AI + Citizen Science for Mosquito Surveillance
The Problem: Mosquito-borne diseases are amongst the most critical public health concerns. More than a million people die each year mainly from malaria, dengue, chikungunya and more that are all spread by mosquitoes. The state of Florida is a domestic epicenter for mosquito-borne diseases, with a devastating Zika outbreak in 2016 and locally transmitted cases of dengue fever in 2019 and 2020. The majority of known mosquito-borne diseases are transmitted by three common mosquito genera, namely Aedes, Anopheles, and Culex. Because there are no vaccines or cures available for many of these diseases, real-time surveillance is critical in deploying countermeasures, such as more targeted insecticide treatment and public information campaigns, to eliminate breeding habitats and mitigate disease outbreaks.
At USF, we are engaged in multi-disciplinary research (with AI as a central theme) to develop a platform for large-scale automated identification of mosquito genera and species via citizen-generated smartphone images and AI algorithms. The platform will enable citizens to upload smartphone images taken in nature to contribute to real-time data on mosquito populations worldwide, which will then be used for superior forecasting of risk prediction maps.
Our Results: Our model based on Convolutional Neural Networks and more than 10,000 images of nine mosquito vectors, yields very good accuracy in detecting the genus and species. Very interestingly, the highest classification accuracies were obtained for two of the deadliest vectors on the planet: Aedes aegypti (vector for dengue, chikungunya and Zika fever) and Anopheles stephensi (vector for malaria). Our model also provides evidence that evolution – as measured by relative divergence times – driving greater anatomical disparity, in turn yields higher classification accuracies when using AI. In other words, our research shows that signals from millions of years of natural selection is indeed being revealed by real-time AI.
Funding: This project is currently funded via an NSF Smart and Connected Health Grant titled – “SCH: INT: Surveillance and Control of Mosquito-Borne Diseases through Automated Species Identification and Spatiotemporal Modeling” (#2014547).
Mona Minakshi, Pratool Bharti, Tanvir Bhuiyan, Sherzod Kariev and Sriram Chellappan (2020). "A Framework based on Deep Neural Networks to Extract Anatomy of Mosquitoes from Images", Scientific Reports.
Mona Minakshi, Tanvir Bhuiyan, Sherzod Kariev, Martha Kaddumukasa, Denis Loum, Nathanael Stanley, Sriram Chellappan, Peace Habomugisha, David W. Oguttu and Benjamin Jacob (2020). "High-accuracy detection of malaria mosquito habitats using drone-based multispectral imagery and Artificial Intelligence (AI) algorithms in an agro-village peri-urban pastureland intervention site (Akonyibedo) in Unyama Sub-County, Gulu District, Northern Uganda", Journal of Public Health and Epidemiology.
Mona Minakshi, Pratool Bharti, Willie McClinton III, Jamshidbek Mirzakhalov, Ryan Carney and Sriram Chellappan (2020). "Automating the Surveillance of Mosquito Vectors from Trapped Specimens Using Computer Vision Techniques", ACM SIGCAS Conference on Computing and Sustainable Societies (COMPASS).
Farhana Shahid, Shahinul Ony, Takrim Albi, Sriram Chellappan, Aditya Vashistha and A.B.M. Alim Al Islam (2020). "Learning from Tweets: Opportunities and Challenges to Inform Policy Making During Dengue Epidemic", 23rd ACM Conference on Computer-Supported Cooperative Work and Social Computing (CSCW).