Artificial Intelligence for Computational Modeling of the Heart

Cover
Title
Copyright
Contents
List of figures
List of contributors
Foreword
Preface
	Computational modeling of the heart: from physiology understanding to patient-specific simulations
	The new age of artificial intelligence
	Computational modeling meets statistical learning
	Book organization
		Part 1: Modeling the beating heart: approaches and implementation
		Part 2: Artificial intelligence methods for cardiac modeling
	Outlook
List of abbreviations
1
	1 Multi-scale models of the heart for patient-specific simulations
		1.1 Models of cardiac anatomy
		1.2 Electrophysiology modeling
			1.2.1 Cellular electrophysiology
				An example: the Mitchell-Schaeffer model
			1.2.2 Tissue electrophysiology
				One example: a monodomain model
				Modeling the electrical conduction system
			1.2.3 Body surface potential modeling
				Bidomain modeling of the coupled heart-body system
		1.3 Biomechanics modeling
			1.3.1 The passive myocardium
				An overview of the Holzapfel-Ogden model
			1.3.2 The active myocardium
				Ionic models
				Phenomenological models
				Lumped models
			1.3.3 The virtual heart in its environment: boundary conditions
				Endocardial pressure
				Attachment to neighboring vessels and tissue
		1.4 Hemodynamics modeling
			1.4.1 Reduced order hemodynamics
				Ventricular model
				Valve model
				Arterial model
				Atrium model
				Venous circulation
			1.4.2 3D hemodynamics
				1.4.2.1 Modeling intra-cardiac blood flow
				1.4.2.2 Fluid-structure interaction
					Valve models and FSI
		1.5 Current approaches to parameter estimation
			1.5.1 Inverse optimization
			1.5.2 Data assimilation
			1.5.3 Machine learning
			1.5.4 Stochastic methods
			1.5.5 Streamlined whole-heart personalization
		1.6 Summary
2
	2 Implementation of a patient-specific cardiac model
		2.1 Anatomical modeling
			2.1.1 Medical image segmentation
			2.1.2 Meshing and tagging
			2.1.3 Computational model of the cardiac fiber architecture
			2.1.4 Torso modeling
		2.2 Electrophysiology modeling
			2.2.1 LBM-EP: efficient solver for the monodomain problem
				LBM-EP evaluation
			2.2.2 Efficient modeling of the electrical conduction system
			2.2.3 Graph-EP: fast computation of tissue activation time
			2.2.4 Body surface potential modeling
				Extracellular potentials computation
				Boundary element model of torso potentials
					2.2.4.1 ECG calculation
		2.3 Biomechanics modeling
			2.3.1 Passive stress component
			2.3.2 Active stress component
			2.3.3 Myocardial boundary conditions
				Endocardial pressure
				Attachment to atria and arteries
				Modeling the effect of the pericardium bag
			2.3.4 Putting it all together: a fast computational framework for cardiac biomechanics
				Description of the TLED finite elements algorithm
			2.3.5 Evaluation of the TLED algorithm
				Validation against analytical solution
				Numerical stability analysis
				Bi-ventricular simulation
		2.4 Hemodynamics modeling
			2.4.1 3D hemodynamics using the lattice Boltzmann method
				The Lattice Boltzmann Method
				Turbulence modeling
			2.4.2 3D fluid structure interaction
				Preparatory step
				Step 1
				Step 2
				Step 3
				Step 4
				Tests of the FSI module
				Output of the FSI system with patient-specific data
		2.5 Parameter estimation
			2.5.1 Windkessel parameters from pressure and volume data
			2.5.2 Cardiac electrophysiology
			2.5.3 Myocardium stiffness and maximum active stress from images
		2.6 Summary
3
	3 Learning cardiac anatomy
		3.1 Introduction
		3.2 Parsing of cardiac and vascular structures
			3.2.1 From shallow to deep marginal space learning
				3.2.1.1 Problem formulation
				3.2.1.2 Traditional feature engineering
				3.2.1.3 Sparse adaptive deep neural networks
				3.2.1.4 Marginal space deep learning
				3.2.1.5 Nonrigid parametric deformation estimation
				3.2.1.6 Experiments
			3.2.2 Intelligent agent-driven image parsing
				3.2.2.1 Learning to search for anatomical objects
				3.2.2.2 Extending to multi-scale search
				3.2.2.3 Learning multi-scale navigation strategies
				3.2.2.4 Robust spatially-coherent landmark detection
				3.2.2.5 Experiments
			3.2.3 Deep image-to-image segmentation
		3.3 Structure tracking
		3.4 Summary
4
	4 Data-driven reduction of cardiac models
		4.1 Deep-learning model for real-time, non-invasive fractional flow reserve
			4.1.1 Introduction
			4.1.2 Methods
				4.1.2.1 Generating synthetic coronary arterial trees
				4.1.2.2 CFD-based hemodynamic computations
				4.1.2.3 Machine-learning based FFR computation
				4.1.2.4 Local features
				4.1.2.5 Features defined based on the proximal and distal vasculature
			4.1.3 Results
				4.1.3.1 Validation on synthetic anatomical models
				4.1.3.2 Validation on patient specific anatomical models
				4.1.3.3 Validation against invasively measured FFR
			4.1.4 Discussion
				4.1.4.1 Use of synthetic data
				4.1.4.2 Limitations
		4.2 Meta-modeling of atrial electrophysiology
			4.2.1 Methods
				4.2.1.1 Atrial electrophysiology models
				4.2.1.2 Learning the action potential manifold for dimensionality reduction
				4.2.1.3 Statistical learning
				4.2.1.4 Application to tissue-level cardiac EP modeling
			4.2.2 Experiments and results
				4.2.2.1 Model parameter selection and sampling
				4.2.2.2 PCA versus LLE
				4.2.2.3 Physical regression model construction
				4.2.2.4 Application in tissue-level EP modeling
			4.2.3 Discussion
		4.3 Deep learning acceleration of biomechanics
			4.3.1 Motivation
			4.3.2 Methods
			4.3.3 Evaluation
				4.3.3.1 Discussion
		4.4 Summary
5
	5 Machine learning methods for robust parameter estimation
		5.1 Introduction
		5.2 A regression approach to model parameter estimation
			5.2.1 Data-driven estimation of myocardial electrical diffusivity
			5.2.2 Experiments and results
				5.2.2.1 Setup and uncertainty analysis
				5.2.2.2 Verification on synthetic data
				5.2.2.3 Evaluation on patient data
		5.3 Reinforcement learning method for model parameter estimation
			5.3.1 Parameter estimation as a Markov decision process
				5.3.1.1 Reformulation of model personalization into an MDP
				5.3.1.2 Learning model behavior through exploration
				5.3.1.3 From computed objectives to representative MDP state
				5.3.1.4 Transition function as probabilistic model representation
			5.3.2 Parameter estimation using Reinforcement Learning
				5.3.2.1 Data-driven initialization
				5.3.2.2 Probabilistic personalization
			5.3.3 Application to cardiac electrophysiology
			5.3.4 Application to whole-body circulation
		5.4 Summary
6
	6 Additional clinical applications
		6.1 Cardiac resynchronization therapy
			6.1.1 Introduction
			6.1.2 Methods
				6.1.2.1 Data acquisition
				6.1.2.2 Computational modeling
			6.1.3 Results
				6.1.3.1 Electrophysiological results
			6.1.4 Discussion
		6.2 Aortic coarctation
			6.2.1 Introduction
			6.2.2 Methods
				6.2.2.1 Generation of a synthetic training database
				6.2.2.2 Three-dimensional flow computations
				6.2.2.3 Pressure drop model for aortic coarctation
			6.2.3 Results
				6.2.3.1 Evaluation of the pressure drop model
			6.2.4 Discussion
		6.3 Whole-body circulation
			6.3.1 Introduction
			6.3.2 Methods
				6.3.2.1 Traditional personalization framework
				6.3.2.2 Deep learning based personalization
			6.3.3 Results and discussion
		6.4 Summary
7
	Bibliography
Index
	Index
Back Cover