Information Technology Reference
In-Depth Information
Fig. 45 Segmentation results with various shape initialization. (i) the initial shape prior, and (ii)is
the final results. The red and yellow colors show the contour of the gold standards and segmented
regions
100
(a)
100
(b)
95
90
95
85
90
80
75
85
70
80
65
60
PCA based Seg, as in [9]
2D-PCA based Seg, Ours
2D-PCA based Seg
PCA based Seg [9]
75
55
50
70
7
14
21
28
35
42
49
56
0
10
20
30
40
50
60
70
80
90
SNR (dB)
# of training shapes
Fig. 46 a The average segmentation accuracy of different segmentation methods on 500 CT
images under various signal-to-noise ratios. b The effect of choosing the number of the projected
training shapes N on the segmentation accuracy
3 Discussion and Conclusion
In this chapter, frameworks which are robust under segmentation challenges,
appropriate for a clinical work
ow, and have theoretical novelty are proposed. This
work is validated with various noise levels and compared with several alternative
methods. To transfer the developed software into the clinical usage, more experi-
ments on increased number of data sets are necessary.
One of the most important contributions of this study is to offer a segmentation
framework which can be suitable to the clinical works with acceptable results. The
proposed method completes the VB segmentation in very low execution time. It
fl
