GIS-based neural network framework for zoonotic cutaneous leishmaniasis risk mapping in Western Iran

Area of Science:

• Environmental science
• Epidemiology
• Artificial intelligence

Background:

• Zoonotic Cutaneous Leishmaniasis (ZCL) poses a significant public health challenge.
• Accurate risk mapping and prediction are crucial for effective control strategies.

Purpose of the Study:

• To develop a Geospatial Artificial Intelligence (GeoAI) framework for high-resolution ZCL risk mapping.
• To analyze environmental drivers and project future ZCL risk scenarios.
• To implement and evaluate advanced neural network architectures for spatiotemporal disease modeling.

Main Methods:

• Integration of Geographic Information Systems (GIS), remote sensing, and neural networks (MLP, 2D-CNN, 3D-CNN).
• Utilized historical ZCL maps and multi-temporal satellite-derived environmental data.
• Employed a 3D-CNN for explicit learning of spatiotemporal transmission dynamics.

Main Results:

• Temperature was identified as the dominant positive environmental driver of ZCL risk.
• The 3D-CNN model demonstrated superior performance in capturing complex spatial and temporal patterns compared to other architectures.
• Current high-risk areas are concentrated in warmer western and southern regions.

Conclusions:

• The GeoAI framework provides a robust tool for ZCL risk assessment and prediction.
• Projected 2030 risk indicates a spatial shift, with decreasing risk in western regions and increasing risk in southern areas.
• Findings support targeted surveillance and intervention efforts for ZCL control.