Self-organizing maps as an approach to exploring spatiotemporal diffusion patterns.

Area of Science:

• Epidemiology
• Data Science
• Geographic Information Systems (GIS)

Background:

• Self-organizing maps (SOMs) are established for pattern identification in large datasets.
• Application of SOMs in the spatiotemporal domain for disease analysis has been limited.
• This study addresses the gap by applying SOMs and Sammon's projection to analyze spatiotemporal disease diffusion patterns.

Purpose of the Study:

• To demonstrate the application of SOMs and Sammon's projection for analyzing spatiotemporal disease diffusion.
• To identify synchrony between spatial locations during disease outbreaks.
• To group epidemic waves based on diffusion pattern similarity and visualize diffusion trajectories.

Main Methods:

• Self-organizing maps (SOMs) were utilized to detect spatial synchrony and group epidemic waves.
• Sammon's projection was employed to generate diffusion trajectories from SOM outputs.
• The methodology was validated using a dataset of Measles outbreaks in Iceland (1946-1970), detailing monthly cases across 50 medical districts.

Main Results:

• Identified both stable and incidental synchrony between medical districts.
• Distinguished two groups of epidemic waves: uniformly structured fast-developing and multiform slow-developing.
• Observed distinct diffusion trajectories: a typical pattern from Reykjavik for fast waves and heterogeneous patterns for slow waves.

Conclusions:

• Self-organizing maps (SOMs), coupled with Sammon's Projection and GIS, are applicable for spatiotemporal diffusion analysis.
• The study successfully visualizes diffusion patterns, highlighting similarities and differences between epidemic waves.
• Integrated analysis of synchrony and diffusion trajectories provides a comprehensive understanding of disease spread dynamics.