Droplet-Transmitted Infection Risk Ranking Based on Close Proximity Interaction

Area of Science:

• Computer Vision
• Public Health Informatics
• Epidemiology

Background:

• Identifying individuals at risk for droplet-transmitted diseases traditionally involves time-consuming manual screening of surveillance data or extensive interviews.
• Current methods can delay crucial public health interventions for communicable diseases such as influenza.

Purpose of the Study:

• To propose an automated, efficient, and accurate method for identifying potentially infected individuals from surveillance videos.
• To reduce the labor and time costs associated with traditional infection risk assessment.

Main Methods:

• A multi-tasking framework was developed to analyze surveillance videos.
• Key sub-tasks include person re-identification (REID) across cameras, depth estimation for spatial context, and pose estimation to calculate proximity between individuals.
• The framework models Close Proximity Interaction to infer infection risk.

Main Results:

• The proposed method significantly reduces time and labor costs compared to manual screening.
• Demonstrated high accuracy and efficiency in identifying high-risk individuals.
• The system effectively identifies diagnosed patients and potentially infected subjects.

Conclusions:

• The automated system accelerates the identification of potentially infected populations.
• This approach contributes to timely public health responses and disease control.
• The method holds promise for improving overall public well-being by enabling faster interventions.