344
M. Abderrahim et al.
Quantitative Evaluation. The average NCCC and NMI values resulting from
the analysis of different mono- and multi-modal registration methods of MRI
brain images from the CAHM dataset are summarized in Table 1, whilst Table 2
illustrates MRI(T1)/PET registration results using the RIRE dataset. It is clear
that both NMI and NCCC values given by the HM for the mono- and multimodal registrations are better than those given by the other three widely used
tools. The superiority of HM is confirmed by the boxplots of the four compared
methods for registering MRI/CT and MRI(PD)/PET scans (Fig. 6). It should be
pointed out that the nature of the images to be aligned can be very diverse and
it affects considerably the choice of the registration method to be adopted. The
hybrid method allows to align effectively images from the same modality as well
as from different modalities. The nature of the modalities considered, as well as
the type of the imaged organ, also influences the choice of the method. Likewise,
the dimensionality of the input images could also be taken into consideration.
Although no available registration method is perfect, research is being done to
improve the results, while reducing the rate of registration error. This could
increase the diagnosis confidence by improving the diagnosis accuracy.
Table 1. Average NMI and NCCC values resulting from the studied mono and multimodal registration methods using the CAHM dataset (best values are in bold).
NMI
NCCC
Image
pair
HM
SPM
ITK 3D Slicer
HM
SPM
ITK 3D Slicer
Monomodal
1
0.0344 0.0087 0.0089
0.0630
0.2666 0.2226 0.0050
0.2623
2
0.0085 0.0047 0.0032
0.0072
0.2010 0.1718 0.0215
0.1995
3
0.0561 0.0075 0.0064
0.0495
0.2609 0.0259 0.0054
0.2520
4
0.0887 0.0093 0.0081
0.0802
0.2381 0.0103 0.0078
0.2358
5
0.0930 0.0075 0.0063
0.0912
0.2567 0.0505 0.0126
0.2468
6
0.1333 0.0526 0.0526
0.1140
0.2518 0.0430 0.2318
0.2566
7
0.1102 0.0351 0.0281
0.1021
0.2135 0.1270 0.0512
0.2048
8
0.1011 0.0513 0.0426
0.0977
0.2017 0.0334 0.1882
0.1975
9
0.0284 0.0476 0.0440
0.0469
0.2543 0.1345 0.1567
0.2491
Multimodal
1
0.0766 0.0578 0.0598
0.0690
0.2408 0.2329 0.2395
0.2343
2
0.0331 0.0161 0.0165
0.0122
0.2682 0.1839 0.1858
0.1561
3
0.0972 0.0728 0.0741
0.0641
0.2486 0.2373 0.2301
0.2362
4
0.0603 0.0356 0.0344
0.0323
0.2580 0.2115 0.1884
0.1777
5
0.0823 0.0794 0.0774
0.0783
0.2351 0.2264 0.2287
0.2279
6
0.0963 0.0831 0.0813
0.0602
0.2483 0.2405 0.2462
0.2444
7
0.0719 0.0623 0.0619
0.0457
0.2556 0.1627 0.2310
0.2131
8
0.0845 0.0674 0.0678
0.0678
0.2381 0.1376 0.2115
0.2099
9
0.0643 0.0439 0.0511
0.0589
0.2467 0.1873 0.2098
0.2125
M. Abderrahim et al.
Quantitative Evaluation. The average NCCC and NMI values resulting from
the analysis of different mono- and multi-modal registration methods of MRI
brain images from the CAHM dataset are summarized in Table 1, whilst Table 2
illustrates MRI(T1)/PET registration results using the RIRE dataset. It is clear
that both NMI and NCCC values given by the HM for the mono- and multimodal registrations are better than those given by the other three widely used
tools. The superiority of HM is confirmed by the boxplots of the four compared
methods for registering MRI/CT and MRI(PD)/PET scans (Fig. 6). It should be
pointed out that the nature of the images to be aligned can be very diverse and
it affects considerably the choice of the registration method to be adopted. The
hybrid method allows to align effectively images from the same modality as well
as from different modalities. The nature of the modalities considered, as well as
the type of the imaged organ, also influences the choice of the method. Likewise,
the dimensionality of the input images could also be taken into consideration.
Although no available registration method is perfect, research is being done to
improve the results, while reducing the rate of registration error. This could
increase the diagnosis confidence by improving the diagnosis accuracy.
Table 1. Average NMI and NCCC values resulting from the studied mono and multimodal registration methods using the CAHM dataset (best values are in bold).
NMI
NCCC
Image
pair
HM
SPM
ITK 3D Slicer
HM
SPM
ITK 3D Slicer
Monomodal
1
0.0344 0.0087 0.0089
0.0630
0.2666 0.2226 0.0050
0.2623
2
0.0085 0.0047 0.0032
0.0072
0.2010 0.1718 0.0215
0.1995
3
0.0561 0.0075 0.0064
0.0495
0.2609 0.0259 0.0054
0.2520
4
0.0887 0.0093 0.0081
0.0802
0.2381 0.0103 0.0078
0.2358
5
0.0930 0.0075 0.0063
0.0912
0.2567 0.0505 0.0126
0.2468
6
0.1333 0.0526 0.0526
0.1140
0.2518 0.0430 0.2318
0.2566
7
0.1102 0.0351 0.0281
0.1021
0.2135 0.1270 0.0512
0.2048
8
0.1011 0.0513 0.0426
0.0977
0.2017 0.0334 0.1882
0.1975
9
0.0284 0.0476 0.0440
0.0469
0.2543 0.1345 0.1567
0.2491
Multimodal
1
0.0766 0.0578 0.0598
0.0690
0.2408 0.2329 0.2395
0.2343
2
0.0331 0.0161 0.0165
0.0122
0.2682 0.1839 0.1858
0.1561
3
0.0972 0.0728 0.0741
0.0641
0.2486 0.2373 0.2301
0.2362
4
0.0603 0.0356 0.0344
0.0323
0.2580 0.2115 0.1884
0.1777
5
0.0823 0.0794 0.0774
0.0783
0.2351 0.2264 0.2287
0.2279
6
0.0963 0.0831 0.0813
0.0602
0.2483 0.2405 0.2462
0.2444
7
0.0719 0.0623 0.0619
0.0457
0.2556 0.1627 0.2310
0.2131
8
0.0845 0.0674 0.0678
0.0678
0.2381 0.1376 0.2115
0.2099
9
0.0643 0.0439 0.0511
0.0589
0.2467 0.1873 0.2098
0.2125