Clinical application of a model-based cardiac stroke volume estimation method. Issue 2 (2020)

• Record Type: Journal Article
• Authors: Smith, Rachel; Balmer, Joel; Pretty, Christopher G.; Shaw, Geoffrey M.; Chase, J. Geoffrey
• Is Part Of: IFAC-PapersOnLine. Volume 53:Issue 2(2020)
• Page Start: 16137
• Page End: 16142
• Publication Date: 2020
• Subjects: Pulse contour analysis -- Pressure contour analysis -- Windkessel model -- Stroke volume -- Cardiac output -- Hemodynamic monitoring -- Intensive care; Automatic control -- Periodicals; 629.805
• DOI: 10.1016/j.ifacol.2020.12.435
• Languages: English
• ISSNs: 2405-8963

Abstract

Abstract: A system is needed for monitoring stroke volume (SV) and cardiac output (CO) in unstable patients which is non-additionally invasive, reproducible and reliable in a variety of physiological states. This study evaluates SV estimation accuracy of a non-additionally invasive pulse contour analysis method implemented using a 3-element Windkessel model. The model lumps the properties of the arterial system into 3 parameters: characteristic impedance of the proximal aorta (Z), and resistance (R) and compliance (C) of the systemic arteries. Parameter products ZC and RC are dynamically identified from measured femoral arterial pressure waveforms, and Z is a static parameter obtained by calibration. The accuracy of the model is evaluated for a cohort of 9 liver transplant patients, using thermodilution as a reference method. Data were obtained from Vital Data Bank (VitalDB). The study thus provides independent assessment of a pulse contour analysis, proven in animal studies, in an uncontrolled clinical environment. The model tracked trends in SV well over the course of the surgery. However, the 95% range for percentage error was -88% to +53%, outside acceptable limits of ±45%. Main areas contributing to error for the model include the changing extent of reflected waves in the arterial system, dynamic response characteristics of fluid-filled pressure catheters, and the assumption of fixed Z parameter. Further investigation is needed to consider the contribution of these factors to SV estimation error by the model.