The three-element Windkessel model is widely used and accepted for analyzing blood flow and pressure in arterial system and cerebral circulation. In most studies, changes in mean arterial blood pressure data is used as input to estimate the model parameters. However, estimation of linear model parameters, using input-output data, requires that the input be persistently exciting. This study examined the efficacy of using mean arterial blood pressure (MABP) sequence as an input stimulus for estimating the parameters of the three-element Windkessel model. Additionally, the study explored the use of a shorter MABP data segment of 1.5 mm as compared to the commonly used 6 mm data. MABP data was obtained from 11 healthy subjects. One thousand three-element Windkessel models, with parameter values randomly selected to be within physiological range, were subjected to seven different input sequences. For each input sequence and model, the values of the model (target-parameters) were estimated. The seven input sequence were: 1) six minutes of MABP measured from subjects; 2-5) four 1.5 mm of measured MABP obtained by dividing the measured six minutes of MABP into non- overlapping contiguous segments; 6) a six-minutes of pseudo random binary sequence (PRBS) with amplitudes comparable to the MABP sequence; and 7) a 1.5 mm of PRBS sequence with amplitudes comparable to the MABP sequence. The MABP data used was randomly selected from the 11 subjects for each estimation run. The model parameter estimation method had two phases of optimization. In the first phase, the parameters were estimated and optimized using the frequency transform of the input and output. In the second phase, the values of the estimated parameters were used as initial estimates and time-domain optimization was carried out to further refine the estimates. Results from the study, comparing the estimated-parameters with the target-parameters, show that for the MABP data, there was no significant difference between using the six minutes or 1.5 mm of data for estimating the target-parameters. Also, parameters estimated from the MABP data were either equivalent or superior to the PRBS results, suggesting that changes in MABP can be used as an effective sequence for linear model estimation.
ASJC Scopus subject areas
- Computer Vision and Pattern Recognition
- Signal Processing
- Biomedical Engineering
- Health Informatics