Short NEX/NSA is 1 or less (partial averaging).
Medium NEX/NSA is 2 or 3.
Long or multiple NEX/NSA is 4 or more.
CONTRAST‐TO‐NOISE RATIO (CNR)
The CNR is defined as the difference in SNR between two adjacent areas. It is controlled by the same factors that affect SNR. All examinations should include images that demonstrate a good CNR between pathology and surrounding normal anatomy so that pathology is well visualized. The CNR between pathology and other structures can be increased by the following:
administration of contrast agents
utilization of T2‐weighted pulse sequences
suppression of normal tissues or pulse sequences that null signal from certain tissues e.g., short TI inversion recovery (STIR), fluid alternated inversion recovery (FLAIR) and magnetization‐prepared pulse sequences
use of pulse sequences that enhance flow (see Pulse sequences).
A note on fat suppression techniques
The CNR can be improved by suppressing signal from tissues that are not important, thereby increasing the visualization of tissues that are. Fat is the most common tissue that is nulled or suppressed in MRI and this is assumed for the majority of protocols described in Part 2, where all of the techniques described below are referred to as fat suppression.
Fat suppression is most commonly used to distinguish between fat and enhancing pathology in T1‐weighted pulse sequences and in a FSE/TSE T2‐weighted pulse sequence where fat and pathology are often isointense. However, signal from any tissue can be suppressed using the inversion recovery (IR) pulse sequence and some saturation techniques are used to null signal from water or background tissue. Further details on suppression of tissues other than fat are provided where relevant in Part 2.
There are several ways in which fat and other tissues are suppressed, including:
Chemical pre‐saturation: a 90° RF pulse is delivered at the specific precessional frequency of the magnetic moments of spins in either fat or water. This pulse is delivered to spins inside the imaging volume before the RF excitation pulse is applied, saturating them. No signal is therefore received from either fat or water when the echo is read.
Spectral pre‐saturation: an RF pulse of a greater magnitude than 90° is applied and inverts the magnetization in a tissue as in inversion recovery (IR) pulse sequences (see Pulse sequences). At the time from inversion (TI) that corresponds to the null point of the tissue, a 90° RF excitation pulse is applied. No signal is therefore received from that tissue when the echo is read.
Dixon technique (either 2‐point or 3‐point): a reconstructed image is obtained from only the spin population in water. This ‘water‐only’ image has no contribution from spins in fat. Images look similar to the pre‐saturation techniques described above but rely on the chemical shift between fat and water (the difference in the precessional frequencies of the magnetic moments of the spin population in fat and water). Images are acquired depending on whether the magnetic moments of fat and water spins are in or out of phase with each other. Unlike saturation techniques, this technique can be used after gadolinium (Gd) and at any field strength and is a very robust suppression method. Some systems use this technique to produce four images in one acquisition (water, fat, in and out of phase).
SPATIAL RESOLUTION
Spatial resolution is the ability to distinguish between two points as separate and distinct. It is controlled by the voxel size. Spatial resolution may be increased by selecting:
thin slices
fine matrices
a small FOV.
These criteria assume a fixed, square FOV so that if an uneven matrix is used, the pixels are rectangular and therefore spatial resolution is lost. Alternatively, square pixels can be maintained so that the phase matrix determines the size of the FOV along the phase encoding axis (phase FOV). Spatial resolution is maintained because the pixels are always square. However, the size of the FOV may be inadequate to cover the required anatomy in the phase direction of the image and any attempt to increase it increases scan time. In addition, SNR is often reduced due to the use of smaller, square pixels. In the interests of simplicity, a square FOV is assumed in Part 2, whereby the phase matrix determines spatial resolution, not the size of the phase FOV.
In Part 2, the following terms and approximate resolution parameters are suggested. The first number quoted is the frequency matrix; the second is the phase matrix (see also Table 2.1):
A coarse matrix is 256×128 or 256×192.
A medium matrix is 256×256 or 512×256.
A fine matrix is 512×512.
A very fine matrix is any matrix 1024×1024 or higher.
A small FOV is less than 16 cm or 160 mm.
A large FOV is more than 30 cm or 300 mm.
A thin slice/gap is 1 mm/1 mm (or thinner) to 3 mm/1 mm.
A medium slice/gap is 4 mm/2 mm to 6 mm/2.5 mm.
A large slice/gap is 7 mm/3 mm or thicker.
SCAN TIME
The scan time is the time required to complete the acquisition. The scan time can be decreased by using:
a short TR
a coarse phase matrix
the lowest NEX/NSA possible.
In addition to the SNR, CNR, spatial resolution and scan time, the following imaging options are also described under the Technical issues subheading:
Rectangular FOV: The use of rectangular FOV is often discussed in Part 2. It enables the acquisition of fine matrices but in scan times associated with coarse matrices. It is most useful when anatomy fits into the shape of a rectangle, for example a sagittal spine. The long axis of the rectangle usually corresponds to the frequency encoding axis and the shorter axis to phase encoding. This is important as certain phase artefacts, such as phase ghosting and aliasing, occur along the short axis of the rectangle. The dimension of the phase axis is usually expressed as a proportion or percentage of the frequency axis, for example 75%. On some systems, rectangular FOV and oversampling are not compatible. If this is so, signal‐producing anatomy existing beyond the FOV along the shorter phase axis is wrapped into the image. This is reduced by increasing the FOV, using spatial pre‐saturation bands to nullify unwanted signal or by expanding the short axis dimension to incorporate all signal‐producing anatomy (see Flow phenomena and artefacts).
Radial k‐space: Radial k‐space is a non‐Cartesian k‐space filling method. Data are not laid out in lines, but k‐space is filled in strips of data that rotate around the centre point of k‐space. The main advantage of this technique