Page 11 - Artificial Intelligence for Computational Modeling of the Heart

x List of figures

Fig. 2.1 Elements, with input and output, of a typical computational model
of the heart. Green (mid gray in print version): input data. Red
(dark gray in print version): processing units. Orange (light gray in
print version): output data. Arrows denote data flow. 44
Fig. 2.2 Overview of the anatomical modeling pipeline based on medical
image segmentation. 45
Fig. 2.3 From left to right: final geometrical models extracted from
computed tomography (CT), magnetic resonance image (MRI) or
ultrasound data. 46
Fig. 2.4 Ventricular models (left images) and valvular models (right
images) are parameterized and tagged. 46
Fig. 2.5 Tagged surface meshes (left and middle panels) and fused
tetrahedral mesh (right panel). 47
Fig. 2.6 Automatic subdivision of the biventricular anatomical model
according the segment definition in [206]. A 17-segment model is
represented for two patient-specific geometries. 47
Fig. 2.7 Fibers and sheets orientation from endocardium to epicardium.
A local coordinate system based on the circumferential (e 0 ),
longitudinal (e 1 ) and radial (e 2 ) directions is used to define a
rule-based model of fiber orientations. 48
Fig. 2.8 Fiber estimation processing. Left panel: apex to base fiber
estimation using a rule-based model. Mid panel: prescription of
fiber orientation around the valve. Right panel: geodesic
interpolation of fibers from the base to the valves. 49
Fig. 2.9 Example of computed fibers and fiber sheets on a patient-specific
anatomy. 49
Fig. 2.10 Image of the torso avatar used for fitting the imaging data, with
the standard 12-lead ECG leads in place. 50
Fig. 2.11 (Left) Snapshot of the activation potential propagation through the
myocardial tissue and (right) the resulting map of the activation
times, computed by solving the monodomain model of tissue
electrophysiology with M-S cellular model and LBM
computational method. 52
Fig. 2.12 Description of the LBM-EP algorithm on a 2-D slice. The first
image shows the pre-collision distribution in a node at the start of
the step. The collision step redistributes the distribution function
values (middle figure) and finally the post-collision values stream
to the corresponding neighbors. 54
Fig. 2.13 Schematic of the problem geometry from [221]. A tissue sample of
size 20 mm× 7mm× 3 mm is stimulated in the cube marked S. The
activation times are reported at points P1 to P9, as well as on the
sliceshownin(B).(Source: [221].) 55
Fig. 2.14 Comparison of the activation time computed from LBM-EP (box L)
with those presented in [221]. Red (mid gray in print version),
green (light gray in print version) and blue (dark gray in print
version) lines represent solutions using a spatial resolution of
0.1 mm, 0.2 mm and 0.5 mm respectively. Codes A, B, C, E, F and H
are finite element codes. Codes D, G, I, J, and K use finite differences.56