An Edge Computing and Ambient Data Capture System for Clinical and Home Environments.

Area of Science:

• Health Informatics
• Computer Science
• Engineering

Background:

• Non-contact patient monitoring facilitates remote care and long-term studies.
• Key challenges include ensuring systems are non-intrusive, privacy-preserving, and affordable.

Purpose of the Study:

• To develop and evaluate an open-source, low-cost edge computing system for ambient data capture in patient monitoring.
• To demonstrate the system's utility across diverse applications, enhancing healthcare insights.

Main Methods:

• Utilized low-cost, readily available hardware for an edge computing and ambient data capture system.
• Developed and tested five distinct applications: occupancy estimation, medical alarm classification, geolocation, ambient light logging, and environmental logging (temperature/humidity).

Main Results:

• Achieved 94% accuracy for occupancy estimation from video data.
• Obtained high F1 scores (0.98/0.93) for medical alarm classification, even with background speech.
• Demonstrated 98.7% room-level accuracy for human geolocation.
• Validated temperature and humidity sensors with low root mean square errors (0.3°C and 1% RH, respectively).

Conclusions:

• The developed low-cost edge computing system effectively captures and analyzes diverse activities in clinical and home settings.
• The system offers privacy-preserving insights into healthcare practices and patient behaviors, supporting remote patient care.