A weighted-principal component regression method for the identification of physiologic systems

Area of Science:

• Physiology
• Systems Biology
• Biomedical Engineering

Background:

• Accurate system identification is crucial for understanding physiologic systems.
• Conventional methods struggle with colored signals common in physiology.
• Linear time-invariant (LTI) models are often used to represent resting physiologic states.

Purpose of the Study:

• To introduce a novel system identification method, weighted-principal component regression (WPCR).
• To enhance the accuracy of LTI model identification for physiologic systems, particularly with colored input signals.
• To incorporate prior knowledge into the system identification process.

Main Methods:

• Weighted-principal component regression (WPCR) for time-domain system identification.
• Incorporation of dominant frequency components from input signals.
• Application to single-input and multi-input single-output systems in open-loop and closed-loop configurations.
• Inclusion of a weighting scheme for incorporating pre-existing system knowledge.

Main Results:

• WPCR enables the construction of data-specific candidate models.
• The method reduces parameter estimation error for colored signals.
• WPCR demonstrated more accurate identification of the system impulse response function compared to conventional methods using both simulated and experimental data.

Conclusions:

• WPCR offers a significant advancement in system identification for physiologic modeling.
• The method's ability to handle colored signals makes it suitable for real-world physiologic data.
• WPCR provides a robust approach for LTI system identification, outperforming existing techniques.