General Description of Parallel Imaging

The key elements of partially parallel imaging (PPI) are, first of all, it uses information obtained from arrays of RF coils sampling data in parallel. Since each coil element is located at a different position, this information can be used to perform some portion of spatial encoding that was usually done by the gradient fields, particularly the phase-encoding gradient field.

Imaging time is reduced because the phase-encoding process is the slowest part of image formation in MRI. So fundamentally PPI speeds up the MR image acquisition process and it does this without needing faster switching gradients, avoiding the tissue stimulation issues they can be caused by using additional RF pulses. In fact, parallel imaging can actually reduce the number of RF pulses so that the RF power deposited is not a big concern.

Now there are two fundamental approaches to parallel imaging. The first is an image based method in which the image is basically reconstructed from the signals detected by each coil element in the phased array with a reduced field of view and then merged. The most common implementation of this is known as SENSE, which stands for sensitivity encoding.

The second approach to producing parallel images is the k-space based approach. In this method the raw image data are analyzed and an interpolation algorithm is used to interpolate missing data in k-space before the Fourier transformation occurs. The algorithm known as GRAPPA is most commonly implemented on clinical MRI systems now, although other variations of this method are under development.

With parallel imaging, to speed up the imaging process, data are acquired for a smaller field of view. For instance, if you want to double the imaging speed, then use an FOV in phase-encoding direction that is half of what you really want. Obviously, this takes half of the imaging time but it also creates problems.

There is also the fact that the surface coils themselves receive MR signals in an inherently nonuniform fashion. These factors make it seem like it’s going to be very difficult to determine spatial properties of the signals based on the geometry and locations of the coils.

The key to parallel imaging is to acquire a little bit of extra data to get the information needed regarding the spatial distribution of RF coils’ sensitivity. One way to do this is by electromagnetic modeling of the coils’ properties applied to phantom measurements. However, because every patient differs and the way every patient couples with a coil differs, this approach is practical only for designing the RF coils used for parallel imaging. In the general clinical situation, extra data collection on the patient, either before or during the imaging, is used to gain a priori knowledge of how the patient and coil work together before parallel imaging reconstruction is performed.

Image Based Parallel Imaging Reconstruction

SENSE is the primary image based form of PPI which is, in some form or another, offered on all the high-end imaging systems of all the major manufacturers. In the common implementation of SENSE, sensitivity profiles are determined for each element of the phased array coil. Measuring the coil sensitivity profiles gives the system the additional information necessary to constrain the reconstruction process so that it can be accurate and efficient.

The two basic methods used for parallel imaging differ in their approaches to obtaining the data to solve the problem of recovering the spatial information from the set of RF coils. The coil sensitivity profiles contain the data that describes the spatial dependencies of each RF coil.