Using generalized additive models to decompose time series and waveforms, and dissect heart-lung interaction...

Area of Science:

• Physiology
• Biomedical Engineering
• Data Science

Background:

• Physiological signals often contain overlapping cardiac and respiratory cycles.
• Understanding these individual components is crucial for accurate patient monitoring and diagnosis, particularly in fluid responsiveness prediction.
• Current methods may struggle to effectively disentangle these cyclical influences.

Purpose of the Study:

• To demonstrate the application of Generalized Additive Models (GAMs) for decomposing physiological time series.
• To illustrate how GAMs can isolate and quantify the distinct effects of cardiac and respiratory cycles within physiological waveforms.
• To showcase the utility of GAMs in analyzing data from mechanically ventilated subjects.

Main Methods:

• Utilized Generalized Additive Models (GAMs) to estimate nonlinear, smooth functions representing physiological cycles.
• Applied GAMs to decompose physiological signals into separate cardiac and respiratory components.
• Developed models for respiratory variation in pulse pressure and central venous pressure waveforms.

Main Results:

• Successfully demonstrated the decomposition of physiological signals into distinct cardiac and respiratory effects using GAMs.
• Modeled the respiratory variation in pulse pressure, isolating the respiratory influence.
• Decomposed central venous pressure waveforms into cardiac, respiratory, and interactive effects.

Conclusions:

• Generalized Additive Models (GAMs) offer an intuitive and flexible approach for analyzing repeating patterns in medical monitoring data.
• GAMs effectively disentangle cardiac and respiratory influences in physiological time series.
• This methodology enhances the analysis of complex physiological waveforms, particularly in mechanically ventilated patients.