Image Processing Reference
In-Depth Information
Fig. 1.
Infant MRI: (from left to right) scans at 6, 12, and 24 months of age. The
modalities of scans are T1W (top row) and T2W (bottom row).
myelination involving processes including tightening of spiral myelin around the
axon [
1
]. This knowledge that each modality captures a distinct maturational
phase highlights the need for joint multimodal analysis.
SI values in MR images also depend on the pulse sequences used for imaging,
along with factors such as scanner type, scanning conditions, and strength of the
main magnetic field. The framework presented here is generic since the method
for joint multimodal analysis of appearance that the authors propose is not
limited to the pulse sequences and modalities discussed in this paper.
This paper applies multivariate nonlinear mixed effects modeling (NLME) to
the problem of joint modeling of white-gray matter contrast in multimodal scan
sets. The NLME modeling uses a nonlinear growth function to estimate popula-
tion and subject-specific trajectories of contrast change [
9
]. The growth function
parameters are jointly estimated for all modalities by taking into account corre-
lations between them, and provide timing and progression information related to
contrast change in different modalities [
6
,
10
], and have future potential as clinical
descriptors of population trends. The study of appearance change in multimodal
image sets also requires quantification of direction of relative white-gray matter
contrast. This is especially true for T1W and T2W images since the signal inten-
sity gradient of white matter with respect to gray matter in these modalities is
inverted. We characterize the direction of relative white-gray intensity gradient
using SIR (Signal Intensity Ratio). The gradient direction at each time point is
compared with the gradient direction in the adult-like image of the same modal-
ity to capture contrast reversal using a measure named CONTDIR (Contrast
Direction). The CONTDIR measure is used to assign a directional attribute or
sign to the contrast measure. Overall, this paper proposes a method to longitudi-
nally model white-gray matter contrast change jointly across multimodal image
sets, in terms of both magnitude and direction. The multivariate NLME proce-
dure and the resulting parametrized growth functions enable characterization of
Search WWH ::
Custom Search