Practical identifiability analysis of a minimal cardiovascular system model

Area of Science:

• Cardiovascular physiology
• Mathematical modeling
• Biomedical engineering

Background:

• Mathematical models of the cardiovascular system offer insights into physiological states like blood volume, vessel elastance, and resistance.
• Accurate estimation of model parameters from bedside patient data is crucial for real-time cardiovascular monitoring.
• This study focuses on a seven-parameter cardiovascular model to assess parameter identifiability.

Purpose of the Study:

• To investigate the unique determination of seven cardiovascular model parameters using clinical hemodynamic indices.
• To evaluate the practical identifiability of these parameters from arterial and venous pressures, and stroke volume measurements.
• To identify limitations in using the model for patient monitoring due to parameter identifiability.

Main Methods:

• An error vector was defined using residuals between simulated and reference hemodynamic indices.
• Sensitivity and collinearity analyses were performed on the error vector with respect to each model parameter.
• Profile-likelihood curves were constructed to assess the practical identifiability of individual model parameters.

Main Results:

• Four out of seven cardiovascular model parameters were found to be practically identifiable.
• Three parameters were identified as practically non-identifiable, hindering precise estimation.
• Inverse correlation between two non-identifiable parameters prevented their unique determination.

Conclusions:

• Practical identifiability analysis revealed that three parameters of the seven-parameter cardiovascular model were not uniquely determinable from limited clinical data.
• The non-identifiability of these parameters restricts the model's utility as a cardiovascular monitoring tool.
• Adjustments to the cardiac contraction function and venous pressure reference range improved model identifiability, suggesting pathways for enhanced clinical application.