Abstract
Introduction
Cardiovascular (CV) parameters can be used as markers of cardiovascular risk, and as inputs to computational models of arterial blood flow. However, they are difficult to measure directly. Several CV parameter estimation(CPE) methods have been proposed to estimate parameters from non-invasive blood pressure and flow waveforms, although it is not clear which perform best. The primary aim of this study was to assess the performance of CPE methods to estimate cardiac, arterial and microvascular parameters. The secondary aim was to use these methods with a model of the circulation to estimate central blood pressure (CBP) from waveforms which can be acquired non-invasively.
Methods
An extensive literature review of current CPE methods was performed: ten methods were identified for estimating characteristic impedance (Z); nine for both asymptotic outflow pressure (Pout) and peripheral compliance (C); six for left ventricular ejection time (LVET); and four for pulse wave velocity (PWV). These methods were implemented in a common framework. They were used to estimate the CV parameters of 258simulated healthy subjects from an in silico pulse wave database (previously described in [2]). Imposed CV parameters and corresponding brachial pressure and central (aortic) flow waveforms from each subject were used as reference data. The performance of CPE methods was assessed individually using the mean percentage error (PE) between estimated and reference values. In addition, the estimated CV parameters were used as inputs to a 3-element Windkessel (0-D) model to estimate CBP. For each subject, CBP was estimated using the CV parameters estimated by every possible combination of CPE methods. The optimal combination of CPE methods was the one which resulted in the lowest mean root-mean-square error (RMSE)between the reference and estimated CBP waveforms for the entire dataset.
Results
The individual analysis indicated that the optimal CPE methods were: a ‘first-derivative of pressure’ analysis for LVET (PE: 8.1 %); 50 % of diastolic blood pressure (DBP) for Pout (8.4 %); an ‘iterative DBP’ method forC (29.7 %); and an ‘early-systole PQ-loop’ method for Z (68.6 %). The mean RMSE of CBP waveforms estimated using this combination of CPE methods was 2.4 mmHg. This was also the optimal combination of CPE methods, since remaining combinations resulted in larger mean RMSE values.
Conclusions
We have identified the best-performing CPE methods through comparison with known reference values in an in silico pulse wave database. When used as inputs to a 3-element Windkessel (0-D) model, these methods provided accurate estimation of CBP. The use of an in silico dataset has the advantage that the reference CV parameters were known precisely and were varied across a wide range of values. Future work will assess the performance of methods in vivo.
References
[1] McEniery et al. Central blood pressure: Current evidence and clinical importance. European Heart Journal, 35(26),1719–1725, 2014.
[2] Charlton P.H. et al. Modelling arterial pulse wave propagation during healthy ageing, In World Congress ofBiomechanics 2018, Dublin, Ireland, 2018.
Cardiovascular (CV) parameters can be used as markers of cardiovascular risk, and as inputs to computational models of arterial blood flow. However, they are difficult to measure directly. Several CV parameter estimation(CPE) methods have been proposed to estimate parameters from non-invasive blood pressure and flow waveforms, although it is not clear which perform best. The primary aim of this study was to assess the performance of CPE methods to estimate cardiac, arterial and microvascular parameters. The secondary aim was to use these methods with a model of the circulation to estimate central blood pressure (CBP) from waveforms which can be acquired non-invasively.
Methods
An extensive literature review of current CPE methods was performed: ten methods were identified for estimating characteristic impedance (Z); nine for both asymptotic outflow pressure (Pout) and peripheral compliance (C); six for left ventricular ejection time (LVET); and four for pulse wave velocity (PWV). These methods were implemented in a common framework. They were used to estimate the CV parameters of 258simulated healthy subjects from an in silico pulse wave database (previously described in [2]). Imposed CV parameters and corresponding brachial pressure and central (aortic) flow waveforms from each subject were used as reference data. The performance of CPE methods was assessed individually using the mean percentage error (PE) between estimated and reference values. In addition, the estimated CV parameters were used as inputs to a 3-element Windkessel (0-D) model to estimate CBP. For each subject, CBP was estimated using the CV parameters estimated by every possible combination of CPE methods. The optimal combination of CPE methods was the one which resulted in the lowest mean root-mean-square error (RMSE)between the reference and estimated CBP waveforms for the entire dataset.
Results
The individual analysis indicated that the optimal CPE methods were: a ‘first-derivative of pressure’ analysis for LVET (PE: 8.1 %); 50 % of diastolic blood pressure (DBP) for Pout (8.4 %); an ‘iterative DBP’ method forC (29.7 %); and an ‘early-systole PQ-loop’ method for Z (68.6 %). The mean RMSE of CBP waveforms estimated using this combination of CPE methods was 2.4 mmHg. This was also the optimal combination of CPE methods, since remaining combinations resulted in larger mean RMSE values.
Conclusions
We have identified the best-performing CPE methods through comparison with known reference values in an in silico pulse wave database. When used as inputs to a 3-element Windkessel (0-D) model, these methods provided accurate estimation of CBP. The use of an in silico dataset has the advantage that the reference CV parameters were known precisely and were varied across a wide range of values. Future work will assess the performance of methods in vivo.
References
[1] McEniery et al. Central blood pressure: Current evidence and clinical importance. European Heart Journal, 35(26),1719–1725, 2014.
[2] Charlton P.H. et al. Modelling arterial pulse wave propagation during healthy ageing, In World Congress ofBiomechanics 2018, Dublin, Ireland, 2018.
| Original language | English |
|---|---|
| Title of host publication | 14th International Symposium on Biomechanics in Vascular Biology and Cardiovascular Disease |
| Publisher | Queen Mary, University of London |
| Pages | 12 |
| Number of pages | 1 |
| Publication status | Published - 2019 |
| Event | 14th International Symposium on Biomechanics in Vascular Biology and Cardiovascular Disease - Imperial College London, London, United Kingdom Duration: 11 Apr 2019 → 12 Apr 2019 https://www.sems.qmul.ac.uk/events/bvbcd2019/ |
Conference
| Conference | 14th International Symposium on Biomechanics in Vascular Biology and Cardiovascular Disease |
|---|---|
| Country/Territory | United Kingdom |
| City | London |
| Period | 11/04/2019 → 12/04/2019 |
| Internet address |