Three factors of image quality, however, are well characterized: the signal-to-noise ratio (SNR), imaging time, and resolution. SNR is the ratio of the strength of the magnetism of precessing spins (MR image signal) relative to the random magnetic fields produced within the patient (noise). Resolution indicates the level of detail that can be seen in an image and is obtained by sampling wave patterns with a wavelength equal to the desired resolution distance (in all directions) as well as a complete set of waves that vary more slowly. The three quantities are related by the following equation:
SNR
a (voxel size)*(imaging time)1/2
where the size of a voxel (short for volume element, the three-dimensional equivalent of a picture element, or pixel) is given by the product of the two in-plane resolutions multiplied by the slice thickness. For example, if an MR image is collected with in-plane resolution decreased from 1.0 mm2 to 0.5 mm2, SNR is reduced by a factor of four but can be restored by increasing imaging time sixteenfold.
The equation represents an unbreakable physical barrier. Technological barriers, however, lower the SNR from its optimal state for a given voxel size and imaging time and limit the minimum resolution and imaging time. These issues are addressed below, and some current and future solutions are considered.
Increasing SNR
A high SNR for a fixed voxel size and imaging time can be obtained by increasing in vivo magnetization or the efficiency of detection. Magnetization can be increased by increasing the magnetic field B0 in which the patient is being scanned. The fraction of parallel, or detectable spins (roughly 10/million), is proportional to B0. Doubling the magnet strength of a scanner doubles this fraction and doubles the MR imaging signal to a first approximation. Higher manufacturing costs, patient safety, and other field strength-related phenomena are hindrances to using higher field strengths, but the simple notion that doubling the SNR can reduce scan time by a factor of four (from the above equation) for many scans is a powerful motivator.
The efficiency of detection can be increased by the proper choice of RF coils. Small coils are less noisy than large ones, but they can only detect signal from a limited area. Recent advances have led to phased array coils,[9] sets of coils that act independently but together measure signal from a wide area. Signal detection with a large set of small coils is typically less noisy than that obtained with one large coil. The drawback of phased array coils is the added hardware and computer power necessary to process multiple signals.
Imaging Time
Imaging time is limited by how fast all the spatial waves in k-space necessary to complete an image with the desired resolution can be collected. These waves are produced by turning on the field gradients. Stronger field gradients can produce a greater variety of waves in a shorter amount of time than can weaker field gradients. To measure a two-dimensional set of waves, these gradients must be able to change direction rapidly so that the signal can be sampled over all of k-space. Thus, larger gradient amplitudes and gradients with higher slew rates (the rate at which the gradient can change direction) allow data to be collected more rapidly. Smarter sampling strategies (e.g., sampling trajectories through k-space) also help reduce imaging time. A recent advance is `real-time' MR imaging. Still in its infancy, this technique includes real-time tracking of interventional devices, interactive movement of the imaging plane, and monitoring the passage of a bolus of contrast agent.
At least two factors involving patient safety may ultimately impose lower limits on imaging time: the specific absorption rate (SAR) and field gradient-induced nerve stimulation. The first factor, SAR, relates to patients becoming heated from exposure to the RF pulses. The RF pulses stimulate charged matter to vibrate at the Larmor frequency and add thermal energy to the patient. Current guidelines from the Food and Drug Administration limit power deposition to an amount which, if the patient had absolutely no cooling mechanisms at all, would heat the patient about 1ยบ C/hour. Consequently, the amount of RF that can be used to excite and refocus spins is limited. The second factor, nerve stimulation, occurs when the field gradients are switched too rapidly, inducing current flow in the affected nerves. Although many have experienced peripheral nerve stimulation, which is probably harmless, ventricular fibrillation must not be induced. Proper design of the coils that create field gradients can minimize, but not eliminate, this effect.
Resolution
Other than limited SNR, which is a significant limitation to high-resolution MR images, other barriers are surmountable. Higher resolution scans require more computer memory and faster computer processing. A set of 128 slices, each producing a 512 x 512 pixel image and sampled from a four-element phased array coil set, requires about a gigabyte of memory. This is now possible but was less feasible just 5 years ago. Component stability, necessary for precise spatial encoding, is also improving through the use of shielded gradient coils, an innovation that is critical to obtaining advanced sequences such as EPI and spiral scanning.
Another barrier to improved resolution is physiologic motion from bulk motion, peristalsis, respiratory motion, and cardiac motion. The different time scales of these motions present different challenges for collecting data. Physiological motion can be circumvented in three ways. First, the motion can be monitored and the imaging data collected only during a quiescent period when tissues are assumed to return to the same state. Second, the motion can be monitored and corrected when the image is reconstructed. Finally, the image data can be collected fast enough to avoid the motion. The first scheme has been the basic way to obtain routine images in the abdomen, with the use of cardiac and respiratory gating. An effective application of the second scheme is the use of navigator echoes[2] in cardiac examinations to correct for respiratory motion. The third scheme is the most robust because motion can often be too unrepeatable and too complex for the first two schemes. Again, however, high-resolution scans collected in a brief time have a limited SNR.
No comments:
Post a Comment