Biomedical Engineering Reference
In-Depth Information
(a)
(b)
(c)
(d)
Figure 12.
Bland-Altman plots for the best circumferential strain predictors of each model
type. Plotted are the difference between Arts' model and the best performer from each model
type of the 224 total models. Also included are the mean of the difference (solid line) and
the 95% limits of agreement (dashed lines). (a) Cartesian-based cylindrical-parameterized
NURBS model (order = 3, number of control points in
(
u
LV
,v
LV
,w
LV
)=(3
,
4
,
4)
).
(b) Cartesian-based prolate spheroidal-parameterized NURBS model (order = 3, number
of control points in
(
u
LV
,v
LV
,w
LV
)=(3
,
4
,
4)
). (c) Cylindrical-based cylindrical-
parameterized NURBS model (order = 3, number of control points in
(
u
LV
,v
LV
,w
LV
)=
(3
,
3
,
3)
). (d) Prolate spheroidal-based prolate spheroidal-parameterized NURBS model
(order = 3, number of control points in
(
u
LV
,v
LV
,w
LV
)=(3
,
5
,
5)
).
See attached CD
for color version.
fitting results, which led to exclusion of the lower 50%). This is illustrated by
plotting the temporal mean and standard deviation of these models for the radial
(Figures 13 and 14), circumferential (Figures 15 and 16), and longitudinal (Figures
17 and 18) strains along with the ground truth predicted by Arts' model.
These plots show that, for this dataset, the NURBS models generally overpre-
dict the subepicardial radial strain values, underestimate the subendocardial radial
strain values, accurately calculate the circumferential strains, and vary for the
longitudinal strain values. As discussed earlier, we can partially attribute this phe-
