Biomedical Engineering Reference
In-Depth Information
Figure 3.
Standard and complex pipelines for deformable models.
the final model. Hence, we can reformulate (7) as:
g,
n
)
n
V
0
)
n
∂
∂t
=
V
0
+
I
∇
+
I
·
(
ακ
−
.
Convergence
Regularizing term
As we have pointed out, we can observe in the former equation that
V
0
ensures
convergence, but it also competes with the curvature for regularity near the region
of interest.
4.
DEFORMABLE MODELS IN THE CLASSIFICATION PIPELINE
Deformable models constitute an important tool for medical image analysis.
They are usually involved in segmentation and/or tracking processes. In this sec-
tion we analyze their application in the segmentation/classification pipeline and
provide alternatives to the standard methodology, which can be exploited by our
new formulation of the
STOP and GO
model.
Generally, any segmentation process involving deformable models is com-
posed of three stages. First, the graylevel image is described in terms of features
in the feature extraction process. Second, a machine learning technique is trained
and used to obtain a meaningful set of regions according to our segmentation goal.
