Biomedical Engineering Reference
In-Depth Information
Figure 12.
Design of the speed function for tracking: value and signal. See attached CD
for color version.
5.2. Results and Discussion: Segmentation and Tracking of CVH
Data
For the tracking task, experiments are carried out on the volumetric CVH
data. Referring to the speed function (Eqs. (40-42)), in addition to the color
information, which is introduced by the distance of (
I
n
−
I
n
+1
)
2
in Eq. (41), we
still incorporate the BPV (Eq. (38)) texture feature as the image content item
C
term in Eqs. (41) and (42), because of its effectiveness as a texture measure and
the convenience of calculation.
Satisfactory results have been obtained (Figure 11). The boundaries of the
AOI are well tracked and delineated using the proposed speed function model.
We tested it in a topology-coherent case (Figure 11) and a topology-changing case
(Figure 13). Figure 13a shows the first slice of a section of CVHdata containing the
brain. The two white dots in Figure 13b depict initialization in the AOI. Figure 13c
gives the boundaries in the subsequent image slices that are automatically tracked.
We can notice the topology changing by comparing the top left and bottom right
images in Figure 13c.
We should note that this method benefitted from the close inter-layer distance
within the CVH data. If the object displacements between slices are so large that
there are no overlapping regions, the model would fail.
6.
CONCLUSION
To summarize, a variational framework for the speed function of level set
methods has been proposed for segmentation and tracking of medical images. For
segmenting the tagged MR images with blurry boundaries and strong tag lines,
