Estimation of Reaction, Diffusion and Restitution Parameters for a 3D Myocardial Model Using Optical Mapping and MRI

Jatin Relan, Mihaela Pop, Hervé Delingette, Graham A. Wright, Nicholas Ayache, Maxime SermesantSTACOM/CESC 2010 · DOI

TL;DR

This paper presents a novel method to create patient-specific 3D cardiac electrophysiology models by estimating key reaction, diffusion, and restitution parameters. By fusing optical mapping and MRI data, they can accurately personalize models, paving the way for improved prediction of cardiac arrhythmias and guiding therapies like RF ablation.

What Problem Does This Paper Address?

Translating computational cardiac electrophysiology models from research into clinical applications requires patient-specific models. Current models often rely on generic parameters, limiting their predictive power for individual patients. The challenge lies in accurately estimating the numerous parameters that govern the complex electrical behavior of the heart, such as how electrical signals propagate (diffusion), how cells generate action potentials (reaction), and how these properties change with heart rate (restitution). This paper addresses this gap by developing a method to personalize a simplified ionic model (Mitchell-Schaeffer model) using a combination of optical mapping and MRI data, aiming to create a truly heart-specific 3D model.

Methods

The authors employ a multi-resolution technique to estimate parameters of the Mitchell-Schaeffer (MS) model. The core of their approach involves:

  • Model Personalization: This is achieved by matching simulated features (Depolarization Times, Action Potential Duration (APD), and Conduction Velocity (CV) restitution) to those extracted from experimental optical mapping data.
  • Parameter Estimation: They estimate several key parameters:
  • Diffusion Tensor (d): This parameter represents tissue conductivity and is estimated using a multi-resolution approach, starting with a coarse domain decomposition and progressively refining it.
  • Reaction and Restitution Parameters (Tin, Tout, Topen, Tclose): These parameters govern the ionic currents and the rate-dependent behavior of APD and CV. They are estimated by minimizing the error between simulated and measured APD and CV restitution curves.
  • Data Fusion: Optical mapping data (providing temporal electrical activity) is fused with Diffusion Tensor MRI (DT-MRI) data (providing anatomical geometry and fiber orientation) to create a spatially accurate 3D model.
  • Iterative Adjustment: A trust region method is used for iterative optimization of parameters, with a focus on minimizing the sum of squared differences between simulated and measured data.

Datasets

The study utilizes ex-vivo porcine hearts. Two hearts were used, each providing optical mapping and MRI datasets:

  • Dataset 1: Optical mapping data from one heart, paced at a single location (LV apex) but at three different frequencies (0.7 Hz, 1.1 Hz, 1.2 Hz). This dataset was used for estimating APD and CV restitution.
  • Dataset 2: Optical mapping data from a second heart, paced at 1.1 Hz but from two different locations (RV epicardium and LV endocardium). This dataset was used for personalization using Depolarization Times and APD maps.

Key Insights and Ideas

  • Fusion of Optical Mapping and MRI is Crucial: Combining the high temporal resolution of optical mapping with the anatomical detail of MRI is essential for creating accurate, patient-specific 3D cardiac models.
  • Multi-Resolution Parameter Estimation: A multi-resolution approach for estimating diffusion parameters (like tissue conductivity) is effective in handling spatial variations and improving accuracy.
  • Restitution Properties Reveal Heterogeneity: APD and CV restitution curves are sensitive to underlying physiological differences and can reveal important heterogeneities in cardiac tissue properties, even in healthy hearts.
  • Robustness to Pacing Location: The personalization method demonstrates low sensitivity to the pacing location, which is a significant advantage for clinical applications where precise pacing information might be ambiguous.
  • Identification of Low Conductivity Areas: The estimated diffusion parameter (d) successfully identified regions of low conductivity, which could correspond to areas of ischemia or other pathologies.

Key Results

  • The personalization method significantly reduced the mean square error of APD and CV restitution curves, achieving an error reduction from 20.35 to 0.54 for Dataset 1.
  • The estimated parameter maps (e.g., for d and Tclose) showed spatial variations, indicating heterogeneity in cardiac tissue properties.
  • The method was able to locate low conductivity areas, consistent with visual inspection of the data.
  • Comparison of results from different pacing locations in Dataset 2 showed similar parameter estimates, confirming the robustness of the method.

Limitations and Future Work

  • Validation with Arrhythmias: The authors acknowledge the need to validate the model predictions for more complex scenarios, such as arrhythmias, which involve varying heart rates and pacing locations.
  • Endocardial Mapping Ambiguity: While the method shows robustness to pacing location, clinical applications often face ambiguity in endocardial mapping. Further work is needed to address this.
  • Model Simplification: The Mitchell-Schaeffer model is a simplified ionic model. Future work could explore personalization of more complex, biophysically detailed models.
  • In Vivo Data: The study used ex-vivo data. Validation with in-vivo data would be a crucial next step.

Summary updated

BibTeX
@inproceedings{relan2010estimation,
  author = {Jatin Relan and Mihaela Pop and Hervé Delingette and Graham A. Wright and Nicholas Ayache and Maxime Sermesant},
  title = {{Estimation of Reaction, Diffusion and Restitution Parameters for a 3D Myocardial Model Using Optical Mapping and MRI}},
  booktitle = {STACOM/CESC},
  year = {2010},
  doi = {10.1007/978-3-642-15835-3\_28},
}
Figure