Applied Radiology

Dr. Uppot is a 4th-year Radiology Resident at the Medical Center of Delaware/Christiana Care Health System, Newark, DE. He received his MD from Louisiana State University Medical Center in Shreveport, LA. After completing his residency, Dr. Uppot will begin a fellowship in Abdominal Imaging and Interventional Radiology at Massachusetts General Hospital in Boston, MA in 2003.

Dr. Lally is the Chairman of the Department of Radiology at the Medical Center of Delaware/Christiana Care Health System, Newark, DE.

Fast imaging techniques for body magnetic resonance (MR) imaging have evolved from faster pulse sequences to the current development of phased-array coils, parallel imaging, and 3.0 T magnets. This report will address the technical challenges of body MR imaging, the historical developments of fast imaging techniques, and the current advances and clinical applications of fast imaging techniques for body MR imaging.

Currently, in most clinical practices, the role of magnetic resonance (MR) in body imaging is to characterize solid organ lesions identified on computed tomography, perform magnetic resonance cholangiopancreatography (MRCP), and evaluate fetal anomalies identified by ultrasound. Although MR imaging historically provides superior tissue contrast, offers multiplanar capabilities, has no ionizing radiation, and seems a natural choice for body imaging, its role in body imaging had been limited because of the slow image acquisition speed.

Technical challenges of body MR imaging

MR body imaging differs from neurologic or musculoskeletal imaging because of the need for high temporal resolution. ¹ High temporal resolution is needed to overcome cardiac and respiratory motion, vascular pulsations, and bowel peristalsis. The benefits of high temporal resolution with fast imaging techniques include decreased motion-related artifacts, improved patient acceptance, and increased patient throughput, reducing per patient examination costs. ^1-8 The tradeoffs of faster imaging, however, are decreased signal-to-noise ratio (SNR) and decreased spatial resolution. The challenge in fast MR imaging, therefore, is to decrease scan time, while maximizing SNR and spatial resolution. Fast MR imaging techniques are a result of advancements in pulse sequence design and k-space acquisition techniques, advancements in gradient technology, development of phased-array coils and parallel imaging, and increases in magnet strength.

Historical developments

The first body MR imaging was performed on July 2, 1977 on an MR scanner named "Indomitable" built by Dr. Raymond Damadian. ^9,10 It was an axial proton density image of the thorax and took 4 hours and 45 minutes to acquire. Since the "Indomitable," there has been a continuous trend toward faster MR imaging (Table 1). ^3,7,9-19

What determines the speed of MR image acquisition?

In general, the speed of MR image acquisition is based on many factors and scan parameters. Historically, attempts to reduce scan times were made by reducing the repetition time (TR), the number of phase-encoding steps, and the number of excitations (NEX) based on the basic formula ⁸ :

TR: TR is the time between successive radiofrequency (RF) excitations. Decreasing the TR reduces the scan time.

Phase-encoding steps: The phase-encoding gradient localizes the signal along the short axis of the anatomy based on the phase of the received MR signal. Scan time is dependent on the phase-encoding gradient and how fast it fills k-space ³ ; k-space is a matrix that serves as a temporary computer data storage space (Figure 1). Each data point in k-space holds information for the entire image (Figure 2A). Data in the center of k-space contains information regarding the image contrast (Figure 2B), and data in the periphery of k-space contain information regarding the spatial resolution of the image (Figure 2C). In conventional imaging, each phase-encoding step fills one line in k-space (Figure 3).

NEX : Increasing the NEX in each phase-encoding step allows more data to be collected in each line of k-space. A lower NEX decreases scan time, but also decreases SNR. In a conventional spin-echo imaging sequence based on a TR of 2 seconds (500 to 4000 ms), 256 phase-encoding steps (in a 256 * 256 matrix), and 1 NEX, the total scan time is 2000 * 256 * 1, or 8.5 minutes. ^3,7

This prolonged acquisition time limits successful body imaging because body physiology defines maximum breath-hold as approximately 20 seconds, normal respiratory cycle as approximately 3 seconds, and cardiac-cycle period as approximately 1 second. ¹ In order to avoid motion artifact, image acquisition times must be shorter than these normal physiologic processes. Strategies to reduce scan times involve reducing the value of each of the scan time parameters (TR, phase-encoding steps, and NEX).

Fast imaging techniques

Faster pulse sequences

The conventional spin-echo sequence was introduced in the mid 1980s. ⁵ It is the standard MR imaging pulse sequence with the application of a 90° RF pulse, a 180° refocusing pulse, the slice selection, and frequency-encoding and phase-encoding gradients (Figure 4A). ²⁰ Each RF excitation results in 1 phase-encoding step that fills 1 line of k-space. Because of the long acquisition times (4 to 8 minutes) with conventional spin-echo sequences, it is not suitable for body imaging. TR may be reduced in conventional spin-echo sequences; however, this affects tissue contrast and results in images that are T1-weighted. ⁵

In 1986, gradient-recalled echo (GRE) sequences were introduced. ²¹ These sequences modify the conventional spin echo by using smaller (<90°) flip angles and by using gradients, rather than a 180° rephasing pulse (Figure 4B). Gradient-recalled echo allows T1-weighted, T1- and T2*-weighted, T2*-weighted, and proton density images to be obtained with very short TRs. Image times were reduced to 200 to 300 msec, for a 128 matrix, fast enough to obtain single-slice breath- hold images of the abdomen. ¹

Since the development of spin-echo and GRE sequences, numerous pulse sequences have been developed (Table 2). ^1,2,7,22-39 These new pulse sequences are the result of changes to the basic spin-echo and GRE sequences combined with variations in how phase-encoding steps are performed and how k-space is acquired. These pulse sequences can be placed into 1 of 3 major categories: multiple spin-echo, GRE, and echoplanar imaging (EPI).

Multiple spin-echo sequences

Multiple spin echo is also known as rapid acquisition relaxation enhanced (RARE) sequence or fast spin echo on systems from GE Medical Systems (Milwaukee, WI) and turbo spin echo on systems from Phillips (Best, The Netherlands) and Siemens Medical Systems (Iselin, NJ). Multiple spin echo, as described by Henning in 1986, ⁴⁰ modifies the conventional spin-echo sequence by application of a series of 180° refocusing pulses and obtaining echoes after each 180° pulse (Figure 4C). With each excitation, multiple lines of k-space are filled. The number of echoes acquired per 90° RF pulse is known as the echo train length (ETL) and imaging scan time is decreased in proportion to the value of the ETL. For example, if the ETL is 4, the imaging acquisition speed is increased by a factor of 4. The echo time (TE) is only an "effective TE" and can be used to weigh the image: for T1-weighted images, a short effective TE is used; for T2-weighted images, a long effective TE is used.

The advantage of a multiple spin-echo sequence is that k-space is filled much more rapidly and image acquisition time is reduced. Multiple spin-echo sequences are insensitive to magnetic susceptibility. ¹

There are many disadvantages of multiple spin-echo sequences, such as blurring caused by T2 decay ⁴¹ --with each refocusing echo, the signal intensity drops, resulting in more distortion. Also, power deposition to the patient from multiple 180° RF pulses is a limiting factor. The gradient and spin-echo sequences have been developed to overcome this effect. ⁴² Another disadvantage of multiple spin-echo sequences is that fat is bright on T2-weighted images because of no signal loss due to J-coupling (carbon-proton interaction) effects. ^43,44 Bright fat can interfere with evaluation of bright signal from pathology (hyperintense lesion on T2-weighted images, or gadolinium-enhancing lesion on T1-weighted images). Fat suppression techniques are therefore important in fast MR imaging. ^2-5

GRE sequences

In GRE sequences, image contrast is determined by TR, TE, and flip angles. ⁴⁵ The rapid acquisition times in GRE allow for electrocardiogram-gated cine images of the heart and true 3D volumetric imaging. ¹ The disadvantage of GRE imaging is that it is susceptible to magnetic field inhomogeneities, which degrade image quality. Also, if TR is less than T2 decay time, echoes from one excitation will remain present to interfere with subsequent echoes, also known as transverse magnetization coherence. Pulse sequences have been developed to eliminate this interference ("spoiled" GRE), or capitalize on it and create constructive interference ("unspoiled" GRE). ¹

Spoiled GRE sequences use RF pulse phase changes to prohibit transverse magnetization coherence. Examples of spoiled GRE include spoiled fast low-angle shot (FLASH) and spoiled gradient-recalled acquisition in steady state (GRASS). Spoiled FLASH and spoiled GRASS sequences are T1-weighting, with T1 contrast controlled by the flip angle.

Sequences that encourage the coherence do so by refocusing the transverse magnetization. Such pulse sequences include fast imaging with steady-state precession (FISP), fast acquisition in steady state (FAST), contrast-enhanced fast acquisition in steady state (CE-FAST), steady-state free precession (SSFP), and reverse fast acquisition in steady state imaging with steady-state precession (PSIF). Both FISP and FAST images are based on the free induction decay after each pulse and are T2-/T1-weighted, while CE-FAST and PSIF are at the point of the subsequent RF pulse (Hahn echo) and are therefore more T2-weighted. ¹ In FISP, only the phase-encoding gradients are refocused, while in true-FISP (described later in this article) all three gradients (slice selection, frequency encoding, and phase encoding) are refocused.

In addition, ultra-fast GRE images using very low-flip angles (5° to 10°) and very short TRs (<5 ms) have been developed that provide a 'snap-shot' of tissue magnetization (snapshot FLASH and magnetization-prepared rapid acquisition with gradient echo [MPRAGE]). ¹ These are proton density­weighted images with very little tissue contrast. Ultra-fast snapshot FLASH can be used to obtain real-time single-slice imaging of the beating heart without cardiac gating. ²⁸

Echoplanar imaging

Echoplanar imaging was described by Mansfield in 1977 ¹¹ and was first used clinically in 1983. ⁴⁶ In EPI, all of k-space is filled after a single excitation (Figure 4D [compare also with Figure 5D]). Echoplanar imaging can be accomplished by applying a single 180° refocusing pulse after the initial excitation pulse (spin-echo EPI) or via rapid gradient switching (gradient-echo EPI). It can be single shot (all of k-space in a single shot) or, if increased spatial resolution is required, it can be interleaved (k-space is only partially filled after each excitation). Because of the need for rapid on/off switching and eddy-current free environments, EPI places great demands on gradient coils. ¹ Recent advances in gradient technology have made EPI a reality.

Ultra-fast scan times (total scan time <100 ms) in EPI have allowed for morphological and functional cardiac imaging ⁷ and hepatic imaging. ² The disadvantages of EPI are that it is limited because of magnetic susceptibility, chemical shift artifact, and lower SNR; and it can cause painful muscle stimulations if the gradients oscillate too rapidly. ⁷

Modified k-space acquisition

Scan times may be decreased by modifying how k-space is acquired.

Rectangular field of view

In conventional imaging, k-space has a square field of view (FOV) and typically 128 to 256 phase-encoding steps are needed to fill k-space (Figure 5A). An alternative method is to change the FOV to a rectangular space. ^5,7 With the phase-encoding gradient oriented along the short axis of the rectangle, less phase-encoding steps need to be performed to fill k-space (Figure 5B). In the abdomen and pelvis, where the cross section is more rectangular and no image information is present in the area anterior to the abdomen, this method can be used to reduce scan times. ⁵

Acquire only a part of k-space--half-Fourier imaging

As defined by the Hermitian conjugacy, ⁴⁷ k-space is mathematically symmetric. Therefore, only part of k-space may be filled and the remainder can be calculated. Two techniques include half-Fourier imaging and fractional echo imaging. Half-Fourier imaging makes use of the fact that k-space is symmetric along the phase-encoding gradient. Therefore, if slightly more than 50% of k-space is filled, the remainder can be calculated mathematically (Figure 5C).

Fractional echo imaging, in contrast, uses the fact that k-space is symmetric along the frequency gradient. Therefore, only a fraction of the signal from the frequency-encoding gradient is collected and the remaining is calculated (Figure 5D).

Fill multiple lines of k-space with each excitation

As described previously, multiple spin-echo and EPI sequences reduce scan times by filling k-space faster (Figure 5E). ⁴⁰

Spiral imaging

In spiral imaging, data in k-space are acquired in a series of spiral trajectories (Figure 5F). ⁴⁸ Spiral scanning decreases the switching demand on the gradient coils when compared with gradient EPI. Acquiring data in a spiral fashion is a more efficient use of gradient power. A spiral trajectory can cover a given area of k-space with fewer repetitions than linear acquisition of k-space. This method can reduce scanning time and still maintain spatial resolution. Spiral MR has less motion artifact than EPI because the spiral trajectory samples the center views of k-space more than EPI. ⁴⁹ Spiral imaging also has less misregistration artifacts (artifacts related to obliquely moving spins relative to phase- and frequency-encoding axes, appearing misplaced on the images). These qualities make spiral imaging effective for the evaluation of flow. Spiral MR imaging requires special reconstruction algorithms. ¹

Keyhole imaging

Because the center of k-space contains contrast information and the periphery of k-space contains spatial resolution information, in a dynamic study such as movement of a contrast bolus, only the central portion of k-space needs to be updated. ⁵⁰ This method saves time by only changing the central portion of k-space and leaving the peripheral portion unchanged (Figure 5G).

Current advances and clinical applications

Recent advances in fast imaging techniques for body MR imaging include advances in software: single-shot RARE, true FISP, and volumetric interpolated breath-hold examination (VIBE); and hardware: advances in gradient technology, phased-array coils and parallel imaging, 3.0 T magnets, and moving table configurations.

Single-shot half-Fourier RARE is a type of multiple spin-echo sequence in which 1 excitation fills half of k-space. The advantages of this sequence include rapid acquisition times of 600 ms per image; a wide contrast range with T1-weighted, T2-weighted, and hydrographic imaging; resistance to magnetic susceptibility artifacts; and reduced tissue-inflow artifacts. The disadvantages include high RF power deposition, image blurring as a result of T2 decay, cross talk, and hydrographic artifacts. ⁵¹ Its uses include MRCP, ⁵² fetal MR imaging, ²⁵ and bowel imaging. ²⁶

True FISP (Siemens Medical Systems), FIESTA (GE Medical Systems) or balanced FFE (Phillips) are unspoiled GRE sequences in which transverse magnetization is preserved and signal is acquired in a steady-state free precession mode. ⁵³ Its uses include cine imaging of the heart, ²⁹ thoracic aortic dissections, ³⁰ and small- and large-bowel imaging. ^31,32

Volumetric interpolated breath-hold examination is a 3D GRE sequence. Three-dimensional GRE may provide thinner sections, fat saturation, higher SNR, and comparable image contrast than 2D GRE; however, its use has been limited because of poor resolution and limited anatomic coverage. An advanced technique, VIBE allows a large volume of acquisitions with ¾2 mm pixel size within the time frame of a single breath-hold. ^54,55 It allows for image acquisition in 1 plane and reformatting in any other plane used for the assessment of liver lesions and high-resolution 3D MR cholangiography. ⁵⁴

Hardware

Gradient technology

Advances in gradient technology have allowed fast MR pulse sequences, such as EPI to become reality. Gradients are electromagnetic coils situated in a 3D plane around the bore of the magnet that perform the slice selection, frequency, and phase encoding. The physical act of switching each of the gradients on and off, or rise time (time to peak gradient strength), usually takes 600 µsec. ⁵⁶ Advances in gradient technology aim to increase the strength of the gradient coils and decrease the rise times. Current gradient coil peak strengths are 20 to 40 milliTeslas per meter (mT/m) and rise times are 150 to 300 µsec. ² High gradient switching rates are limited because they could induce currents that may stimulate the heart and peripheral nerves. ⁵⁷

Phased array coils and parallel imaging

Surface coils improve SNR by their proximity to the body. However, surface coils can only image a limited area over which the coil is placed. Phased-array coils are multiple surface coils distributed throughout the area of interest that are decoupled magnetically. The advantages of phased-array surface coils are a higher SNR due to the proximity of the coils to the body surface and the stronger signal that is received. Current phased-array coils contain four to six elements. ²

Parallel imaging makes use of each of these elements to provide spatial information. This spatial information acquired reduces the number of phase-encoding steps and, therefore, allows for a 2- to 5-fold reduction in scan times. ⁵⁸ Theoretically, all phase-encoding steps can be reduced with a very large number of receivers. ⁵⁹ However, most current efforts are directed toward partial parallel techniques in which a few phased-array elements are used to reduce the number of phase-encoding steps. Two partial parallel imaging sequences recently introduced include simultaneous acquisition of spatial harmonics (SMASH) and sensitivity encoding (SENSE).

First described in 1997, SMASH is a partial parallel imaging technique. ¹⁷ In SMASH, multiple receiver coils acquire signal simultaneously with different spatial sensitivities. These differences in spatial sensitivity are used to reconstruct the missing data in k-space from the reduced phase-encoding steps. Fewer phase-encoding steps allow for faster scan times. A 2- to 4-times savings in image acquisition time is possible with SMASH.

When the number of data points sampled in k-space is reduced, the acquisition time is reduced. This less dense k-space preserves spatial resolution, but reduces the FOV and introduces an aliasing, or fold over, artifact. In SENSE, multiple receiver coils are used and each receiver coil receives a different signal voltage from the imaged object based on the position of the receiver coil relative to the object. ¹⁸ Computational algorithms can use these differences in signal voltage to subtract the aliasing effect and create a whole image. The result of SENSE imaging is the acquisition of an unaliased image with only half the k-space data and, therefore, half the scan time. The disadvantage of SENSE is a decrease in SNR because of fewer k-space data points, and the raw computational power necessary for the complex reconstruction algorithms. ⁵⁸

Increased magnet strengths

Currently, most clinical practices have 1.0 and 1.5 T magnets. The FDA has approved 3.0 T magnets for clinical use. The main advantage of increasing gradients is increased SNR. MR imaging scanners with increased SNR supplement the SNR that is lost in faster pulse sequences.

Moving table configurations

Historically, conventional MR imaging is restricted to a single body region, ie, thorax, abdomen, or pelvis. With the development of real-time gradient-echo imaging with true FISP, imaging of multiple body regions is now feasible. Using a rolling table platform moving at a speed of 5 cm/sec and real-time true FISP imaging, MR imaging covering 150 cm, with 8-mm sections, and 110% overlap was achieved in 30 seconds and could detect 8-mm hepatic and pulmonary lesions. ¹⁹ This feasibility study opens up the possibility of whole-body MR imaging screening.

Conclusion

Fast imaging techniques for body MR imaging are in constant evolution. The choice of pulse sequence and hardware will depend on the clinical situation, the available hardware and software, and an understanding of the tradeoffs in spatial resolution, SNR, FOV, tissue contrast, and scanning time.

Current advances in fast techniques for body MR imaging offer the potential of moving from anatomic imaging to physiologic imaging. They offer the potential of moving from static imaging to real-time anatomic evaluation, as in MR fluoroscopy, and in real-time manipulation, as in MR intervention. ¹ Future advances in fast MR imaging will involve continued increases in the number of elements in phased-array coils and increases in magnet strengths.