Page 13 - Artificial Intelligence for Computational Modeling of the Heart

xii List of figures

that is subject to shear, the second letter denotes the direction
of shear. 76
Fig. 2.26 A linear elastic beam of length l and height h is subject to a
suddenly applied shear stress S at one end while the other end is
fixed. 77
Fig. 2.27 Displacement of the point at the center of the loaded boundary
face, computed with different spatial resolutions and with the
analytical solution of the problem. 77
Fig. 2.28 Example of bi-ventricular electromechanics simulation, from
end-diastole to systole to relaxation. Color encodes the computed
electrical potentials. 78
Fig. 2.29 15-velocity lattice structure. 80
Fig. 2.30 Fluid structure interaction system for cardiac haemodynamics
computation. The interactions between the electromechanical
model, valves and the CFD model are controlled by the FSI
interface module. 83
Fig. 2.31 Aortic and mitral 3D valves are controlled by 0D opening phase
functions whose dynamics is governed by pressure gradient forces. 84
Fig. 2.32 Cross-sectional flow variation with the peristaltic amplitude. An
excellent match with theory is obtained. 87
Fig. 2.33 Time variation of the geometry of the expanding and contracting
vessel. 87
Fig. 2.34 Flow ratesvstime. 88
Fig. 2.35 Cardiac cycle (systole on top, diastole on bottom) computed using
the FSI framework introduced in this chapter. Velocity magnitude
in the left ventricle is visualized using a standard rainbow
colormap with constant positive slope transparency map.
Myocardial stress magnitude is also visualized with a black-body
radiation colormap. 88
Fig. 2.36 Different steps involved for the estimation of the Windkessel
parameters of the arteries. Illustration on a pulmonary artery and
right ventricular data. In the left panel, intra-cardiac pressure is
shown in blue (dark gray in print version); arterial pressure in red
(mid gray in print version) ; ventricular volume in green (light gray
in print version) . Pressure is in mmHg, volume in ml. 90
Fig. 2.37 Inverse problem framework for personalizing the biomechanical
model parameters from clinical data. 92
Fig. 3.1 Visualization of uniform feature patterns versus self-learned,
sparse, adaptive patterns. 100
Fig. 3.2 Schematic visualization of the marginal space deep learning
framework applied for the sake of example to object localization in
3D echocardiographic images. The same approach can be used to
parse images from different imaging modalities. 102
Fig. 3.3 Schematic visualization of the learning-based boundary
deformation step. 103
Fig. 3.4 From left to right: segmentation results for all four heart chambers
visualized in an orthogonal slice-view of a cardiac CT image
(computed using MSL); a detected bounding box around the aortic
heart valve in a 3D TEE Ultrasound volume (the detected box is