Mathematical methods for visualization and anomaly detection in telemetry datasets.

Area of Science:

• Computational biology
• Infectious disease modeling
• Animal health monitoring

Background:

• Biological data acquisition and analysis advancements enable new health modeling approaches.
• Telemetry data provides continuous physiological monitoring for disease detection.

Purpose of the Study:

• To explore temperature pattern evolution in mice during infection.
• To identify anomalies in temperature time series for early disease detection.

Main Methods:

• Utilized Laplacian Eigenmaps and Multidimensional Scaling for data visualization.
• Applied Radial Basis Functions and Multivariate State Estimation Technique for anomaly detection in nonlinear time series.
• Analyzed telemetry data from 306 Collaborative Cross mice infected with Salmonella typhimurium.

Main Results:

• Demonstrated visualization techniques to understand overall dataset patterns.
• Developed models of healthy states using anomaly detection tools.
• Showcased the ability to detect disruptions in circadian patterns and infection severity within three days.

Conclusions:

• Temperature time series analysis is effective for early infectious disease detection in mice.
• Data-driven modeling can reveal subtle physiological changes indicative of illness.
• Methods presented offer a promising approach for real-time animal health monitoring.