Machine Learning in Rehabilitation Assessment for Thermal and Heart Rate Data Processing

Area of Science:

• Biomedical Engineering
• Physiological Computing
• Rehabilitation Technology

Background:

• Multimodal signal analysis using noninvasive sensors, communication systems, and machine learning is increasingly applied in various fields.
• Physiological data analysis is crucial for monitoring health and rehabilitation progress.

Purpose of the Study:

• To analyze physiological data (heart rate, breathing temperature) acquired during exercise rehabilitation using noninvasive sensors.
• To determine fitness levels and detect potential medical disorders through pattern recognition and machine learning.

Main Methods:

• Utilized 56 experimental datasets (40 min each) of heart rate and breathing temperature from exercise bike sessions.
• Combined machine learning for thermal camera data analysis and adaptive image processing for breathing frequency evaluation.
• Applied a neural network model with specific transfer functions for individual temperature value determination and statistical methods for correlation analysis.

Main Results:

• Evaluated a mean delay of 21 s for heart rate drop post-activity change, aligning with real cycling conditions.
• Determined average changes in breathing temperature (167 s) and breathing frequency (49 s).
• Successfully correlated exercise activity with selected physiological functions.

Conclusions:

• The developed methodology effectively analyzes multimodal physiological data for rehabilitation monitoring.
• Machine learning and adaptive image processing provide accurate insights into fitness levels and physiological responses.
• The findings support the use of noninvasive sensors and advanced analytics in personalized rehabilitation programs.