Cardiovascular Lumped Parameter Model (LPM)

• Cardiovascular lumped parameter models (LPM) are zero-dimensional models that represent blood circulation using electrical circuit analogies with resistors, capacitors, and inductors.
• They simulate global hemodynamics with low computational cost and support multi-scale coupling, patient-specific calibration, and uncertainty quantification.
• Underlying equations like Ohm’s law and elastance formulations enable analysis of complex cardiovascular conditions and inform clinical interventions.

Searching arXiv for recent and foundational papers on cardiovascular lumped parameter models to support the encyclopedia entry. arxiv_search query: "cardiovascular lumped parameter model zero-dimensional elastance Windkessel" max_results: 10 Cardiovascular lumped parameter modeling (LPM), also termed zero-dimensional (0D) modeling, represents the circulation as a network of discrete compartments linked by resistive, compliant, inertial, and valvular elements. In the electrical-hydraulic analogy used throughout this literature, blood pressure is treated as electric potential, blood flow as current, volume as charge, vascular resistance/compliance/inertance as resistance/capacitance/inductance, and valves as diodes. The resulting ordinary differential equation or differential-algebraic equation systems are used to simulate global hemodynamics, complete closed-loop circulation, multiscale 3D-0D coupling, and patient-specific inference with comparatively low computational cost (Ghasemalizadeh et al., 2014).

1. Electrical-hydraulic formulation

The canonical LPM formalism maps hemodynamic quantities to circuit variables and elements. This representation is explicit in whole-body cardiovascular models, ocular-cardiovascular couplings, and closed-loop circulation models, and it provides the common language through which systemic vessels, pulmonary vessels, cardiac chambers, and valves are encoded (Ghasemalizadeh et al., 2014).

Physiological quantity or structure	Electrical analog
Blood pressure	Voltage
Blood flow	Current
Blood volume	Charge
Vascular resistance	Resistor
Vessel compliance	Capacitor
Blood inertia	Inductor
Heart valve	Diode

In detailed vessel-level formulations, each vessel is represented by a three-element $RLC$ block: a resistor for viscous resistance, an inductor for blood inertia, and a capacitor for elastic compliance. In the 2014 whole-body model, the corresponding element formulas are

$$R = \frac{8\mu l}{A^2}, \qquad L = \frac{9l\rho}{4A}, \qquad C = \frac{30\pi r^3}{2Eh}.$$

The same paper states the unit conversions $1~\text{mmHg} = 1~\text{V}$, $$1~\text{Pa}\cdot\text{s}/\text{ml} = 1~\text{k}\Omega$$, $0.01~\text{ml/Pa} = 1~\mu\text{F}$, and $$1~\text{Pa}\cdot\text{s}^2/\text{ml} = 1~\mu\text{H}$$ (Ghasemalizadeh et al., 2014).

The chamber representation depends on physiology. Atria may be modeled as non-contractile capacitors plus resistors, whereas ventricles are commonly modeled through time-varying capacitance or, in elastance-based formulations, time-varying pressure-volume relations. Valves may be ideal diodes, non-ideal diodes with pressure-dependent resistances, or dynamic orifice elements with explicit cusp motion and pressure-loss laws, depending on the required level of detail (Ghasemalizadeh et al., 2014).

2. Governing equations and constitutive structure

The mathematical core of a cardiovascular LPM is conservation of mass and momentum expressed at the compartment level. Across the cited models, the recurrent relations are the fluid analog of Ohm’s law,

$Q = \frac{\Delta P}{R},$

the compliant-compartment law,

$\frac{dP}{dt} = \frac{Q_{in}-Q_{out}}{C},$

and the inertial relation,

$L\frac{dQ}{dt} = \Delta P.$

In volume-pressure form, compliant compartments satisfy relations such as $V = C\cdot P$, while chamber models often use

$$R = \frac{8\mu l}{A^2}, \qquad L = \frac{9l\rho}{4A}, \qquad C = \frac{30\pi r^3}{2Eh}.$$0

with $$R = \frac{8\mu l}{A^2}, \qquad L = \frac{9l\rho}{4A}, \qquad C = \frac{30\pi r^3}{2Eh}.$$1 a time-varying elastance (Tonini et al., 2023).

This framework supports several algebraic and dynamical specializations. In five-compartment and multi-compartment pulsatile models, chamber and vascular states are written in state-space form $$R = \frac{8\mu l}{A^2}, \qquad L = \frac{9l\rho}{4A}, \qquad C = \frac{30\pi r^3}{2Eh}.$$2, with outputs such as ventricular pressure and volume used for calibration and uncertainty quantification (Marquis et al., 2017). In more general metamodeling studies, 0D circulation is expressed as a system of differential-algebraic equations,

$$R = \frac{8\mu l}{A^2}, \qquad L = \frac{9l\rho}{4A}, \qquad C = \frac{30\pi r^3}{2Eh}.$$3

which accommodates nonlinear resistances, inertances, compliances, valves, and time-varying compliance elements in a unified formulation (Hanna et al., 2024).

Valve laws are a major differentiator among models. At the simplest level, diodic logic permits forward flow only under a favorable pressure gradient. More elaborate formulations use quadratic pressure-drop laws, valve-angle dynamics, or geometry- and flow-regime-dependent loss coefficients. This progression from ideal-diode closure to dynamically resolved valvular loss models is one of the main axes along which cardiovascular LPMs gain physiological specificity (Laubscher et al., 2022).

3. Architectural variants and physiological granularity

Cardiovascular LPMs span a broad range of structural complexity. The whole-body model of “Exact Modeling of Cardiovascular System Using Lumped Method” describes a 36-vessel model and cardiac system, includes pulmonary circulation, both atria, left and right ventricles with equivalent circuits, and treats bifurcations to capture wave reflection and transmission. The model is subdivided into heart, upper body, and lower body, and the circulation is explicitly closed so that total current is recirculated without leakage (Ghasemalizadeh et al., 2014). A closely related 2014 model divides the ascending aorta into 27 segments and states that the full system uses about 120 $$R = \frac{8\mu l}{A^2}, \qquad L = \frac{9l\rho}{4A}, \qquad C = \frac{30\pi r^3}{2Eh}.$$4 segments distributed across 36 distinct vessels and the cardiac system (Ghasemalizadeh et al., 2014).

Simpler but still mechanistic architectures are also common. A recent comparative study investigates a three-compartment single-ventricle model consisting of Left Ventricle, Systemic Arteries, and Systemic Veins, and an eight-compartment four-chamber model comprising all cardiac chambers plus systemic and pulmonary arteries and veins (Sasikumar, 28 Dec 2025). A six-compartment hemodynamic model, CVSim-6, uses left and right ventricles, systemic arteries and veins, and pulmonary arteries and veins, with 23 parameters and six time-dependent pressures as states (Tong et al., 2024). For intraoperative hypotension, a previously developed 0D LPM with 12 compartments represents the heart and major circulatory regions and uses elastance-driven cardiac pressure-volume relations together with systemic and pulmonary $$R = \frac{8\mu l}{A^2}, \qquad L = \frac{9l\rho}{4A}, \qquad C = \frac{30\pi r^3}{2Eh}.$$5 elements (Thiel et al., 17 Sep 2025).

Architectural expansion may also be domain-specific. The Eye2Heart model is a closed-loop mathematical framework that couples a cardiovascular subsystem to an ocular subsystem and formulates the entire problem as 23 coupled ordinary differential equations. In addition to heart, aorta, vena cava, lungs, and systemic circuits, it includes retinal circulation and a parallel eye branch, connected to the aorta and vena cava through dedicated $$R = \frac{8\mu l}{A^2}, \qquad L = \frac{9l\rho}{4A}, \qquad C = \frac{30\pi r^3}{2Eh}.$$6 pathways (Sala et al., 8 Apr 2025). This suggests that LPM architecture is typically chosen according to the target observables and the physiological interfaces that must be resolved.

4. Calibration, identifiability, and uncertainty quantification

Because LPMs are often overparameterized relative to available observations, calibration and identifiability are central problems. Classical deterministic workflows use physiologically plausible nominal values, sensitivity analysis, subset selection, and nonlinear least squares. In a pulsatile five-compartment model calibrated to rat left ventricular pressure and volume data, sensitivity analysis and structured correlation analysis reduced the inferable parameter set to five identifiable parameters, after which frequentist and Bayesian uncertainty quantification showed that the computed intervals capture measurement and model error (Marquis et al., 2017).

Recent work has proposed more specialized calibration procedures. A correlation matrix calibration method uses the sample Pearson correlation matrix between parameters and outputs as a surrogate for the local loss-function gradient, while a hybrid CMC-L-BFGS-B workflow combines this global randomized search with local quasi-Newton refinement. In in silico tests, the reported success rates were 19 out of 20 for CMC, 12 out of 20 for L-BFGS-B, and 17 out of 20 for the hybrid approach; the same study states that CMC was faster than the combined method and L-BFGS-B and achieved the best results in estimating the parameters in the case of noisy data (Tonini et al., 2024).

Bayesian calibration has been used to address non-uniqueness and to propagate uncertainty into treatment-relevant predictions. In a 12-compartment LPM for intraoperative hypotension, a non-intrusive polynomial chaos expansion surrogate was trained on 5,000 LPM simulations, and Bayesian Markov chain Monte Carlo with NUTS in PyMC was used to infer parameter distributions. The same study reports diagnostic thresholds such as rank-normalized split $$R = \frac{8\mu l}{A^2}, \qquad L = \frac{9l\rho}{4A}, \qquad C = \frac{30\pi r^3}{2Eh}.$$7, uses measurement-uncertainty models for pressure and flow directly in the likelihood, and shows that sequential updating can reduce the coefficient of variation of parameter posteriors by up to 94% (Thiel et al., 17 Sep 2025).

Alternative inference methods reflect the same identifiability concern. The inVAErt framework was used on a six-compartment hemodynamic model to reconstruct full solution manifolds rather than single point estimates, explicitly targeting structural and practical non-identifiability; the study reports that 12 principal modes explain 99% variance in the decoded solution manifold (Tong et al., 2024). A modified unscented Kalman filter for a 10-parameter, four-observable cardiovascular LPM addresses rank deficiency by increasing the observation dimension through a Kalman interval and reports recovery of almost all parameters to over 98% accuracy, over 90% of the time, on a challenging target data set of 50 ten-parameter samples (Thornton et al., 24 Dec 2025). For non-invasive left-ventricle elastance estimation, a hybrid Adam-L-BFGS optimization with forward-mode automatic differentiation yielded mean absolute percentage errors ranging from 6.67% to 14,14%, with errors of approximately 2% in simulated aortic stenosis and mitral regurgitation when arterial pressure and valvular flow rate measurements were included (Laubscher et al., 2022).

5. Multiscale coupling and boundary-condition roles

Although 0D models are intrinsically reduced-order, they are routinely embedded within multiscale frameworks. In a geometric multiscale model of the human left heart, a 3D Navier-Stokes simulation of the left heart is coupled to a 0D closed-loop circulation model representing the remaining circulation. The coupling is implemented at the pulmonary veins and ascending aorta through a two-way, time-dependent Dirichlet–Neumann interface: updated 0D pressures are applied as Neumann boundary conditions to the 3D fluid problem, and 3D-derived inlet and outlet flow rates are returned to the 0D model at each time step (Zingaro et al., 2021).

A closely related line of work couples 3D cardiac electromechanics to a 0D circulation model through a left-ventricular pressure-volume constraint. In this formulation, the 0D left ventricle is replaced by the 3D electromechanical ventricle, $$R = \frac{8\mu l}{A^2}, \qquad L = \frac{9l\rho}{4A}, \qquad C = \frac{30\pi r^3}{2Eh}.$$8 acts as a Lagrange multiplier, and the volume consistency condition $$R = \frac{8\mu l}{A^2}, \qquad L = \frac{9l\rho}{4A}, \qquad C = \frac{30\pi r^3}{2Eh}.$$9 is enforced within a segregated numerical framework. The derivation emphasizes energy-consistent boundary conditions and proves an overall mechanical energy balance for the coupled 3D-0D problem (Regazzoni et al., 2020). The numerical counterpart shows that the segregated algorithm remains stable through the different phases of the heartbeat and reproduces physiologically relevant responses to changes in preload, afterload, and contractility (Regazzoni et al., 2020).

At the outlet level, LPMs are widely used as reduced downstream circulation models. A recent discussion of cardiovascular outflow conditions formulates the three-element Windkessel model at an outlet $1~\text{mmHg} = 1~\text{V}$0 as

$1~\text{mmHg} = 1~\text{V}$1

with $1~\text{mmHg} = 1~\text{V}$2 the outlet flow rate, $1~\text{mmHg} = 1~\text{V}$3 the computational outlet pressure, $1~\text{mmHg} = 1~\text{V}$4 the proximal pressure, and $1~\text{mmHg} = 1~\text{V}$5 the distal pressure. The same source notes that these equations are solved jointly with the 3D Navier-Stokes system and that parameter assignment can be based on physiological measurements or patient-scaled quantities such as outlet areas and systemic vascular resistance (Mirzaiyan et al., 4 Mar 2025). A similar reduction principle appears at the valve scale in a reduced 3D-0D fluid-structure interaction model of the aortic valve, where leaflet motion is compressed into a single opening coefficient governed by a momentum-balance ODE with curvature-based elasticity (Fumagalli et al., 2021).

6. Disease modeling, devices, and organ-to-organ coupling

A principal use of cardiovascular LPMs is the controlled study of pathology and intervention. In a stochastic model of paroxysmal atrial fibrillation, irregular RR intervals are represented by an exponentially modified Gaussian distribution, atrial contraction is removed by setting atrial elastances constant, and left-ventricular contractility is reduced according to $1~\text{mmHg} = 1~\text{V}$6. The reported consequences include a decrease in cardiac output from $1~\text{mmHg} = 1~\text{V}$7 l/min to $1~\text{mmHg} = 1~\text{V}$8 l/min, a decrease in ejection fraction from $1~\text{mmHg} = 1~\text{V}$9 to $$1~\text{Pa}\cdot\text{s}/\text{ml} = 1~\text{k}\Omega$$0, and an increase in left atrial volume from $$1~\text{Pa}\cdot\text{s}/\text{ml} = 1~\text{k}\Omega$$1 ml to $$1~\text{Pa}\cdot\text{s}/\text{ml} = 1~\text{k}\Omega$$2 ml (Scarsoglio et al., 2014).

Patient-specific LPMs have also been used to reveal latent disease signatures not directly apparent in raw measurements. In a COVID-19 study calibrated on clinical data from 58 patients hospitalized for COVID-19-related pneumonia, 29 patients could be successfully calibrated. Statistical analysis of reliable model outputs showed significant increases in right ventricular systolic pressure, diastolic and mean pulmonary arterial pressure, and capillary wedge pressure, whereas systemic vascular parameters were not significantly altered. The same work states that raw clinical data, without the support of a mathematical model, were unable to detect the effects of COVID-19-related pneumonia (Tonini et al., 2023).

Cardiovascular LPMs are increasingly extended to coupled organ systems and extracorporeal support. The Eye2Heart model bridges systemic circulation and ocular hemodynamics, reproduces standard cardiovascular indices such as EDV, ESV, SV, CO, and EF together with retinal blood-flow variables, and is used to study the effects of intraocular pressure elevation and reduced left-ventricular compliance on both local ocular and global systemic circulation (Sala et al., 8 Apr 2025). In intensive-care applications, a comprehensive LPM including extracorporeal membrane oxygenation and continuous renal replacement therapy was calibrated on 30 clinical data points from eight veno-arterial ECMO patients and achieved $$1~\text{Pa}\cdot\text{s}/\text{ml} = 1~\text{k}\Omega$$3 between simulation and experimental data; across nine CRRT-ECMO connection schemes, pulmonary artery pressure changes of up to 202.5 % were reported, highly dependent on ECMO flow (Thiel et al., 2024).

Valvular and vascular applications further illustrate the range of the approach. A multi-compartment cardiovascular model with a geometry- and flow-regime-dependent valve loss formulation reports that previously published valve models under predicts expected severely stenosed peak and mean transvalvular pressure drops by approximately 47% and 30% respectively, whereas the newly proposed model under predicts the peak pressure drop by 20% and over predicts mean pressure drop by 7% (Laubscher et al., 2022). In a segmented arterial-tree model, pressure waveforms in the brachial artery were found to be so near to aortic pressure that the brachial signal was proposed as a usable surrogate, and localized obstructions in the ascending aorta and brachial artery were examined by directly modifying segmental properties (Ghasemalizadeh et al., 2014).

7. Model dependence, limitations, and current directions

Comparative studies show that parameter importance is not invariant across model structures. In a direct comparison between a simplified single-ventricle model and a detailed four-chamber model, sensitivity rankings differed substantially, with dominant parameters in the reduced model not matching those in the more detailed one. The same study observed that $$1~\text{Pa}\cdot\text{s}/\text{ml} = 1~\text{k}\Omega$$4 and $$1~\text{Pa}\cdot\text{s}/\text{ml} = 1~\text{k}\Omega$$5 were nearly identical in the tested settings and stated that broader or more realistic parameter ranges may reveal wider interaction effects (Sasikumar, 28 Dec 2025). A plausible implication is that sensitivity results are model- and output-dependent rather than transferable as universal rankings.

Identifiability remains a structural limitation. The inVAErt study distinguishes structural non-identifiability, in which a manifold in parameter space is mapped to a common output, from practical non-identifiability due to limited data, model misspecification, or noise corruption (Tong et al., 2024). The Bayesian hypotension study likewise reports that deterministic calibration yielded many local solutions with notably different sensitivities, whereas posterior-based uncertainty-aware sensitivity analysis produced tighter credible intervals and more stable parameter rankings (Thiel et al., 17 Sep 2025). These results indicate that a single calibrated parameter set may be insufficient when clinical observables are sparse or noisy.

Boundary treatment and multiscale closure are also nontrivial. The outlet-condition study emphasizes that wrong outflow setup can lead to unphysical results and notes that LPM-based outlet models assume uniform pressure or stress across the outlet section and cannot always account for complex feedback between local 3D hemodynamics and downstream vascular properties (Mirzaiyan et al., 4 Mar 2025). Conversely, the COVID-19 study shows that model-derived latent variables can reveal clinically important pulmonary abnormalities that are not statistically evident in raw measurements alone (Tonini et al., 2023). Current directions documented in this literature therefore include patient-specific parameterization, continual or amortized updating, non-invasive personalization, surrogate-assisted uncertainty quantification, and broader multiscale digital-twin applications in personalized medicine and clinical decision support (Hanna et al., 2024).