Image Processing Reference
In-Depth Information
The first corresponds to tagging information: the isoparametric curves of the
model are deformed to align with the tag strips. Simultaneously, the B-solid is
attracted toward the LV boundaries by integrating a distance function to edge
points on the epicardial and endocardial surfaces. Therefore, in this method,
boundary and tag information are incorporated in a unified approach. Because this
method has been applied in combination with short-axis tagged images only, it
yields in-plane 2-D displacements. Later, Huang et al. [181] extended the method
to analyze true 3-D deformations using a spatiotemporal model. The method differs
from the one of Radeva et al. in that no boundary information is now incorporated.
On the other hand, a spatiotemporal B-solid is constructed through a 4-D tensor
product spline (three dimensions and time). The fitting process to SPAMM data
is governed by a normal constraint that ensures that attraction produced by each
tag plane is in its normal direction. Because multiple orthogonal tag planes are
available, this allows a full 3-D reconstruction of the deformation field. A related
work is that of Amini et al. [174], in which the same parametric B-splines are
used to reconstruct the tag planes and track myocardial beads in three dimensions.
More recently, Chen and Amini [182] have proposed a maximum
a posteriori
framework for tag line detection using oriented filters, which is used to recover
the parameters of 3-D and 3-D
t deformable solids.
Kerwin and Prince [83] have developed an alternative projection technique
to accurately estimate the 3-D, location of the intersection points of the tag grid.
The deformation field between two frames is recovered using thin-plate spline
interpolation. Myocardial points are distinguished from those in static tissues by
checking whether they pass across the imaging plane over time. For points that
do not fulfill the preceding criterion, a test is performed to check their inclusion
within the outlined myocardial borders prior to rejection from the analysis. Such
a rejection scheme is important for proper visualization and analysis of myocar-
dial motion.
Young [180] introduced the concept of
model tags
, which represent the
material surfaces within the heart tissue that are tagged with magnetic saturation.
Model tags are “attached” to the heart and deform with it. They are embedded
within a 3-D FE model describing the geometry of the LV; this model is linear
in the transmural direction and employs bicubic Hermite interpolation in the
circumferential and longitudinal directions. Instead of finding the 3-D location
of the tag plane intersections, this approach finds the intersections of the model
tags with the imaging planes (model tag intersections or MTIs). The FE model
is subsequently deformed so that the MTIs match the tag stripes in each image
plane. Matching is carried out by a local search algorithm guided by an orientation
filter. Additional mechanisms are incorporated to allow efficient user interaction
and to correct for erroneous MTI matches.
Wierzbicki et al. [155] evaluated the application of nonrigid registration for
tracking cardiac structures over the cardiac cycle. A manual segmentation of the
epicardium, LV and RV, and atria in the end-diastotic (ED) frame is nonrigidly
registered to the remainder of the sequence, and several registration similarity mea-
sures are evaluated in animal experiments and healthy volunteers. The mean-square
+
Search WWH ::
Custom Search