Evaluation of Personalised Canine Electromechanical Models

Sophie Giffard-Roisin, Stéphanie Marchesseau, Loïc Le Folgoc, Hervé Delingette, Maxime SermesantSTACOM 2014 · DOI

TL;DR

This paper presents a novel computational model that integrates electrical and mechanical aspects of the canine heart to simulate its function. By calibrating this model with real-world canine heart data, the researchers demonstrate its ability to accurately predict key physiological indicators like pressure-volume loops and ejection fraction, paving the way for personalized cardiac therapy simulations.

What Problem Does This Paper Address?

Understanding and predicting the behavior of the heart is crucial for diagnosing and treating cardiovascular diseases. This involves simulating the complex interplay between the heart's electrical system, which initiates contraction, and its mechanical system, which pumps blood. While models exist for each aspect, integrating them into a cohesive, personalized simulation that can predict responses to therapies remains a significant challenge. This paper tackles this by developing and validating an electromechanical model for the canine heart, a common preclinical model, aiming to bridge the gap between fundamental cardiac physics and clinical applications.

Methods

The authors employ a sophisticated computational framework built on the SOFA platform to couple electrophysiology and mechanics.

  • Electrophysiology: An Eikonal model is used to simulate the propagation of the electrical wave across the myocardium. This model captures the depolarization front's movement, with parameters like local conduction velocity and anisotropic ratios derived from DTI imaging.
  • Mechanics: The Bestel-Clement-Sorine (BCS) model is utilized for simulating the active and passive mechanical behavior of the heart. This model accounts for microscopic muscle fiber phenomena and thermodynamic laws, incorporating both active contraction (actin-myosin dynamics) and passive elasticity. A Windkessel model is used to represent the circulation.
  • Personalization and Calibration: The model is personalized using anatomical meshes and fiber directions from individual canine hearts. A crucial aspect is the parameter estimation process, which involves a sensitivity study and a quantitative calibration using the Unscented Transform Algorithm. This algorithm efficiently calibrates key parameters by minimizing the difference between simulated and measured physiological data (e.g., ejection fraction, flow).

Datasets

The study utilizes data from the STACOM'2014 challenge, which provided:

  • Left Ventricular (LV) geometry for four healthy canine hearts.
  • Myocardial fiber directions.
  • LV volume and LV pressure curves for a complete cardiac cycle.
  • Tagged MRI data for ground truth displacement fields.

This dataset represents a benchmark for evaluating cardiac models, offering a comprehensive set of physiological measurements for validation.

Key Insights and Ideas

  • Electromechanical Coupling is Key: The paper strongly emphasizes that a realistic simulation of heart function requires the tight integration of electrical and mechanical processes. The timing of electrical depolarization directly influences the mechanical contraction, and this coupling is essential for accurate physiological prediction.
  • Personalization Matters: By calibrating the model with individual canine heart data, the researchers show that personalized models can capture the unique physiological responses of each heart, including variations in ejection fraction. This is a critical step towards developing patient-specific cardiac simulations.
  • Model Robustness: The BCS mechanical model's ability to capture phenomena like the Starling effect (where stroke volume adapts to end-diastolic volume) highlights its potential for simulating various cardiac conditions.
  • Validation is Crucial: The quantitative comparison with ground truth data from the STACOM challenge provides strong evidence for the model's validity and its ability to reproduce key cardiac indicators.

Key Results

The study demonstrates promising results in simulating canine cardiac function:

  • Realistic Pressure-Volume Loops: The simulated LV pressure-volume diagrams closely match the ground truth measurements for all four canine cases, indicating accurate overall cardiac function simulation.
  • Accurate Ejection Fraction: The model successfully reproduces the variability in ejection fraction across the four cases, with estimated values closely aligning with measured values (e.g., Case 1: Measured 53%, Estimated 43%; Case 3: Measured 30%, Estimated 33%).
  • Reproducible Volume and Flow Curves: The simulated LV volume and outgoing flow curves show good agreement with measured data over a cardiac cycle, further validating the model's dynamic response.
  • Realistic End-Systolic Geometry: Visualizations of the simulated end-systolic geometry align well with the ground truth, demonstrating the model's ability to capture the heart's shape changes during contraction.

Limitations and Future Work

  • Myocardium Incompressibility: The authors acknowledge that the simulated myocardium's wall thickness variation is not perfectly reproduced, suggesting the muscle might not be modeled as perfectly incompressible. This could be an area for future refinement.
  • RV Influence: For simplicity, the model primarily focuses on the LV, approximating the RV's influence as a boundary condition. Future work could incorporate a more detailed RV model.
  • Pathology Simulation: While the current study focuses on healthy hearts, the ultimate goal is to simulate pathological conditions and predict treatment responses. This paper lays the groundwork for such future investigations.
  • Further Calibration: The authors suggest that further calibration of more parameters could lead to even more precise simulations.

Summary updated

BibTeX
@inproceedings{giffardroisin2014evaluation,
  author = {Sophie Giffard-Roisin and Stéphanie Marchesseau and Loïc Le Folgoc and Hervé Delingette and Maxime Sermesant},
  title = {{Evaluation of Personalised Canine Electromechanical Models}},
  booktitle = {STACOM},
  year = {2014},
  doi = {10.1007/978-3-319-14678-2\_8},
}
Figure