obviously harms their ability to progress.
Sight Diagnostics developed a malaria diagnostics device, and one of my jobs was to test them. We ran these tests in public hospitals in Mangalore and Mumbai, India.

In India, you have some places with a lot of malaria and others without. I was in an area with a lot of malaria. Now, I’m a nervous parent – when my children have a fever; I get really stressed out. But then you see the moms with young children who are very feverish, they are scared, because it’s not just a fever that will probably be gone by tomorrow, it’s malaria. And when I saw that, the difference between malaria diagnostics and malaria elimination struck me.

In many countries, Israel for example, malaria was a problem – a big problem. Here, malaria was all but eliminated in the 1920s and 1930s, after the stagnant water bodies where the Anopheles mosquito breed were thoroughly targeted. So, if we have such a big problem, but one that could be fully eliminated, why don’t we do it? This is what caused me to say, okay, diagnostics are very important, but we need something more radical. We need to recreate in modern Africa the successful malaria elimination operations of Cyprus, Egypt and many other countries.

Is that why you specifically founded Zzapp Malaria?

Exactly. Zzapp Malaria is about moving from malaria control to malaria elimination. This takes artificial intelligence and data. When people tried to treat water bodies in Africa, they did it with partial success. This is because tropical Africa has two rainy seasons and there are wide areas that must be searched for water bodies that once detected need to be treated regularly. A very high percentage of water body coverage is required, which is a difficult result to reach.

With our system, fieldworkers go into the field with a smartphone that guides them exactly to the areas they need to search, enables them to upload information about the water bodies they find, and, once water bodies are detected, shows them which water body has already been treated, which needs to be treated again, which houses should be sprayed etc. So we really have a lot of information about the exact location of water bodies and the overall situation of the operation in terms of the treatment of houses and mosquito breeding sites.

Did the company have a concrete idea already?

Did they have a team? How did you start this up? Where were you when you started the company?

We actually started with an app for house spraying, but quickly moved into larviciding (the treatment of water bodies). Today, our system combines a lot of different features that, combined, are meant to provide the most cost-effective solution per location. The ideas and the team were built as we progressed. We have been receiving a lot of help from many people. It is nice to see that people understand the importance of malaria, and really try to do their best to fight it.

What were the backgrounds of the other members?

Zzapp has a strong and dedicated team with expertise in science, technology implementation, marketing and community engagement. We have leading software developers, PhD holders in veterinary medicine and communication, and team members with extensive experience in leading field operations, against mosquitoborne diseases and other impactful causes, in Africa and elsewhere.

You had a diverse set of skills when you started out. Did you have a very concrete idea already?

We started out with a clear goal that is still with us: eradicating malaria. We knew, almost from the start, that the key is locating and treating the water bodies. We have since improved our ability to do so, and have integrated in our operations other methods, but yes, we are still big believers in larviciding.

That sounds interesting. So, you stuck with your original ideas – more or less – to implement and to turn your idea into reality rather than to do a lot of research in 2016? Can you explain a little bit about how the company grew and how the project grew over time?

Malaria is unique in that the means for eliminating it are known. It’s not like having to develop a new medicine. So the trick is to think of the best waysthat are the most cost-effective ways – to do so under challenging conditions and with limited resources. So that is where our R&D goes. We began with a small budget from our parent company Sight Diagnostics, and a grant from the Israel Innovation Authority (IIA). Working lean, and gradually hiring more staff, we collaborated with leading scientists, for example Andy Hardy with whom we operated in Zanzibar. In that project, which was funded by IVCC through the Bill and Melinda Gates Foundation, we used drones to map water bodies.

Winning the grand prize in the IBM Watson XPRIZE AI for good provided us with $3 million, which is obviously a significant improvement in terms of our resources. We were able to hire additional staff and push our technology forward. It also helped us create connections with African governments. We have been going to malaria conferences in which we present our technology to malaria researchers and implementers. We have also been contacting government ministries of health and proposing our solution. So far we have been working in several African countries, including Ethiopia, Kenya, Ghana, Tanzania, Mozambique and São Tomé and Príncipe.

Can you explain the core idea that you had to combat the spread of malaria? If I understand correctly, it has to do with identifying stale pools of water? How do you prevent the spread of malaria?

In the past, the best solution to fight malaria was treating the stagnant water bodies. But today in Africa, most countries prefer bed nets. You distribute them and hope that people sleep under them. It’s a simple and effective tool, which indeed protects many people. However, it depends on peoples’ behaviour, which as we know from mask mandates during COVID, is unfortunately not always reliable. We want to make use of the proven method of larviciding and apply it in Africa.

To do so successfully, you must overcome two problems. One is coverage: you need to find all of the water bodies. The second problem is budget. Because you don’t have the money to scan every single square kilometer in Africa, you must prioritise. Based on satellite imagery and data on topography and rainfall, our system decides when and where to scan for water bodies. This is how we reduce operation costs. The mobile app helps coordinate the operation and the monitoring. We make sure that all the areas that were assigned for scanning were searched, and that all of the water bodies and the houses that were selected were treated.

So, you’re basically saying that using satellite data, and then also artificial intelligence, you can pinpoint people much more precisely to those water bodies that need to be treated, and by increasing the spacial distance coverage, basically you can eliminate malaria better? What was the idea initially? How do you scale this across a continent like Africa?

Yes, treating the water bodies is the idea, and scale is the challenge. Besides technology, it’s about finding good local partners, be it the government or a large NGO. We provide the technology and training, but it is the fieldworkers and the local management who have the ownership over these operations. Ultimately, they are the ones who serve their own communities.

Therefore you are bringing the technology, and the partners are bringing the people to make it scalable?

Yes. We bring technology and we also bring knowledge about which agent to put in the water. We are experienced because we did it in other countries, then we go to new countries to share our knowledge.

You already mentioned satellite data. There’s obviously a lot happening in that space at the moment. Can you talk a little bit about which kind of data you’re using and how you’re using the satellite data?

One of the challenges that we faced is resolution. If you have a large water body, you just see it from the satellite imagery. It then becomes a standard machine vision problem and you also have the near infrared, a channel where water is very distinct. But then you have water bodies that are smaller than your resolution. So it’s not about finding the water bodies themselves, but rather the areas that are suitable for water bodies. This is interesting because if you go to look at something smaller than your resolution, the context becomes very important. For example, you won’t find malaria mosquitoes in a river because they only breed in standing water. However, near that river, you’ll probably find many water bodies.

We started with a small neural network that detected only what we wanted it to detect. We then enlarged the networks so as to get better usage of the context of the area. It’s similar with topography (and this isn’t even artificial intelligence, it’s more traditional models of topography): you have water, going from a high area to a low catchment, so you can understand that it’s not only the absolute height of a specific point on the map, it’s also how it relates to other points in the area. We use both traditional models of topography and a neural network to understand where the water bodies are likely to occur. Again, if you give it more areas, it works better.

You mentioned you’re using conventional neural networks, basically to identify areas where there could be water bodies… what are the inputs for those models? You mentioned satellite data, but you’re not only using visible light channels, but also infrared – anything else?

We’re using satellite data, then topography, and also land use. We also use data on rain and humidity. For that, as part of our prize, IBM Watson also helped us in machine learning projects. They focused on the temporal part of it, because you need to know not only where the bodies of water are, but also when.

Obviously, at the start of the rainy season, you will find many more water bodies than after the dry season. But knowing the exact amount of rain, humidity and temperature, helps to better predict the abundance and location of water bodies.

That’s quite sophisticated modelling that you’re doing there already. For the data sources, are they open source or do we have to buy this kind of data?

This is up to the country we work with. There is free satellite imagery produced by the European Union, but also high resolution satellite imagery that one can buy. We help countries understand the pros and cons of each and make a decision. So far, we mostly focused on the low resolution part because it will be more convenient for many of our potential customers. We hope to gain more experience with high resolution satellite imagery.

Basically then, this model takes all of this input data and then predicts where water bodies could be, and that information goes in through an app to the workers in the field. Is that correct?

Yes. We have one more layer, which is locating the houses. If you have a water body in the middle of the jungle, you don’t care because malaria is transmitted only from person to person. The mosquito is merely the vector. So, we have one component that maps the water bodies and then one component that maps the houses.

The last component is about the proximity of houses to areas that potentially have many water bodies in it. Then, based on that, and based on their allocated budget, we define where to scan for water bodies. We then have one more component that takes it to the mobile app.

Malaria attacks poor communities and then prevents them from getting out of their poverty because it’s difficult to go forward. It’s a vicious cycle of poverty and disease that we want to help solve.

And so these workers, what do they then do? Do they take the app, then take the information, and then they go to these locations and then they treat the stale water bodies with chemicals?

Yes, they search for the water bodies. If we found water bodies from a drone or from satellite imagery, we can just direct them, but they can also report water bodies from the field. This is important because that’s how we feed the system. It’s machine learning and we need new data. That’s one thing. Then, they treat the water bodies. Treatment of the water bodies in the past involved chemicals, but this isn’t good because animals, and sometimes even people, drink from the water bodies.

The World Health Organisation has very strict regulations about which materials can be put into the water, but mostly we use a biological agent that is called Bti (bacillus thuringiensis israelensis). The good thing about Bti is that except for mosquitoes it doesn’t harm other animals. Not people, not cows, not frogs – not even other insects. Only mosquitoes and black flies (that transmit river blindness). So, it’s very environmentally friendly.

Can you talk about some specific user cases or field tests that you did with communities?

One interesting thing that we’ve done is with an NGO in Ghana. They fight malaria in their town and in the villages surrounding it, and they were very successful before us. They sprayed houses and did community activities with the village inhabitants, because education is also important so that people use the bed nets and go to the doctor if they have a fever.

They wanted to achieve zero cases, and so they approached us and together we implemented an operation against the water bodies. As I said, the very interesting part is that we managed to reduce more than 60% of the mosquito population in the town and the villages, which is an outstanding result.

It was a controlled trial: some areas we treated and in some not, so we were able to compare the impact. The cost was only 20 cents per person protected , which is extremely low. Other interventions cost about $5 per person protected, so it’s a very big difference. This operation spanned 100 days, but because it was so inexpensive, we could have used the budget they usually spend to run the operation year-round and scale it up to more villages in the area.

Another interesting operation was in Ethiopia, where we worked in a few villages and mapped the area. We learnt a lot from this operation because we saw how different fields are correlated with the existence of water bodies. For example, teff fields (teff is a kind of grain) did not have suitable water bodies for Anopheles mosquitoes, whereas in the grazing area where the cows were we saw hundreds.

How does the collaboration work? Do you have to be in the field when you work with these people in Ghana or Ethiopia, or is this a remote collaboration?

I’m a big fan of being in the field. It’s about user experience. It’s about science. It’s about the quality of training. It’s about understanding the specific problems. In each of our operations, even during COVID, we visited in person except for one operation in Kenya where, because of COVID, we weren’t able to make it. You learn a lot from such partners.

How do you measure how successful you are?

You have two measures. One is about mosquitoes. We catch mosquitoes. We don’t catch them to kill them, we catch them to count them – again inside the intervention area and then outside the intervention area to confirm the reduction rate of mosquitoes. And then, the most important measure is to count malaria cases to see if you reduce malaria cases.

And what are the results?

In Ghana, this is the first time we did an end-to-end trial. The results of mosquito reduction is amazing, it’s more than 60%.

What timeframe is that 60% reduction over?

It’s 100 days, less than four months. Now we start our most ambitious operation, which is about malaria elimination – really elimination to zero. For that, we collaborate with the government of São Tomé and Príncipe. São Tomé and Príncipe are two islands, forming an African island nation. Because they are islands, it’s a closed system, so you don’t get incoming mosquitos and they don’t go out.

We want, in two years, to not only target the water bodies but integrate other interventions based on artificial intelligence planning, to understand where to do what, and then to eliminate the disease. If that happens, it opens up the opportunity to approach larger countries and offer them malaria elimination. Malaria elimination will save many people and will boost the economy. People understand that if you eliminate malaria from countries, the impact on the GDP and on their economy will be more than 10% in a few years. It affects agriculture, tourism, education – everything.

Obviously, there is an imminent health care cost or health care impact of malaria. But you mentioned at the beginning that it drives poverty, it has economic implications…

The sad thing about malaria, and actually about many things in our world, is that it impacts the poor more than it impacts the rich. You have more malaria in poor villages than in the wealthy neighbourhoods in the cities. For example, in India, you have more malaria in some tribes in the mountains than all of the rest of the population. Malaria attacks poor communities and then prevents them from getting out of their poverty because it’s difficult to go forward. It’s a vicious cycle of poverty and disease that we want to help solve.

What’s happening in terms of more widespread adoption across Africa, are you hoping that one day this will be across the whole of Africa, or is that a challenging thing to scale to that degree?

For us, it is not challenging. It’s software, basically. It’s not very difficult to scale. So scaling up is up to the Ministry of Health. If countries in Africa adopt the system, it will happen. If one country uses it, they will help not only themselves, but also neighbouring countries, because, as you know – mosquitoes don’t believe in borders. I really hope to see it across the world, and not only in Africa – in South America and in India where there is malaria as well. We must do our best to eliminate this disease. I think it’s very strange that in the 21st century we still have such disasters happening. We experienced 18 months of the Coronavirus pandemic and we know how difficult it is, so why do we allow a disease such as malaria to persist for decades or centuries? We must stop it.

Most of the deaths are in children. Is that right?
Yes. Children under five is where most of the deaths are. And then there are also pregnant women who are greatly affected by the disease.

You see varying estimates. What would you put the estimates of the number of people per year affected by malaria?

I don’t have better estimates than the World Health Organisation, which estimated about 400,000 people last year. This figure grew by 15% or so because of COVID. With COVID, it became difficult to provide bed nets and health clinics were less available to treat malaria patients, which is what caused a surge in malaria cases.

You’ve worked a lot in this field of malaria prevention over the last couple of years. Are there any other organisations that you would want to mention that work on different approaches, who are maybe tackling the same problem?

If we improve the available vaccine it will be amazing, and some teams are trying to do so. Some tried to do this based on MRA technology, and some are trying to just take the existing vaccine, and with a few modifications, enhance its capability. Other groups try to do so by engineering mosquitoes – they want to put genes in the mosquito that means the mosquito itself will have a drug against malaria. Then the mosquito will not be infectious. This is a very innovative approach.

If it happens, it will be the first time in history where humanity has taken a species from nature and replaced it with other species with different genes. This is interesting, but we are still trying to understand how to do it in terms of technology and the safety of this method.

If people want to support you, what can they do?

Thanks. I actually prefer if people support the malaria NGOs. For example, Malaria No More and Only Nets are two good NGOs that really save many, many lives. If you buy, for example, 10 bed nets, it costs only $40 and you are potentially saving lives.

Where can people find out more about your work and about Zzapp Malaria?

Our website is

“We must do our best to eliminate this disease. I think it’s very strange that in the 21st century we still have such disasters happening.”


Back to blogs
Share this:
© Data Science Talent Ltd, 2024. All Rights Reserved.