Image Processing Reference
In-Depth Information
the number of averages, the acquisition time T
acq
becomes
TNNNN R
acq
=
(11.5)
x
y
z
acq
Due to signal-to-noise and quantification requirements of the measured spec-
tra, repetition time TR of the sequence has to be long enough (typically, TR
=
1500 msec). For eight phase-encoding steps along each direction (which is the
minimum used number, as discussed later), TR
=
1500 msec, N
acq
=
1, and the
12,8 min. Because in clinical measurements, more encod-
ing steps are used to achieve better resolution, T
acq
becomes too long. Therefore,
CSI sequences are mostly used in 2-D mode.
Various fast CSI sequences, suitable for 3-D CSI, have been adopted (19),
such as sequences using multiple echoes for phase encoding (20,21), sequences
using time-varying gradients during the readout period (22-27), sequences
derived from the steady-state MR imaging sequences (28,29), or the recently
implemented parallel spectroscopic imaging techniques (30,31).
acquisition time T
acq
=
11.3.2
A
VOIDING
U
NDESIRED
E
XCITATIONS
In many cases, examined tissue contains areas with spurious signals, such as areas
with poor magnetic field homogeneity, bones, air-containing structures (sinuses),
or fatty tissue. These signals are potential sources of the contamination of spectra
in other regions. A typical example is
1
H CSI of the brain, in which strong lipid
signals from extracranial subcutaneous fat can contaminate spectra within the brain.
Therefore, methods eliminating signals from problematic regions are desirable. Two
most-often-used methods are volume preselection and outer-volume suppression.
The idea of volume preselected CSI sequences is to incorporate a volume
selection used in PRESS or STEAM SVS into a CSI sequence (32,33). In this
case, only the desired part of the sample, the VOI is excited using the PRESS or
STEAM sequence, while the position of the CSI voxels is coded by phase
encoding (see
Figure 11.14a
, in which a 2-D PRESS CSI sequence is depicted).
Apart from elimination of spurious signals, this approach has another advantage.
As is known from MR imaging, to prevent aliasing artifacts, all regions of the
sample contributing to the measured signal have to be inside the FOV (because
only phase encoding is performed, frequency low-pass filters cannot be used).
This fact dictates the minimum FOV size and also the minimum voxel size
achievable per fixed time. Because by using volume preselected sequences only
a restricted area of the sample is excited, FOV can be reduced correspondingly,
resulting in smaller voxel size without the occurrence of aliasing artifact. How-
ever, due to imperfections of pulse profiles, areas outside the selected VOI are
also partially excited and contribute to the measured signal. Therefore, the FOV
should always extend beyond the VOI to encode positions of these signals (by
how much depends on used pulse profiles and the chemical shift artifact), and
Search WWH ::
Custom Search