This project aims to develop clinically validated multiscale models of cardiac dynamics that integrate fluid dynamics, electromechanical coupling, and fluid-structure interaction (FSI) to simulate intracardiac flows and blood coagulation in atrial fibrillation (AF). AF is the most common sustained arrhythmia in the U.S. and is associated with serious complications, including thromboembolism and stroke. Anticoagulation is commonly prescribed to patients who have an elevated stroke risk. However, current risk assessment indices, which lack individualization based upon atrial structure or function, classify most AF patients as being at intermediate risk. The overarching hypothesis motivating this research is that treatment guidelines using current risk assessment metrics result in many AF patients receiving unneeded anticoagulation and unnecessary monitoring for thrombosis. The long-term objective of this research program is to develop new, broad-spectrum approaches to clotting risk assessment in AF that provide personalized risk prediction. The premise of this specific project is that comprehensive models of atrial dysfunction will enable mechanistic studies of flow and clotting in AF that will ultimately facilitate individualized treatment.
The specific aims of this project involve developing dynamic computational models of blood flow and coagulation dynamics in the left atrium (LA) and left atrial appendage (LAA), and to use these models in the context of percutaneous LAA exclusion (e.g. via the WATCHMAN device), and LAA isolation in catheter ablation therapy. Clinical data will be obtained for both types of procedures within the project.
| Data for building the model | Published? | Private? | How is credibility checked? | Current Conformance Level / Target Conformance Level |
|---|---|---|---|---|
| in vitro (primary cells cell, lines, etc.) | yes | no | expert review | partial |
| ex vivo (excised tissues) | yes | no | expert review | partial |
| in vivo pre-clinical (lower-level organism or small animal) | ||||
| in vivo pre-clinical (large animal) | ||||
| Human subjects/clinical | yes | yes | expert review | partial |
| Other: ________________________ |
| Data for validating the model | Published? | Private? | How is credibility checked? | Current Conformance Level / Target Conformance Level |
|---|---|---|---|---|
| in vitro (primary cells cell, lines, etc.) | yes | yes | expert review | partial |
| ex vivo (excised tissues) | yes | no | expert review | partial |
| in vivo pre-clinical (lower-level organism or small animal) | ||||
| in vivo pre-clinical (large animal) | ||||
| Human subjects/clinical | yes | yes | expert review | partial |
| Other: ________________________ |
| Who does it? | When does it happen? | How is it done? | Current Conformance Level / Target Conformance Level | |
|---|---|---|---|---|
| Verification | project team | continuously | automated test suite | adequate |
| Validation | project team | as new data sets are acquired | manual comparisons | partial |
| Uncertainty quantification | future work | N/A | N/A | insufficient |
| Sensitivity analysis | future work | N/A | N/A | insufficient |
| Other:__________ | ||||
| Additional Comments |
| Disclaimer statement (explain key limitations) | Who needs to know about this disclaimer? | How is this disclaimer shared with that audience? | Current Conformance Level / Target Conformance Level |
|---|---|---|---|
| Uncertainty in anatomical geometry (assumed atrial wall thickness) | All potential users of the model and model results | Disclaimer will be clearly stated in model dissemination | partial |
| Rule-based fiber architectures (not subject specific) | All potential users of the model and model results | Disclaimer will be clearly stated in model dissemination | partial |
| Literature-based mechanical properties (not subject-specific) | All potential users of the model and model results | Disclaimer will be clearly stated in model dissemination | partial |
| Newtonian blood analogue | All potential users of the model and model results | Disclaimer will be clearly stated in model dissemination | partial |
| Current Conformance Level / Target Conformance Level |
|---|
| Extensive |
| Naming Conventions? | Repository? | Code Review? | |
|---|---|---|---|
| individual modeler | yes | yes | yes |
| within the lab | yes | yes | yes |
| collaborators | yes | yes | yes |
| Current Conformance Level / Target Conformance Level | |
|---|---|
| Code commented? | extensive |
| Scope and intended use described? | adequate |
| User’s guide? | partial |
| Developer’s guide? | partial |
| Current Conformance Level / Target Conformance Level |
|---|
| adequate |
| Target Audience(s): | “Inner circle” | Scientific community | Public |
|---|---|---|---|
| Simulations | institutional file systems | in progress | in progress |
| Models | https://github.com/IBAMR (private repository) | in progress | in progress |
| Software | https://github.com/IBAMR | https://github.com/IBAMR | https://github.com/IBAMR |
| Results | in progress | in progress | in progress |
| Implications of results | in progress | in progress | in progress |
| Current Conformance Level / Target Conformance Level |
|---|
| partial |
| Reviewer(s) name & affiliation: | TBD - U.S. Food and Drug Administration |
|---|---|
| When was review performed? | planned for mid 2020 |
| How was review performed and outcomes of the review? | Medical Device Development Tool program review (for prototype IVC filter model) |
| Current Conformance Level / Target Conformance Level |
|---|
| insufficient |
| Yes or No (briefly summarize) | |
|---|---|
| Were competing implementations tested? | no -- unaware of alternative implementations that provide needed fluid-structure interaction capabilities |
| Did this lead to model refinement or improvement? | N/A |
| Current Conformance Level / Target Conformance Level |
|---|
| insufficient |
| Yes or No (briefly summarize) | |
|---|---|
| Are there operating procedures, guidelines, or standards for this type of multiscale modeling? | standards for organ-scale multiscale models do not appear to exist at present |
| How do your modeling efforts conform? | N/A |
Simulation methods
- Computational fluid dynamics (CFD): finite volume-type methods for the incompressible Navier-Stokes equations with block-structured adaptive mesh refinement (AMR)
- Computational solid dynamics (CSD): stabilized mixed nodal finite element methods for incompressible and nearly incompressible nonlinear elasticity
- Computational fluid-structure interaction (FSI): immersed boundary (IB) type methods
- Computational electrophysiology (EP): nodal finite element methods
- Computational thrombogenesis: finite volume-type methods for reaction-advection-diffusion equations with structured adaptive mesh refinement (SAMR)
Software
The IBAMR software (ibamr.github.io) will serve as the primary software framework for CFD, FSI, and continuum thrombosis modeling, and for developing integrative models. IBAMR is a C++ infrastructure for constructing FSI models using the immersed boundary (IB) method and related methods. The majority of its development occurs within the open IBAMR GitHub project pages (github.com/IBAMR).
The BeatIt software (github.com/rossisimone/beatit) will serve as the primary software framework for CSD and EP modeling. Although IBAMR provides a nonlinear solid mechanics module, we are currently developing FSI coupling algorithms that will allow us to use BeatIt to handle the solid mechanics, electrophysiology, and electro-mechanical coupling.
IBAMR and BeatIt both rely on the libMesh finite element library (libmesh.github.io) to treat complex geometries, and PETSc (www.mcs.anl.gov/petsc) for core computational infrastructure (distributed vectors and matrices, and linear and nonlinear solvers and preconditioners).
Verification and validation
Software and method verification will primarily be carried out via benchmark problems with known analytic solutions (e.g. using the method of manufactured solutions) , or via benchmark problems with consensus solution values. An automated verification test suite, based on the test suite for the deal.II finite element library (dealii.org), already exists for the IBAMR software. Similar tests will also be developed for the BeatIt software. We shall improve the methods used for reporting the results from these tests (e.g. using a “dashboard” on GitHub).
Validation tests will include comparisons to data from in vitro models of the fluid dynamics of the heart and great vessels (e.g. obtained from a commercial pulse duplicator), including quantitative flow mappings obtained using particle image velocimetry, and by comparisons to in vivo data on flow patterns, local tissue stiffnesses, and electrical activation in the LA obtained intraoperatively during catheter ablation along with pre-/post-operative patient imaging data.
For CFD, we shall rely on benchmark datasets collected within the FDA’s Critical Path program on validating cardiovascular fluid dynamics (pubmed/21428676, pubmed/25180887, pubmed/28114192), which include laboratory studies of flow-induced hemolysis. Additionally, we shall carry out qualitative comparisons to previous clinical measurements of LA blood flow (e.g., pubmed/11559688).
For cardiac mechanics, we shall use the cardiac mechanics benchmark (pubmed/26807042) developed by Land et al. as an initial test of the mechanics solvers, although these tests focus on the response of the left ventricle. Because strain imaging is used for the direct evaluation of LA function, validation tests on the LA mechanics will compare strains and strain rates during sinus rhythm (pubmed/22909795). Validation of the electromechanical model will compare the evolution of strain and strain rates data during AF, as reported previously (pubmed/20133512, pubmed/22265458). Within the project, we also shall collect acoustic radiation force impulse (ARFI) data that will characterize in vivo tissue mechanics, and these data will be used in the project for additional validation studies.
For FSI, we aim to follow a previously described cardiac FSI benchmark (pubmed/27813272) developed by Nordsletten and co-workers. We also shall collect data during the duration of the project within ongoing projects that will provide additional validation datasets.
For cardiac electrophysiology, we shall apply the N-version benchmark (pubmed/21969679) described by Niederer et al. This involves using the converged consensus result from multiple simulation codes as a gold standard. An online database is available to facilitate the verification of cardiac tissue electrophysiology simulation codes such as the one we are developing. In addition, we shall collect multisite recordings on the posterior wall of the LA from multiple patents during clinical procedures, which will provide additional validation data sets with regard to conduction velocities and electrogram amplitudes and morphologies in diseased tissue.
For thrombosis, we shall compare simulation model outputs to in vitro microfluidic platelet deposition and fibrin formation assays (pubmed/18983510, pubmed/18203955, pubmed/23001359, pubmed/24236042, pubmed/28529666, pubmed/29472230), and to in vitro platelet deposition in stenotic arteries measurements (doi/10.1007/s13239-014-0180-z, doi/10.1007/s13239-012-0086-6).
Distributing verification and validation model software
Dedicated web pages will be developed to document key verification and validation tests and to summarize results from those tests along with the software implementation of these tests. Where feasible, model geometries will be provided for validation models. Although we shall endeavor to use in vitro test geometries that can be readily distributed, redistribution of validation model geometries may be restricted if required by the intellectual property or human subjects research offices of the participating institutions. To improve reproducibility, we shall document specific software versions used to generate the model results.