Parallel MRI: Concepts and Methods - Advance Image Processing in Magnetic Resonance Imaging - page 57

These advances have led to ultrafast multiecho pulse sequences that go by names
such as RARE, EPI, and BURST but have not been able to address the speed
requirements in certain applications. Dynamic imaging applications such as cardiac
and interventional imaging would still be greatly served if an order of magnitude
reduction in scan time were achieved without sacrificing spatial resolution and signal-
to-noise ratio (SNR). During the 1990s, the answer to the imaging speed needs was
starting to take shape in the parallel acquisition paradigm. The idea that multiple RF
receivers could be used in parallel to speed up the image acquisition, as is the case
in computer-aided tomography (CAT), was gaining momentum. Moreover, this new
field of parallel imaging can be combined with the previous ultrafast multiecho
methods to further increase imaging speed.
The theoretical feasibility of fast data acquisitions using multiple detectors
in MRI was first described by Hutchinson and Raff in 1988 (2), and subsequently
by Kwiat et al. in 1991 (3). Both groups investigated methods to solve the inverse
source problem on MR signals received in multiple RF receiver coils, requiring
the use of a number of closely packed RF coils, equal to the number of pixels in
the image, as well as greatly increased receiver coil sensitivities, in order to
eliminate the requirement of phase encoding by gradient switching. These require-
ments are quite impractical in conventional MR imaging and have prevented these
techniques from being applied in practice.
In 1993, Carlson and Minemura (4) described the use of a two-coil array, using
one coil with homogeneous sensitivity over the field of view (FOV) and the other
having a linear gradient in sensitivity. Partial data sets were acquired in each coil,
and the missing lines in k-space were generated using a series expansion in terms
of other phase-encoded lines. This approach yielded twofold image acceleration and
was the first technique that produced accelerated images using coil-sensitivity infor-
mation; however, it remained impractical due to design conditions required of the
sensitivity of the coils.
Later that same year, Ra and Rim (6) introduced the first feasible parallel imaging
method, which used coil sensitivity as a way to remove the aliasing in regularly
undersampled images acquired with multiple coils. Although this technique later
constituted the basis for the sensitivity encoding (SENSE) method, which is currently
enjoying wide commercial use, Ra and Rim's original paper presented only phantom
reconstructions. However, during the time when the technique came out, more
research emphasis was being given to sequential fast imaging methods, and little
effort was made to develop it.
In 1997, Sodickson et al. introduced the simultaneous acquisition of spatial
harmonics (SMASH) method (7), which used the sensitivity profile of receiver
coils as a complementary encoding function. In its initial embodiment, SMASH
tried to fit these profiles to sinusoidal harmonics in order to emulate the effect
of phase encoding. It was the first parallel acquisition method that produced

in vivo imaging and launched the field of parallel imaging using sensitivity encod-

ing. SMASH has since undergone a large number of changes and adaptations, as
a number of practical limitations prevented it from reaching commercial reliability.