Reduced-Order Computational Models for Central Blood Pressure Estimation: Two Clinical Scenarios

TY - CHAP

T1 - Reduced-Order Computational Models for Central Blood Pressure Estimation: Two Clinical Scenarios

AU - Mariscal Harana, Jorge

AU - Charlton, Peter Harcourt

AU - Sherwin, Spencer J.

AU - Alastruey-Arimon, Jordi

PY - 2018/9

Y1 - 2018/9

N2 - Introduction Central blood pressure (CBP) is a better cardiovascular risk indicator than brachial pressure. The current gold standard for the assessment of CBP is a catheter, which is invasive. We propose a new method to assess CBP from aortic blood flow and peripheral blood pressure, which can be measured non-invasively. This clinical data is used to generate reduced-order (0-D or 1-D) computational models which provide accurate CBP estimates. Our analysis is based on basic haemodynamic principles and on phenomena occurring directly in the ascending aorta. The aims of this study are twofold: to compare reduced-order CBP estimation models; and to find optimum parameter estimation methods for two common clinical scenarios. Methods Two 0-D models, namely the two-element and three-element Windkessel models, and one 1-D model of the aorta were studied. A population of ‘virtual’ (computed) healthy subjects was generated using a 1-D model of the systemic arteries based on [ref]. Using each subject’s unique pressure and flow waveforms as reference data, parameter estimation methods were developed. The optimal method was identified based on the errors between reference and estimated systolic (SBP) and pulse (PP) pressure values. Two common clinical scenarios were considered: (i) a pressure waveform is available, and (ii) only diastolic (DBP) and mean (MBP) blood pressure values are available. These optimum methods were applied to two clinical cohorts: hypertensives and normotensives. Results & Discussion For scenario (i), average relative errors in SBP and PP were <3% and <8%, respectively, across both clinical cohorts. Average relative errors in SBP and PP for scenario (ii) were <6% and <17%. In both scenarios, average SBP and PP errors were higher for the hypertensive cohort. Conclusion Parameter estimation methods which account for two different clinical scenarios were developed and compared. All three models performed similarly when estimating landmark values of blood pressure. However, the 1-D model was superior at reproducing space-dependent features of the pressure waveform.

AB - Introduction Central blood pressure (CBP) is a better cardiovascular risk indicator than brachial pressure. The current gold standard for the assessment of CBP is a catheter, which is invasive. We propose a new method to assess CBP from aortic blood flow and peripheral blood pressure, which can be measured non-invasively. This clinical data is used to generate reduced-order (0-D or 1-D) computational models which provide accurate CBP estimates. Our analysis is based on basic haemodynamic principles and on phenomena occurring directly in the ascending aorta. The aims of this study are twofold: to compare reduced-order CBP estimation models; and to find optimum parameter estimation methods for two common clinical scenarios. Methods Two 0-D models, namely the two-element and three-element Windkessel models, and one 1-D model of the aorta were studied. A population of ‘virtual’ (computed) healthy subjects was generated using a 1-D model of the systemic arteries based on [ref]. Using each subject’s unique pressure and flow waveforms as reference data, parameter estimation methods were developed. The optimal method was identified based on the errors between reference and estimated systolic (SBP) and pulse (PP) pressure values. Two common clinical scenarios were considered: (i) a pressure waveform is available, and (ii) only diastolic (DBP) and mean (MBP) blood pressure values are available. These optimum methods were applied to two clinical cohorts: hypertensives and normotensives. Results & Discussion For scenario (i), average relative errors in SBP and PP were <3% and <8%, respectively, across both clinical cohorts. Average relative errors in SBP and PP for scenario (ii) were <6% and <17%. In both scenarios, average SBP and PP errors were higher for the hypertensive cohort. Conclusion Parameter estimation methods which account for two different clinical scenarios were developed and compared. All three models performed similarly when estimating landmark values of blood pressure. However, the 1-D model was superior at reproducing space-dependent features of the pressure waveform.

M3 - Poster abstract

SP - 327

BT - BioMedEng18 Conference Proceedings

T2 - BioMedEng18

Y2 - 6 September 2018 through 7 September 2018

ER -