Biomedical Engineering Reference
In-Depth Information
However, it is not uncommon to use serially interleaved methods, in which,
for example, every second slice is obtained in one complete acquisition, fol-
lowed by a separate, repeat run to acquire the omitted slices. With this latter
approach, subject motion can result in spatially inconsistent data, so that the
two scans cannot be treated as one, but neither component scan has full
enough spatial coverage to be reformatted without error.
Susceptibility artifacts may be very largely replicated if the patient is in vir-
tually the same position on both examinations. However they are inherently
anisotropic with respect to B
0
and may not therefore be correctable by rigid
body translation and rotation. The problem is likely to be greater with T2-
weighted sequences than the T1-weighted sequences and the T2-weighted
spin echo sequences used in this study.
The difference images provide an inbuilt check of the fidelity of registra-
tion. If the program fails to obtain a satisfactory match, monophasic or mul-
tiphasic changes will be seen in border zones. Interplateau shifts may also be
seen. The changes may be widespread and generally will have the character
of a whole brain shift superimposed on any underlying changes. Unregis-
tered images provide a useful guide in this context.
Difference images tend to reveal artifacts more strongly than the original
anatomical images. This is because subtraction of registered images cancels
out the signal from unchanged anatomy, whereas the artifacts present in the
source images are likely to be different for each acquisition and so do not
cancel. A notable feature of many difference images is biphasic changes along
strong edges that run perpendicular to the primary or secondary phase
encode directions. These are often caused by patient motion at a level that is
insufficient to produce overt artifacts on the source images, but which has the
effect of slightly reducing the image resolution by making the extreme edges
of the raw k-space data inconsistent.
10
Changes in scanner calibration can also often readily be detected on differ-
ence images. For example, changes in gradient scaling, which alter the appar-
ent size of imaged objects, are manifest as progressively increasing difference
signals from edges perpendicular to the scaled axis. Methods for correcting
calibration errors are discussed in Chapter 5.
7.7
Approach to Diagnosis of Changes to the Brain
on Difference Images
It is necessary to relate the causes of changes on difference images (i.e., pure
changes in signal intensity, throughplane and inplane border zone shifts, and
interplateau shifts) to the appearances they produce, i.e., (a) nonspecific
changes in signal intensity; (b) monophasic curvilinear changes; (c) mul-
tiphasic curvilinear changes, and (d) interplateau changes. Changes of differ-
ent types may be present on the same image. Their relation to underlying
causes is summarized in Table 7.1. A formal approach to image interpretation
is described in Table 7.2.
