Biomedical Engineering Reference
In-Depth Information
MR image distortions can originate from limitations in the scanner or prop-
erties of the imaged object.
14-16
The factors that are scanner-dependent, and
therefore of greatest interest here, are
1.
Gradient field nonlinearity;
2.
Static field inhomogeneity;
3.
Error in the field-of-view or slice thickness due to variations in the
gradient field strengths;
4.
Fields due to Eddy currents induced in scanner components by the
switching gradients;
5.
Imperfect slice or volume selection pulses.
In most standard imaging applications, the two most important sources of
geometric distortion are gradient nonlinearity and field inhomogeneity,
14
12,17
although Schad et al.
seem to attribute a large part of the distortions to the
effect of eddy currents. Gradient eddy currents have been much reduced in
modern scanners by the widespread use of self-shielded gradient coils.
Although signal nonuniformity due to RF inhomogeneity is not a geomet-
ric distortion, it can impede image registration in the most extreme cases
(e.g., images acquired using surface coils, see Studholme et al.
18
) and there-
fore merits discussion.
5.3.1
Magnitude of Geometric Distortions in MRI
The amount of geometric distortion can be measured using phantoms and
or
6,8,12,14,19,20
The amounts of distor-
tion observed vary widely and reflect the design and performance of different
scanners (different manufacturers, levels of technological development,
design, etc.), the slice orientation and imaging sequence used, and the design
of the stereotactic frames. The magnitude of the distortion generally varies as
a function of position, being larger in the peripheral region. Bakker et al. in a
phantom study found localization errors of up to 7 mm and 4 mm in plane over
a 40 cm field of view due to static field inhomogeneity and gradient nonlinear-
ity, respectively.
phantoms combined with stereotactic frames.
14
Michiels et al. concluded that the degree of localization accu-
racy achieved after optimization of the image acquisition and application of
correction for both field inhomogenity and gradient nonlinearity is adequate
for stereotactic neurosurgery.
21
However, Alexander et al. found discrepancies
in stereotactic localization of the order of 4 mm between MRI and CT.
22
They did
not attempt to correct the geometric distortions in MRI themselves, but rather
proposed image fusion with CT as a means of assuring spatial accuracy while
benefiting from the superior image contrast of MRI. In a frame-based localiza-
tion study, Walton et al. found that 3D (volumetric) acquisitions give more accu-
rate stereotactic localization.
23
Ramsey and Oliver, in a recent study on modern
CT and MRI scanners, have found linear distortions in the range of 0 to 2 mm in
