Information Technology Reference
In-Depth Information
Fig. 8 The separation of the VB region. a 3D view of three adjacent VB. b Automated placement
of 4 seeds on cortical bone and disc. c Separation of VB shown with red color
Next, the histogram for a neighborhood around each seed is obtained. The
histogram represents the number of voxels whose intensity values are above
200 Houns
eld Unit (HU). This value is obtained empirically. Vertical boundaries
of a VB show higher gray level intensity than inner region of the VB and disks.
Figure
9
shows histograms (the red line), and thresholds (the black line). To search
vertical limits of the VB, the following adaptive threshold equation is used as
follows:
TH
¼
lð
A
Þþj
½
max
ð
A
Þlð
A
Þ;
ð
6
Þ
where
j
¼
3 which is derived from experiments by trial-and-error, where A
represents each histogram vector with the red line as shown in Fig.
9
, max
0
:
ð
A
Þ
and
lð
are the maximum and average values in the histogram vector.
In the separation step, 30 patients which totals to 117 VBs are used. The results
can be classi
A
Þ
ed as in [
17
]. There are
five respective categories as described below.
Excellent: The VB is successfully separated without any misclassi
cation.
Vertical limits are correctly obtained.
Good: The VB is separated with small parts of adjacent disk or VB. Around
90 % or vertical limits are correctly obtained.
Bad: The VB is separated, however noticeable parts of it are missed. Around
70
90 % of vertical limits are correctly obtained.
-
Poor: Large portions of anatomical structure of VB are missed. Around 50
-
70 %
of vertical limits are correctly obtained.
Fail: The VB is not separated due to challenges.
