| |
Results
163
The initial atlas (overlay from initial atlas in Figure 50, top left) does not match the reference
(“measured”) phase-wedge image closely. However, after the matching procedure, the final
result (Figure 50, top right) is very similar to the gold standard (Figure 50, bottom left). For
this case, the gold standard is known because the displacement fields from the manual defor-
mation procedure were used to create the gold standard visual-area ROIs. Since gold standard
visual-area ROIs are known, this is a good opportunity to calculate the quality measure:
Stage # Areas " All Visual Areas V1
V2
V3
Initial
0.93173
0.96671
0.94183
0.96551
Very-low Res.
0.94405
0.97116
0.95313
0.97302
Low Resolution
0.96142
0.98071
0.96761
0.98077
High Resolution
0.98492
0.99411
0.9893
0.99087
Table 1: Quality for artificially created reference image
The table shows that the quality increases during the matching process. If this were not the
case, the definition of the quality or the definition of the energy would have to be re-evaluated.
Another nice thing to realize is that the visual field sign map, calculated based on the
noisy phase-wedge and phase-ring images, looks as expected. This shows that the visual-field-
sign-map calculation works in principle.
The deformed noisy and rotated atlas images that were used in this section can only show
that basic principles of the algorithm work. Therefore, various measured data sets were also
tested to evaluate the performance of mrFindBorders. Two of those data sets are presented
next.
|  |
|
| |
|
|
