Intelligent Systems for Vector-Borne Disease
Coordinator
Dr. Dorothe Poggel (HWK)
Speaker
- Prof. Dr. Peter Haddawy, Faculty of ICT, Mahidol University, Thailand (HWK alumnus)
Members
- Prof. Dr. Gabriel Zachmann, AG Computer Graphics and Virtual Reality, Universität Bremen
- Prof. Dr. Anna Förster, ComNets Lab, Universität Bremen
- Prof. Dr. Soerge Kelm, Glycobiochemistry, Universität Bremen
- Dr. Thomas Barkowsky, Bremen Spatial Cognition Center, Universität Bremen
- Dr. Tim Ziemer, Bremen Spatial Cognition Center, Universität Bremen
- Assoc. Prof. Dr. Saranath Lawpoolsri, Faculty of Tropical Medicine, Mahidol University
- Dr. Patchara Siriwichai, Faculty of Tropical Medicine, Mahidol University
- Assoc. Prof. Dr. Panisadee Avirutnan, Siriraj Hospital, Mahidol University
- Assoc. Prof. Dr. Anuwat Wiratsudakul, Faculty of Veterinary Science, Mahidol University
- Dr. Myat Su Yin, Mahidol-Bremen Medical Informatics Research Unit, Mahidol University
- Dr. Akara Supratak, Faculty of ICT, Mahidol University
- Prof. Dr. Nuno Nunes, Technical University of Lisbon
- Dr. Dumrong Mairiang, Biotec, National Science and Technology Development Agency, Thailand
- Dr. Dominique Bicout, VetAgro Sup
- Dipl.-Phys. Michael Weber, Biogents AG
Duration
15 June 2022 - 14 June 2025
Statement of Problem
Mosquito vector-borne diseases such as malaria, dengue, and Zika constitute some of the most severe public health burdens in tropical and sub-tropical countries. In addition, climate change is increasing the geographic range of many mosquito vectors, introducing diseases into regions not well equipped to deal with them.
For effective treatment and control of these diseases, timely access to high-quality data presented in a form that can inform decision making is crucial. This is true in both the clinical and public health domains. In fact, the two domains must be linked since information about patients diagnosed with illness in hospitals must be quickly and effectively communicated to public health personnel and information from public health monitoring can inform clinical preparedness measures. Focusing on dengue and malaria, this project addresses the entire information collection and management pipeline, including data collection, data analysis, and communication. It will seek to do this in an integrated fashion so that data flows effectively between clinics and public health workers. The first component focuses on mosquito vector surveillance, i.e., counting the number and identifying the species of mosquitoes in the field. This is achieved through the development of bio-acoustic sensor networks to provide adult mosquito vector counts at far lower cost and in a timelier fashion than through current means. The second component studies the relationship between human mobility and malaria transmission, with the objective to identify routes of potential transmission and simulations to evaluate alternative control strategies that take into account the effects of mobility. In the same vein, work will be conducted to model the exposure of populations to mosquito vectors of malaria and dengue. The last component is an integrated system to support dengue clinical care and public health control measures. We will explore approaches to provide advanced patient data management and decision support to hospitals and to communicate data on dengue patients to public health workers. From a technical perspective, we will use technology and approaches from Artificial Intelligence, Internet of Things (IoT), and Mobile Computing. The study group brings together a highly interdisciplinary group of researchers from the fields of Computer Science, Acoustics, Epidemiology, Medicine, Microbiology, and Entomology.
Goals
The work of the study group is organized into three work packages: WP1) bio-acoustic sensors for mosquito vector counts, WP2) human mobility and disease transmission, and WP3) decision support for dengue clinical care and public health response. These work packages cover a spectrum including vector monitoring, disease transmission models, clinical care, and control of disease transmission. By addressing the full length of the information and action pipeline from monitoring to clinical care to control, we aim to identify and support crucial information dependencies for effective decision-making in disease prevention and response. For example, vector counts from the bio-acoustic sensors in WP1 can provide information to the transmission models in WP2 and the public health control in WP3. Furthermore, information from the simulations in WP2 can inform public health responses in WP3.
WP1: Bio-acoustic sensor networks for mosquito vector counts
Mosquito vector-borne diseases such as malaria, dengue, chikungunya, yellow fever, and Zika constitute some of the most serious public health burdens in tropical and sub-tropical countries. Common public health interventions for these diseases include the provision of bed nets, distribution of vaccines and curative drugs (available for some diseases), and mosquito vector control. In order to effectively target such interventions, as well as to monitor the effectiveness of vector control efforts, accurate information about mosquito vector population density is essential. Since different diseases are transmitted by specific species of mosquitoes, with some species serving as vectors of multiple diseases and others not serving as disease vectors, vector surveillance requires obtaining population estimates categorized by species. This greatly complicates the surveillance task.
Current surveillance techniques for gathering adult mosquito vector counts are highly time-consuming, requiring specialists to place traps and manually count and classify the species caught in them. A promising alternative is acoustic monitoring, where the wingbeat sounds produced by mosquitos are used to identify different species in the field automatically. This project will involve the development and testing of networks of bio-acoustic adult mosquito vector sensors. Such sensor networks have the potential to greatly decrease the cost of vector monitoring, making data collection possible over large areas and for extended periods of time. The work will involve the development of key algorithms, the development of the sensor hardware platform, and field testing to calibrate and validate the sensors. Work will be carried out with our industry partner Biogents, who produce advanced mosquito monitoring systems. Initial work has explored a number of issues in construction of such sensor networks [6] and has produced excellent results in classification of mosquitos into sex and species using deep learning techniques in a controlled laboratory setting [7]. The latter publication received the conference’s best paper award for that year.
WP2: Human mobility and malaria transmission
Understanding mechanisms of malaria transmission in low-transmission settings is essential in order to develop strategies for malaria elimination. In such settings, it is believed that human migration plays a major role in malaria transmission and even in the emergence of drug resistance. Malaria infections may show up in low incidence areas because people become infected during time spent in high-risk areas. In the case of northern Thailand, mobility may be due to normal daily patterns, labor migration, and seeking of refuge. It is known that a significant factor in malaria transmission in Thailand along the border with Myanmar is due to cross-border migration, largely due to seeking of labor. While attempts have been made to understand movement patterns through systematic interviews, answers depend on subjects’ memory and thus the data may not be accurate and precise. A more precise approach is to use a GPS device to collect mobility data. This provides not only locations visited but also precise durations spent and time of day, which is important to detect the period of mosquito-human exposure.
We have developed a smartphone app and used it to track the movement of approximately 400 participants over the course of one year in Tak province in northern Thailand along the border with Myanmar. The app records each participant’s location on an hourly basis. This data can be used to gain a number of important insights. If we know the latent risk of infection associated with different geographic areas at different times of the year, the mobility data can be used to compute each person’s individual risk, depending on how much time they spend in each area. In addition, the mobility data can be used to determine geographic routes of transmission and importantly routes of importation. Finally, we can use the above information to develop simulations to determine the effectiveness of various control efforts that take into account mobility. We will address all these issues in this project.
The framework and tools we will develop in this work package are applicable not only to vector-borne diseases but also to infectious diseases generally and to health conditions resulting from exposure to environmental hazards. Since non-vector-borne infectious diseases are spread through human contact in one form or another, understanding mobility patterns is crucial to understanding transmission. We have, in fact, made use of the mobility data we gathered to understand how government travel restrictions during the COVID-19 pandemic affected mobility patterns along Thai-Myanmar border.
WP3: Decision support for dengue clinical care and public health response
Dengue is a widespread disease in countries in SE Asia. Most people infected recover well, but a small percentage develop dengue hemorrhagic fever (DHF) which can be life-threatening. It is thus important to be able to identify those patients at risk of DHF in order to provide close monitoring and care. This work package will develop a dengue patient clinical information system, which will include a prognostic prediction model to provide running predictions of whether a patient will develop DHF and the time this may occur, called day zero. We will apply machine learning techniques to a unique dataset of clinical data from pediatric patients gathered over a period of 18 years from two hospitals in Thailand. The prognostic prediction model will focus initially on pediatric patients since they are one of the high-risk groups. A 2-year grant on “Advanced statistical analyses and predictive model construction to identify correlation between clinical data and disease severity of dengue patients” has recently been secured from Thailand’s National Science and Technology Development Agency and Siriraj Hospital. The grant will support model and software development, as well as evaluation.
Once a patient is diagnosed with dengue, public healthcare personnel must visit their home to investigate and eliminate potential sources of further infection. A rapid response is crucial to limiting further disease spread. While hospitals in Thailand are currently required to report every dengue case to the Ministry of Public Health, the process of getting this information to the responsible public health workers so that they may take action can be slow. In addition, such information is currently primarily sent through the Line app, which lacks data management capabilities and is prone to data privacy leaks. To speed this process, provide better data management, and provide more secure data communication, we will develop an information system that will directly route confirmed dengue case information to responsible healthcare workers. The system will provide this data in a form that supports the health workers to investigate patient households for the presence of dengue vector breeding sites and other risk factors. As the work in WP1 and WP3 matures, the mosquito sensors from WP1 will be integrated into this system so that the healthcare workers have access to information about dengue vector counts, which is very useful both for potential vector contact risk and to evaluate vector control actions.
The work plan includes a study group meeting each year. In order to facilitate close communication and coordination among the study group members, frequent online meetings will also be held. Several high-rank publications will help to disseminate the results, as well as one international workshop held in conjunction with a major related conference. Prototype sensor systems, prediction tools, and efficient simulation are expected to be the main results of the Study Group. Further, a common grant application for establishing a Network of Excellence will serve to continue the work of the Study Group and to further disseminate and to apply the knowledge created.