After receiving a BA in mathematics from the University of Pennsylvania (1988), Frank Rybicki received a combined MD, PhD degree from the Harvard-MIT Division of Health Sciences and Technology (1994). Between 1994 and 1996, he served as a Research Associate in the Department of Radiology, Brigham and Women's Hospital, Boston, MA. In 1996 Dr. Rybicki completed a medical internship at Mount Auburn Hospital, Cambridge, MA. He is currently a fourth-year resident in diagnostic radiology at Brigham and Women's Hospital.

Dr. Mulkern is an Associate Professor of Radiology in the Department of Radiology, Children's Hospital and Harvard Medical School, Boston, MA.

Despite the wide availability of magnetic resonance imaging (MRI) techniques which can image faster than one slice per second, sub-second imaging, and specifically "single-shot" MRI, has not replaced slower "multi-shot" MRI in routine practice. In this article, "single-shot" refers to imaging a single slice in a single repetition time (TR). Specifically, the entire image is obtained via a train of echoes after a single slice selective excitation. Two types of echoes are described in this paper, gradient echoes (GE) and spin echoes (SE). Single-shot MRI (SS-MRI) sequences that use GE trains are generally termed Echo Planar Imaging (EPI); ^1-3 sequences that use SE trains ^4-5 have a variety of acronyms including single-shot fast-spin echo (SSFSE), rapid acquisition with relaxation enhancement (RARE), and Half-fourier turbo spin echo (HASTE). As defined, SS-MRI excludes an important fast imaging alternative: short TR gradient echo imaging in the steady state. This technique has comparable scan times and clinical roles that can be found in the literature. ^6-8

Before discussing the principles of SS-MRI, we answer the question, "Why hasn't SS-MRI completely replaced multi-shot MRI (MS-MRI) sequences that utilize many TR periods to acquire image data over several minutes?" The main reason is that, in general, multi-shot methods yield images of higher quality than single-shot methods. To illustrate, imagine yourself in a unique art gallery, and your task is to see the information in 64 paintings. The gallery is unique because the only light is overhead, and it is turned on for a brief moment and then quickly dimmed. While the gallery is illuminated, you must obtain information from all 64 paintings, one at a time. This is nearly analogous to SS-MRI. Turning on the overhead light represents the excitation, and the MRI signal begins at a maximum and rapidly decays (or dims). The information in the 64 paintings represents the patient data with each painting corresponding to an echo.

The analogy requires a modification based on the fact that in MRI, each echo must be sampled at fixed rate, for example, 128 samples per echo. This translates into a requirement that you blink your eyes 128 times per painting! Consequently, your task in the unique gallery (and SS-MRI) is not a leisurely exercise. Instead, SS-MRI requires the acquisition of a complete image data set (in this example 64 echoes, each sampled 128 times) after only a single excitation.

The same image obtained with MS-MRI shares the requirement of 128 samples for each of the 64 echoes. However, MS-MRI uses multiple excitations or "shots," and each excitation corresponds to turning on the overhead light. Thus, MS-MRI may use 64 excitations for 64 echoes, corresponding to turning on the overhead light for each painting. Therefore, the sampling (blinking) rate is slower rate for each echo (painting). In the gallery, you obtain more overall information. In MRI, multi-shot images generally have better image quality, but at the price of a longer imaging time. When more time is acceptable, SS-MRI has not replaced MS-MRI. However, when rapid imaging is needed, SS-MRI has become invaluable. The goals of this article are to describe the principles of SS-MRI with an emphasis on how the choice of GE or SE trains affects the image properties, and to illustrate how the properties of SS-MRI sequences are used in six clinical applications.

Principles of single shot MRI

MRI data acquisition --To form an image, data (e.g., 64 echoes, each with 128 samples) is used to generate the "k-space" representation of the image. Understanding the mathematical relationship between the k-space representation and the image itself is less important than understanding three basic points: 1) the k-space representation of the image exists. That is, there are two representations for each slice: the anatomic representation viewed by the radiologist, and the k-space representation; 2) like the anatomic representation, the k-space representation is a two dimensional matrix of pixel values. In the gallery example, k-space is a 64 * 128 matrix. Each of the 64 lines is occupied by one echo, and each of 128 entries per line is occupied by a sample; 3) each pixel in the k-space matrix is an individual sample of the signal, or transverse magnetization, measured after the slice selective excitation. The excitation (turning on the overhead light) tips the equilibrium longitudinal magnetization from a slice of spins into the "transverse" plane, the plane perpendicular to the main magnetic field. Once there, the precession of spins forms the signal which is acquired using appropriately oriented radiofrequency (RF) coils. The signal is acquired as a series of echoes used to construct the k-space representation of the image. The MRI terminology for this process is "filling k-space." While numerous books and articles describe k-space, ^9-11 this paper focuses on the echoes that fill k-space and how the choice of an echo (GE versus SE) leads to several image properties.

Gradient echoes and spin echoes --The two basic types of echoes used to fill k-space are GE and SE. A third type, "stimulated echoes," has a role in clinical MRI and is described in the literature. ¹² A GE is generated by successive application of two magnetic field gradients of opposite polarity. The first gradient dephases the spins; the second gradient, applied shortly after the first, rephases the spins. The GE peaks when the effects of the dephasing and rephasing gradients are balanced. The generation of a GE is often illustrated with a racing analogy. ¹³ This article uses a horse race to describe how spins can form echoes. Several horses (spins), each with a different but constant speed, begin a race at the starting line. During the first half of the race (the dephasing gradient), the horses spread along the track. At the midpoint of the race (the application of the rephasing gradient), the horses reverse their direction. Since each horse maintains its constant speed, they return to the starting line together (the peak of the GE).

The generation of a SE begins like a GE with a gradient dephasing the spins. However, instead of a rephasing gradient, a 180 degree refocusing pulse is applied. The refocusing pulse effectively reverses the polarity of all gradients that are present, including the dephasing gradient. This refocusing or reversal leads to the generation of the SE. In the horse race analogy, the first half of the race is identical. However, at the midpoint of the SE race, the refocusing pulse instantaneously transports each horse to its mirror position on the opposite side of the starting line. Each horse continues to run in the same direction at its constant speed, but after the refocusing pulse the slower horses are ahead of the faster ones. The SE peaks when the spins are refocused, corresponding to the time when the horses all return to the starting line.

Gradient and spin echo trains determine the image properties in SS-MRI --Unlike MS-MRI where multiple excitations are performed to fill k-space, in SS-MRI every line of k-space is filled after a single excitation by collecting multiple echoes as so-called "echo trains." A GE train produces a GE-SS-MRI sequence while a SE train produces a SE-SS-MRI sequence (figure 1). The fact that either echo train can fill all of k-space in one excitation was noted by Mansfield who coined the term "echo planar imaging" or EPI. ^1,2 However, the term EPI has become linked to GE-SS-MRI while sequences that use SE trains have acronyms such as SSFSE, RARE, and HASTE. In the following comparison of GE and SE methods, all four acronyms (EPI, SSFSE, RARE, HASTE) refer to single-shot imaging.

There are three differences between GE and SE trains that contribute to the image characteristics and ultimately to the clinical uses GE-SS-MRI and SE-SS-MRI.

Gradient echoes can be collected approximately four times as fast as spin echoes (figure 1). The reason is that SE trains require refocusing pulses; the time for each refocusing pulse, associated gradients, and the spin echo is approximately 4 msec. In contrast, GEs can be sequentially sampled a 1 msec intervals. Consequently, GE-SS-MRI is faster than SE-SS-MRI for an equivalent number of echoes.

The rate at which GEs fade (T2*) is faster than the rate at which SE fades (T2). In the gallery analogy, the overhead light dims faster for GE trains. The difference between T2 and T2* is based on two sources of signal loss, "reversible" and "irreversible." ^14,15 The etiology of "reversible" signal loss is unwanted magnetic field gradients that can be reversed by the refocusing pulse in the SE. Reversible gradients are either global (main field inhomogeneity) or local (air-tissue interfaces or metallic implants). Since reversible gradients can not be refocused in a GE, the associated signal is lost in GE-SS-MRI. "Irreversible" signal losses are random in nature; they can not be refocused by the 180 degree refocusing pulse and therefore contribute to signal loss in both GE-SS-MRI (T2*) and SE-SS-MRI (T2). As shown below, the T2* decay in GE trains makes perfusion and functional MRI (fMRI) of the brain possible. Therefore, these applications use predominantly EPI.

The artifact created from susceptibility shifts and the chemical shift between fat and water is refocused with SE trains. In EPI, both shifts cause oscillations of the signal which lead to spatial distortions. Complete descriptions ^16,17 include the fact that these distortions are primarily along the dimension of the image known as the "phase encoding" dimension. Both susceptibility (from bone and air) and chemical shift (from fat) provide obstacles for MRI in abdomen and contribute to the fact that SE-SS-MRI is more often used for abdominal applications such as fetal imaging and MRCP.

Before the applications of SS-MRI, we discuss one important feature that is independent of the type of echo train: the echo time (TE) which ultimately determines the degree of T2 or T2* weighting in the image. Since a single-shot MR image is generated from an echo train and each echo has a different T2* or T2 weighting, what is the "effective TE" of the image? Or, to make the question more specific, how can a SE-SS-MRI sequence be tailored for fetal MRI (which requires intermediate T2-weighting) and MRCP (which requires heavy T2 weighting)? The answer is related to how individual echoes are used to fill k-space, ¹⁸ but ultimately the radiologist chooses an effective TE which determines the T2 weighting.

Applications of SS-MRI

Echo planar imaging -- Perfusion imaging of the brain: Perfusion imaging evaluates relative cerebral blood volume and flow and has the potential to contribute to the management of patients with cerebral ischemia. ¹⁹ Imaging is usually performed by the bolus tracking method, which is enabled by two features of EPI: image speed (for example, 10 slices of the brain every 2.5 seconds) and T2* decay. Between 15 and 30 seconds after bolus intravenous administration of a gadolinium chelate contrast agent, the concentration in the cerebral vessels peaks, and the vessels achieve a markedly different susceptibility than the surrounding tissue. ²⁰ This difference in susceptibility creates local microscopic field gradients around the vessels. Similar to local gradients at an air-tissue interfaces, these gradients cause a reversible loss which contributes to T2* decay to a far greater extent than T2 decay. In perfusion imaging, the degree of signal loss is captured over time by EPI and is then analyzed ^21,22 to determine relative cerebral blood flow (figure 2).

Diffusion imaging of the brain: Diffusion-weighted MR images provide the earliest imaging of acute cerebral infarction with intense signal generated by the decreased apparent diffusion coefficient (ADC) at the infarct site. The feature of EPI that enables diffusion MRI is speed. Diffusion weighting in EPI is achieved by a modification to the pulse sequence as shown in figure 1A. Specifically, additional gradients (diffusion sensitization gradients) are placed around an additional RF pulse, ²³ which only minimally compromises the image speed that is required to freeze the macroscopic motion of the normal brain pulsation from the beating heart. With macroscopic motion frozen, MRI can detect the microscopic random motion of water molecules within and between cells. In acute infarction, there is decreased microscopic motion and thus a decreased ADC. ^24,25 On diffusion-weighted images, regions of low ADC are hyperintense (figure 3). One interpretation of the decreased ADC in infarcted brain is cell swelling (cytotoxic edema), leading to more intracelluar water within a given voxel. Because speed is the essential requirement for diffusion imaging, SE-SS-MRI can also been used. However, for technical reasons, ²⁶ EPI is more commonly employed in clinical practice.

Functional MRI of the brain: Functional MRI (fMRI) maps regions of the brain that become "activated" by a task performed by the patient, for example finger tapping or viewing a strobe light. These tasks would be expected to activate the motor strip and visual centers, respectively. fMRI has expanded from research protocols to a promising tool for presurgical diagnosis. For example, fMRI maps identify the motor cortex or primary visual centers preoperatively so these areas can be avoided as much as possible during surgery. ^27,28

As with perfusion imaging, the speed and T2* sensitivity makes EPI the most widely used sequence for fMRI. The physiology measured by fMRI is the activation of regions in the brain that receive more oxygenated blood than neighboring or "unactivated" regions. ²⁹ Since deoxyhemoglobin is paramagnetic and oxyhemoglobin is not, the brain surrounding vessels with a higher concentration of deoxyhemoglobin has more rapid T2* decay than brain surrounding vessels with more oxyhemoglobin. The change in T2* of brain tissue surrounding a vessel secondary to differences in deoxyhemoglobin concentration (fMRI) is less than the change in T2* from perfusion MRI with the bolus tracking method. This translates into relatively less contrast in fMRI. Thus, measurements (20 EPI slices every 2 seconds) are repeated with and without the activation task; for example, 10 cycles of 20 seconds of finger tapping (the "on state") followed by 20 seconds without tapping (the "off state"). Various statistical methods ³⁰ can then be used to determine which parts of the brain are responding to the on/off states, and color maps of those regions can be overlaid on the anatomical images (figure 4).

Limitations of EPI: Each of the three applications of EPI provides complementary information to routine brain MRI protocols that rely heavily on MS-MRI. The fact that each application images the brain may suggest a limitation of EPI in other body parts. Although EPI is capable of freezing motion in the abdomen, ^31,32 artifacts from susceptibility and chemical shift has limited the clinical use of EPI outside the brain (figure 5). In clinical evaluation of the abdomen (and chest), SE-SS-MRI, although slightly slower than EPI, has predominated, primarily because of the better image quality.

Spin echo SS-MRI -- Obstetric MRI: Applications of obstetric MRI include evaluation of maternal disease and evaluation of the fetus when sonography is either limited or indeterminate for an anomaly. When obstetric MRI is performed for the fetus, ^33,34 SE-SS-MRI is used to meet three requirements: speed (to freeze fetal motion), the removal of susceptibility and chemical shift artifact, and intermediate T2 weighting. The need for intermediate T2 weighting in fetal MRI introduces T2-related signal loss in the later echoes. This obstacle is overcome by utilizing a mathematical relationship between the information contained in the first half of the echoes and the information in the second half of the echoes. Consequently, only the first half of the echoes (or, in practice, slightly greater than the first half) is required in so-called "half-Fourier" imaging ^35,36 which freezes fetal motion (figure 6).

MR cholangiopancreatography (MRCP) --Single-shot MRCP provides a noninvasive breath-hold evaluation of the biliary tract. ³⁷ The imaging requirements for MRCP are similar to fetal imaging: speed (to freeze motion), the removal of chemical shift and susceptibility artifact (typically from air in the duodenum or surgical clips), and T2 weighting. Because of its very long T2, bile is discriminated from shorter T2 tissues such as liver and pancreas. Clinical MRCP is typically performed with breath-hold half-Fourier SE-SS-MRI with a single thick slice (e.g., 5 cm) acquisition followed by multiple thin slice (e.g., 3 mm) acquisitions which can be viewed after maximum-intensity projection (MIP) reconstruction (figure 7).

MRI of the pulmonary parenchyma --The many challenges imposed by MRI of the pulmonary parenchyma include motion and large local gradients created from the alveolar air/soft-tissue interfaces. Although the many attempts at pulmonary MRI have included multi-shot methods, the speed and insensitivity to susceptibility of SE-SS-MRI has been used for breath hold, ECG-triggered MRI (figure 8) of many pulmonary disorders. ³⁸

Conclusion

SS-MRI uses only a single excitation to generate a complete image data set. This data is acquired via a train of either gradient echoes or spin echoes, and the type of echo train used in a single-shot method determines several imaging properties: image speed, T2* versus T2 decay, and artifact from susceptibility and chemical shifts. We hope that this paper has provided a concise review of SS-MRI, image properties in single-shot imaging methods, and clinical applications.

Acknowledgements

We thank Drs. Clare Tempany, Mary Frates, Gregory Sica, Liangge Hsu, Richard Schwartz, Russel Blinder, Carol Barnewolt, and Michael Rivkin for their assistance with the figures.