AI-powered surveillance may be Africa’s new weapon against malaria resurgence

Malaria remains one of the most devastating diseases globally, infecting over 250 million people and killing more than 600,000 every year. To address this menace, a multinational research team has unveiled an artificial intelligence-powered tool designed to detect invasive mosquito species capable of spreading malaria in resource-limited regions. The work marks a breakthrough in the use of artificial intelligence and citizen science for disease vector surveillance.

The study, titled "Artificial Intelligence and Citizen Science as a Tool for Global Mosquito Surveillance: Madagascar Case Study," and published in the journal Insects demonstrates how AI image recognition models, combined with mobile-based citizen data collection, can identify invasive mosquitoes like Anopheles stephensi, a species that has rapidly spread across Africa and poses a major public health threat due to its ability to thrive in cities.

A new approach to malaria surveillance in Africa

In Africa, the arrival of Anopheles stephensi, an invasive Asian mosquito species first detected on the continent in 2012, has significantly complicated malaria control efforts. Unlike native African malaria vectors that breed in natural water sources such as rice paddies and puddles, An. stephensi adapts to urban life, breeding in artificial containers like tires, barrels, and buckets.

In Madagascar, where malaria incidence has surged dramatically since 2015, the risk of invasion by An. stephensi is considered among the highest in East Africa. The new research integrates AI and citizen science to strengthen early detection and rapid response efforts. The team trained deep learning models using thousands of smartphone photos of laboratory-authenticated mosquito larvae from eight species, including An. stephensi, An. gambiae, and Aedes albopictus.

Through NASA's GLOBE Observer app, citizens across Madagascar uploaded photos of mosquito larvae found in artificial containers. One image submitted in March 2020 by local participants in Antananarivo became the focal point of the study. Although physical specimens were unavailable for genetic testing, the AI model accurately identified the larva in the photo as An. stephensi, confirming that mobile-based citizen reporting could serve as a reliable tool for vector surveillance.

The team deployed multiple AI architectures, EfficientNet, Inception-ResNet-V2, and Xception, achieving up to 99.34 percent confidence in the classification of An. stephensi larvae, with a false-positive rate of less than one percent. The explainable AI feature used in the analysis pinpointed anatomical traits such as abdominal segment pigmentation, validating the accuracy of the prediction and even determining the sex of the mosquito with 100 percent confidence.

AI models turn citizen photos into surveillance tools

The research explores how AI can transform smartphone images into actionable surveillance data. By training convolutional neural networks on verified mosquito larvae images, the team created models that can distinguish An. stephensi from morphologically similar species. The EfficientNet-B0 model, which delivered the highest classification accuracy, used photo augmentation techniques such as rotation, contrast adjustment, and cropping to enhance recognition precision.

Explainable AI was particularly valuable in validating the predictions. The Grad-CAM and Class Activation Mapping methods revealed which anatomical features the AI considered most relevant, offering transparency and biological insights. These visualizations identified dark pigmented areas on the larva's abdomen, corresponding to testes in male mosquitoes, a discovery that confirmed both the sex and species classification.

Following the AI analysis, the researchers conducted extensive larval surveys from 2022 to 2023 across six districts of Antananarivo. Nearly 2,000 artificial containers were inspected, yielding over 27,000 larvae from three genera, Aedes, Anopheles, and Culex. Notably, no An. stephensi specimens were detected during this follow-up study, suggesting that the 2020 finding may have been an isolated incident or that community-based habitat elimination prevented local establishment.

The results validate AI's potential to provide early warning signals when paired with citizen-collected imagery. Even when physical specimens are unavailable for molecular confirmation, archived images can be analyzed retrospectively using machine learning models. This allows health agencies to monitor vector movements, identify invasion patterns, and mobilize interventions faster than through traditional entomological surveys.

Global implications for vector control and public health

The integration of artificial intelligence and citizen science has global implications for combating vector-borne diseases. With malaria transmission expanding due to climate change, urbanization, and insecticide resistance, scalable surveillance tools are urgently needed. The study's approach offers a cost-effective model for global deployment, leveraging smartphone technology, open data, and community participation.

According to the researchers, citizen involvement is vital for identifying breeding habitats and eliminating standing water sources before mosquitoes mature. The GLOBE Observer app, developed under NASA's Earth Science Education Collaborative, empowers volunteers in over 120 countries to document mosquito habitats and larvae, creating a global database accessible to both scientists and public health officials.

The team recommends that citizen scientists capture multiple images of each larva, from different angles and body regions, to improve AI detection reliability. They also encourage the rearing of larvae found in artificial containers for further species confirmation, particularly in urban or livestock-rich environments where An. stephensi thrives.

To scale up this approach, the researchers have made their AI tools and data visualization platforms freely accessible. The Global Mosquito Observations Dashboard (mosquitodashboard.org) integrates citizen science data from GLOBE Observer, iNaturalist, and Mosquito Alert. The complementary platform mosquitoID.org offers AI-powered identification of both larval and adult mosquitoes. Together, these systems form a real-time global network for tracking invasive and disease-carrying mosquito species.

The study also highlights potential applications of AI beyond surveillance. Automated larval sexing, for instance, could support eco-friendly mosquito control programs such as Wolbachia infection, genetic modification, and sterile insect techniques, all of which require precise sex separation to reduce breeding populations. The researchers envision expanding AI-enabled smart traps to target adult mosquitoes, integrating these systems with community reporting networks.