AI to reduce malaria in a warming worldAI to reduce malaria in a warming world

It is often noted that the most dangerous animal in the world is not a fearsome and powerful creature like the shark, tiger or cobra but is in fact the ubiquitous and tiny mosquito.

Hundreds of thousands of people die yearly from mosquito-borne diseases, with malaria being the deadliest. The World Health Organization estimates that every year approximately 250 million people contract malaria, which is caused by a parasite that is spread through mosquito bites. More than 600,000 people died from the disease in 2021.

One way to reduce the impact of malaria is to better understand mosquitoes and the environments that are conducive to their spread. These bugs thrive in hot, humid environments, and mosquito-borne diseases like malaria are widespread in regions near the equator, such as sub-Saharan Africa and parts of Asia and the Americas.

Related: AI leads the way in advancing early disease detection

As the Earth becomes hotter and wetter due to global climate change, malaria outbreaks are poised to become more severe, last longer and occur in regions they haven’t in the past, putting more people at risk for the disease.

A team of scientists at MBZUAI, led by Professor of Computer Vision Abdulmotaleb El Saddik, is developing artificial intelligence-powered applications to help physicians and public health officials in Indonesia reduce the impact of malaria on the country’s population of 270 million.

The applications El Saddik and his team are developing perform sensory data fusion, a process that combines data from a variety of sensors to generate a virtual representation of the environment in the form of a “digital twin.” This approach provides precise weather forecasting and generates a near real-time representation of the environment, providing officials with detailed information about where malaria outbreaks may occur.

Abdulmotaleb El Saddik

Artificial intelligence techniques and clustering analysis tools are applied to the data collected from sensors with the goal of identifying recurring features that contribute to malaria outbreaks. In the future, these applications have the potential to be used in other countries at risk for malaria as well.

The initiative is supported by Reaching the Last Mile, a portfolio of global health initiatives and programs funded by the philanthropy of His Highness Sheikh Mohamed bin Zayed Al Nahyan, President of the UAE. Associate Professor of Computer Vision Mohammad Yaqub and Director of Research Engagement Hosni Ghedira both of MBZUAI, are also involved in the project.

Power of prediction

When it comes to reducing the impact of malaria on people, preventative measures and early treatment are the most effective approaches. With foresight, officials can notify at-risk populations, implement additional mosquito-control measures and send local clinics the resources necessary to deal with an outbreak in advance. If health officials anticipate where outbreaks will occur before they happen, they will likely be able to save many lives.

To provide these tools, El Saddik and his team are developing a machine-learning algorithm that analyzes atmospheric, epidemiological, geographic and other kinds of data. The system is designed to draw on information captured by remote sensors that are placed throughout the country — including in remote villages — and analyses the data through sensory data fusion, with the goal of creating a comprehensive and accurate representation of where malaria hotspots may arise.

“Some of the most important parameters related to the spread of malaria are temperature, humidity and rain,” El Saddik said. “But it’s complicated, because it’s not a linear relationship,” as more rain or hotter temperatures don’t always bring about more cases of malaria. For example, a heavy rainfall that causes rainwater to be drained away quickly may result in a lower risk for malaria than a minor rain event that results in more standing water for mosquitoes to breed in.

“Predicting malaria is both a challenge of data collection and data analysis, and that’s why it’s an interesting research problem.”

Improving diagnosis and speeding care

Today, there are several tests that can be used to diagnose a patient with malaria. The most accurate approach is to draw a small sample of a patient’s blood, stain it with special dyes and examine the sample under a microscope. Accurate diagnosis to determine the species of the parasite is essential for effective treatment.

Many people, however, contract malaria in remote areas where resources are limited. In a small village there might not be a medical provider who has been trained to properly interpret the results of a malaria test.

Related: AI-powered growth in healthcare: early trends and learnings

El Saddik and his team are developing an application to analyse microscope images and determine the type of infection and stage of the disease. It draws on principles of computer vision to identify characteristics in images that are indicative of malaria.

It’s not an easy task. The type of malaria that is found in Southeast Asia is different from that found in Africa, with resulting differences in how symptoms of the disease appear in microscope images. “Like other challenges with computer vision, we need enough good data to train the model, and we have been identifying annotated data sets and images so that they can train our models to do the proper prediction,” El Saddik said.

The programme also needs to be efficient in terms of computing power and the amount of data it receives and sends over the network. “We want the application to be able to be used over any kind of mobile communication, particularly in a remote area like the jungle of Central Papua,” a region in Indonesia, El Saddik said. “The application also needs to be lightweight so that it can function on a mobile app.”

While the computing and analysis of the image would be done locally on the mobile device, when a case of malaria is detected, it would share that information with health authorities so that they were aware of cases and could take appropriate steps. Data collected through the app could be used to inform the prediction model El Saddik is developing.

Idea to implementation

El Saddik and his team began working on the project earlier this year and have already developed a version of the mobile app that can detect malaria from images. The project is expanding beyond Indonesia as well. They are collaborating with physicians in India to learn what kinds of features would be most helpful for doctors.

In the next several months, El Saddik hopes to provide the first full version of the applications to his partners in Indonesia and test the system. “The most important aspect of our work is to use the potential of artificial intelligence to save as many human lives as possible, to use AI to serve humanity and help society,” El Saddik said.

Get the latest insights from the healthcare industry straight to your inbox. Subscribe for a bi-weekly roundup of the latest healthcare tech, trends, innovation and patient experience, accompanied by exclusive interviews from the industry’s leaders.

Back to Technology