Biomedical Engineering Reference
In-Depth Information
Table 13.4:
Number of optimization steps used by each of the three
binning techniques for second level.
Binarization
Linear binning
Nonlinear binning
Data Set
Mean
Mean
Mean
δ
δ
δ
Patient 1
127
24.31
123
18.81
113
17.16
Patient 2
171
37.50
129
25.3
120
27.76
Patient 3
218
43.81
212
36.75
188
42.06
Patient 4
185
86.14
160
50.48
149
56.73
Patient 5
197
46.25
187
70.06
157
50.12
Patient 6
274
110.03
198
55.19
149
41.49
Patient 7
260
74.11
163
34.40
151
39.41
When it comes to the second level (Table 13.4), the binarization approach
turns out to be the worst method for all seven patients. Not only the number
of optimization steps, but also the standard deviation for each patient data sets
is inferior. The nonlinear binning based method performs the best, on average,
while linear binning is in the middle.
The binarization approach can accelerate the registration in the coarser level,
where the image volumes are greatly decreased and so the time to do the entire
registration is reduced, as seen in the results section. When applied to the original
images, binarization requires more steps to be taken to make up for the over-
simplified bin structure in the previous level. This results in a longer registration
time than the other two techniques.
The nonlinear binning method is slightly better than the linear method in
the first level, but it dominates in the second level. Because the nonlinear bin-
ning method takes anatomical information into consideration, and gives more
details than the linear binning, its superiority over the other two is no surprise.
Running times given in Section 13.2.2.2 also proves this. The nonlinear binning
technique can accelerate the convergence of the registration thereby reducing
total registration time, compared to the binarization approach or linear binning,
without loss of accuracy.
13.2.3
Summary
In practice, the multiresolution approach has proven to be helpful for mul-
timodality image registration. In our approach, we implemented a two-stage
