Biomedical Engineering Reference
In-Depth Information
Figure 2.10
(a--d) Tracing results using MW-CTTS compared to other algorithms. Tracing results
on a vascular image is shown for three different algorithms (scale bar = 20
m
m, mouse cortex with
vasculature stained with India ink and cell nuclei stained with Nissl, imaged using conventional
light microscope). All three algorithms were initiated with the same single seed point in (a). Our
MW-CTTS algorithm exhibits the most extensive trace.
(2) the fiber width, (3) the size of the Gaussian filter, and (4) total number of
moving windows. With the fiber width
n
, moving window side length of 2
n
e
(e is a small parameter between 1.1 and 1.5), a Gaussian filter size of 5
×
5,
k
and the total number of moving windows of
2
n
e
, the number of pixels come out
k
2
n
e
to
(
2
n
e
×
4
)
×
(
5
×
5
)
×
=
100
k
. That is, the complexity of the algorithm is
O
. This calculation shows that our algorithm can scale up quite well to large
volumes of data. It also helps that vascular and neural data are very sparse (
(
k
)
<
6%
for vascular and
1--2% for neural data).
In the next section, we will show how a similar approach can be extended into
<
3D.
Search WWH ::
Custom Search