A Survey of Three-Dimensional Modeling Techniques for Quantitative Functional Analysis of Cardiac Images - Advance Image Processing in Magnetic Resonance Imaging - page 306

MR tagging techniques for reconstruction of myocardial motion or tissue
deformation deserve separate attention. Most in vivo animal and human studies
have reported on the Monte Carlo analysis of sensitivity to errors in tag local-
ization and tracking, and on the ability to recover the location of tags in different
frames [78,83,177,178,240]. * Several models have been used in the literature to
benchmark the accuracy of motion and deformation recovery. These evaluations
were based, for instance, on spherical and cylindrical models of cardiac motion
[69,78,82,177] FE solutions with realistic geometries [178], artificially generated
motion trajectories [122], or synthetic images using the cardiac motion simulator
[83,181,250] developed by Waks, Prince, and Douglas [236] that builds upon the
kinematic model of Arts et al. [251]. A study was carried out by Declerck et al.
[252] that thoroughly compared four techniques [82,177,253,254] for motion
tracking from tagged MR. This paper provides results on normal and pathological
subjects. Although the general trends of motion were captured correctly by all
methods, this study shows that there are noticeable differences in the displacement
and strain computations provided by each technique.
Finally, the third category includes studies that reported application on human
volunteers and patients, including studies that provided quantitative results in terms
of cardiac functional parameters. The size of the populations in most of these studies
was small: with only three exceptions, all studies were conducted on less than a
dozen volunteers or patients.

9.5.2 P ERFORMANCE C RITERIA

In the following subsections we elaborate on the criteria that we have used to
compare the different methods.

9.5.2.1 Model Complexity or Flexibility

The complexity or flexibility of a technique has been categorized in four groups
according to the number of DoF or parameters involved. ** The four categories
are: (1) Compact models with only a few parameters (on the order of a dozen)
of which prototypical examples are superquadrics, (2) flexible models with large
number of DOFs and parameterized with global-support basis functions, of which
representative examples are harmonic parameterizations of several types, (3)
flexible models with large number of DOFs and parameterized with local-support
basis functions, members of which family are B-spline and polyhedral models,

* Remarkably, a large number of evaluations involving canine models have been acquired with the
dynamic spatial reconstructor. However, the reduced clinical availability of this technique and its
specific image properties make it difficult to extrapolate the results of the evaluation to other clinical
imaging techniques.
** Validation MR tagging itself for describing tissue deformation has been addressed by Young et
al. [249] using a silicone gel phantom. Strains derived from MR tagging were compared to the
analytic equilibrium strains under a Mooney-Rivlin material law.