Related Experiment Video
Updated: Aug 7, 2026
09:11
Visualizing Dengue Virus through Alexa Fluor Labeling
Published on: July 9, 2011
A spatiotemporal U-Net++ deep learning framework for dengue risk mapping in Colombia
Daira Velandia1,2, Javiera Contador3, Juan Zamora3,4
1Institute of Statistics, Universidad de Valparaíso, Valparaiso, Chile. daira.velandia@uv.cl.
Scientific Reports
|August 5, 2026
Summary
This study developed a deep learning framework to map dengue risk in Colombia, integrating climate, environmental, and socioeconomic data. The approach effectively identifies high-risk areas, aiding public health surveillance and interventions.
Area of Science:
- Epidemiology
- Geospatial analysis
- Artificial intelligence
Background:
- Dengue transmission is influenced by climate, human behavior, and vector distribution.
- Spatiotemporal understanding of dengue dynamics is crucial for public health.
- Aedes aegypti is the primary vector in tropical and subtropical regions.
Purpose of the Study:
- To propose a spatiotemporal deep learning framework for high-resolution dengue risk mapping in Colombia.
- To integrate diverse data sources including climatic, environmental, demographic, and socioeconomic information.
- To evaluate different spatial approaches for dengue risk assessment.
Main Methods:
- Development of a U-Net++ based deep learning framework.
- Integration of satellite and census-derived data.
- Evaluation of high-dimensional geography (HDG) and low-dimensional geography (LDG) spatial approaches.
Main Results:
- The model integrating climatic, social, and environmental data performed best.
- The HDG configuration achieved a test mIoU of 0.6646.
- The LDG configuration achieved an average test mIoU of approximately 0.73.
Conclusions:
- Deep learning frameworks can effectively characterize dengue risk patterns.
- Integrating heterogeneous data sources enhances dengue risk prediction.
- The framework supports high-resolution surveillance and public health decision-making for dengue control.