Image Processing Reference
In-Depth Information
129,130,224]. Both factors influence the robustness and reproducibility of the
derived functional information.
Finally,
Table 9.2
also indicates the degree of user guidance (automation) of
the fitting procedures for any given input data (preprocessing) and set of
ad hoc
parameters. Three degrees of automation were used to classify the approaches:
relying on substantial human guidance (=), manual interaction can be necessary
for guiding or correcting the deformation (
), and fully automated (+). In general
terms, the larger the need for human intervention during the fitting procedure,
the less robust a technique, and the more prone it is to inter- and intraobserver
variability of the final results.
−
9.6
CONCLUSIONS AND SUGGESTIONS
FOR FUTURE RESEARCH
In this chapter we have reviewed the techniques for 3-D geometric modeling and
analysis of cardiac images. In particular, we have focused on those techniques
leading to traditional indices of cardiac function. We have proposed a systematic
classification of the approaches based on the type of representation of the geo-
metric model, and the type of input data required for model recovery (
Table 9.1
).
Furthermore, we have given a critical assessment of these approaches according
to the type of functional parameters that they provide, their degree of evaluation,
and the performance achieved in terms of modeling flexibility, complexity, and
effective automation (Table 9.2).
From the surveyed literature, four main lines of future efforts can be distin-
guished:
1.
Research on modeling and model deformation techniques: The last two
decades have witnessed an enormous amount of work on 3-D models of
the LV and RV. This holds true for all imaging modalities (cf. Table 9.2).
In spite of the large number of attempts, no approach has simulta-
neously achieved robustness, automation, model flexibility, and com-
putational speed. Manual outlining and analysis of cardiac images is
still the most popular technique in clinical environments.
Several issues will require more attention in order to integrate the
advances of modeling techniques into clinical practice. Accurate 3-D
modeling techniques are, in general, computationally intensive. Explora-
tion of flexible modeling techniques that make efficient use of their DOF
will be worthy of further research. So far the main flow of efforts has
been focused on adopting generic geometrical representations to build
cardiac shape models (e.g., superquadrics, B-splines, polyhedral meshes,
Fourier descriptors, etc.). As a consequence, in generating a realistic LV
shape, the representations are either too restrictive or require a consider-
able number of parameters. The question arises as to how to infer a
compact representation giving rise to realistic shapes, possibly learned
Search WWH ::
Custom Search