Skip to main content
. 2013 Dec;17(8):1137–1150. doi: 10.1016/j.media.2013.07.004

Table 3.

Registration of 27 US images of younger dataset with MR template. Average volume overlaps of four manually segmented structures and their standard deviations after registration of the source image (third column) to the US images. We also present p-value showing statistical significance when compared to the proposed method. The best performance is highlighted in bold.

Registration Similarity Source CP VC CH CSP Average
No alignment 0.18 ± 0.14 0.10 ± 0.15 0.09 ± 0.13 0.09 ± 0.14 0.12 ± 0.14
Gradient descent NMI MRI 0.28 ± 0.19 0.18 ± 0.18 0.27 ± 0.19 0.22 ± 0.17 0.24 ± 0.18
p-value 1 × 10−8 2 × 10−9 2 × 10−7 3 × 10−6 6 × 10−12
Gradient descent NCC Pseudo US 0.45 ± 0.16 0.34 ± 0.19 0.31 ± 0.20 0.40 ± 0.17 0.37 ± 0.18
p-value 2 × 10−4 0.022 1 × 10−6 0.21 7 × 10−5
Gradient descent LNCC Pseudo US 0.42 ± 0.25 0.35 ± 0.26 0.45 ± 0.25 0.32 ± 0.18 0.38 ± 0.23
p-value 3 × 10−4 1 × 10−3 0.021 4 × 10−3 5 × 10−4
Block-matching CR MRI 0.54 ± 0.13 0.43 ± 0.17 0.46 ± 0.15 0.39 ± 0.13 0.46 ± 0.14
p-value 0.048 0.30 3 × 10−4 6 × 10−3 5 × 10−3
Block-matching CR Pseudo US 0.58 ± 0.11 0.42 ± 0.17 0.54 ± 0.16 0.39 ± 0.11 0.48 ± 0.14
p-value 0.92 0.043 0.31 0.010 0.13
Block-matching NCC Pseudo US 0.58 ± 0.10 0.46 ± 0.15 0.57 ± 0.09 0.43 ± 0.10 0.51 ± 0.11