Qualitative and Quantitative Ultrashort Echo Time Imaging of Musculoskeletal Tissues

Abstract Ultrashort echo time (UTE) sequences represent a group of clinically compatible techniques that are capable of using echo times < 1 ms. With these techniques, direct imaging of short T2/T2* tissues or tissue components can now be performed. Continuing modifications to the UTE techniques have allowed for faster and more robust sequences now comparable with conventional clinical sequences. UTE also allows for morphological imaging and quantitative evaluation in a manner not previously possible with conventional imaging sequences utilizing much longer echo times. Numerous potential clinical applications have emerged that are discussed in this review article.

Since the earliest years of clinical MR imaging, the primary focus has been on tissues with long T2 relaxation. 1,2 Included in this group are tissues that normally demonstrate a relatively long T2 such as the brain and liver. It has long been recognized certain tissues with short mean T2 components typically lack detectable signal using conventional sequences, such as normal cortical bone and tendon. 3 These tissues can only be visualized when affected by advanced pathology, causing an increase in water and water mobility, such as when fluid is located in the gap of a torn tendon. However, there is no opportunity to diagnose disease in earlier states, and contrast manipulation of tissue constituents or the evaluation of contrast enhancement is impossible. Furthermore, quantitative measurements of tissue (including T1, T2, T2 Ã , T1ρ, proton density [PD], and magnetization transfer) are not feasible.
In more recent years, a group of clinically compatible sequences has emerged that can image tissues with very short transverse relaxation times, together referred to as the ultrashort echo time (UTE) group of sequences. 4 Common to this group is a reduction of echo time (TE), which allows for the detection of a short T2/T2 Ã signal. With detectable signal from short T2/T2 Ã tissues, both morphological and quantitative evaluation become possible. This promises to revolution-ize musculoskeletal imaging. Rather than relying on pathologic increases in T2 to visualize structures such as bone, tendons, ligaments, and menisci, we now have the opportunity to visualize the tissue directly and evaluate tissue components and properties selectively.
This article reviews the UTE group of sequences in three parts. First, we review the UTE sequence and subsequent modifications used to improve coverage, speed, or contrast. Second, we review morphological and quantitative imaging techniques that have emerged using UTE sequences. Finally, we discuss potential the clinical applications of UTE sequences including highlighting anatomy that has not been seen using conventional clinical sequences.

Techniques
The slice-selective UTE sequence was first described in 1989 by Pauly et al. 5 The TE is limited only by the time required to turn off the radiofrequency (RF) excitation and to enable the receiver. K-space is filled with a radial trajectory, with the acquisition starting at the center. By beginning at the center of k-space, multiple delays can be eliminated including those associated with the phase-encode gradient, dephasing gradient, and first half of the readout gradient. 5,6 Data sampling occurs as the acquisition gradient is being ramped up and continues through the plateau (►Fig. 1). Two-dimensional (2D) imaging utilizes slice selection, performed with two half excitations: the first with a positive gradient and the second with a negative gradient, combining to add as a single complete excitation pulse. Three-dimensional (3D) imaging can be performed with a nonselective short rectangular pulse and 3D radial acquisition. 6 By using appropriate coils and prescription of the field of view, aliasing can be avoided. After acquisition, the raw k-space data are mapped onto a grid and reconstructed by inverse Fourier transform. UTE sequences suffer from relatively long acquisition times. With 2D UTE sequences, two TR periods per line of k-space are required (compared with one used in conventional 2D sequences), and a single radius of k-space is mapped (compared with a Cartesian line in conventional sequences). In addition, radial mapping is less effective than rectilinear (Cartesian) methods because there is tremendous oversampling of low frequencies and undersampling of high frequencies (π/2 times more TRs are required for the k-space periphery to satisfy the Nyquist criteria). 7 However, more advanced trajectories have been proposed to improve efficiency including spiral and cones projection methods for k-space filling. 8,9 Other sequences in the UTE family used on clinical scanners include the acquisition-weighted stack of spirals sequence 10 that uses selective excitation, variableduration slice encoding, and mobile spiral readout to sample a cylindrical volume in k-space and the variable TE sequence 11 that uses Cartesian k-space sampling with asymmetric RF pulses, partial echoes, and ramp sampling. With UTE sequences, TEs as short as 8 μs (0.008 ms) have been achieved, although TE values ranging from 20 to 100 μs are more typical.
Another group of sequences does not switch off the readout gradients between successive TR intervals but only slightly adjusts them. Because encoding starts at full speed immediately at the time of excitation, a nominal TE of zero is achieved, hence the name zero TE (ZTE). 12,13 ZTE uses a nonselective hard pulse excitation followed by 3D centerout radial sampling. The RF pulse can also be a longer, frequency-modulated pulse with an interleaved transmitreceive operation, known as sweep imaging with Fourier transformation (SWIFT). 14 The finite RF transmit-receive switching times between excitation and reception lead to a spherical gap in the center of k-space. Multiple methods of addressing this problem have been proposed including acquisition oversampling and algebraic reconstruction in the ZTE technique 12 or filling with a Cartesian single-point imaging technique, known as pointwise encoding time reduction with radial acquisition (PETRA). 15 ZTE techniques have potential for imaging spins with extremely short T2/ T2 Ã s, such as collagen protons, which may not be directly detectable with UTE sequences.

Morphological Evaluation
UTE sequences detect signal from tissues with both long and short/ultrashort mean transverse relaxation times. However, signal-to-noise and contrast-to-noise ratios are not synonymous, and there can be insufficient contrast between structures for optimal morphological evaluation when using UTE sequences. The simplest technique for selective visualization of short T2/T2 Ã components is with image subtraction between a UTE image (which detects long and short T2 Ã components) and an image with a longer TE (which detects longer T2 Ã components) (►Fig. 2). Variable weighting of the subtraction images can provide improved contrast of tendons, cartilage, bone, and menisci. 16 Selective visualization of short T2/T2 Ã components can also be achieved with the use of preparation pulses including saturation or inversion pulses. Saturation of long T2 Ã components can be achieved with a long RF pulse followed by a crusher gradient and subsequent UTE imaging. 17,18 This technique relies on the fact that the short T2 Ã components experience significant transverse relaxation during the long RF pulse and cannot be saturated. Thus significant longitudinal magnetization exists after the long saturation pulse and provides the short T2 Ã signal and contrast. The use of two consecutive saturation pulses and spoilers can be used for combined water and fat suppression, known as water-and fat-suppressed projection imaging (WASPI). 17 Off-resonance saturation with UTE acquisition and subtraction can provide high contrast images of short T2 Ã components. 19 This technique exploits the broad spectral linewidths (several kilohertz) that short T2 Ã components demonstrate. An RF pulse placed at a frequency far from the narrow linewidths of long T2 water and fat is expected to selectively suppress the short T2 Ã components, either through direct saturation, 20 cross- Fig. 1 (a) The two-dimensional ultrashort echo time pulse sequence. (b) Magnetization diagram. Two-dimensional imaging utilizes slice selection performed with two half excitations: the first with a positive gradient and the second with a negative gradient, combining to add as a single complete excitation pulse. Data acquisition begins while the readout gradient is being ramped. DAW, digital acquisition window; FID, free induction decay; RF, radiofrequency. relaxation between collagen protons and protons in bound water 21 or chemical exchange between bound and free water. 22 Subtraction of the two UTE images, acquired with and without the off-resonance preparation pulse, will highlight the short T2 Ã components.
Short T2 Ã components can also be selectively imaged using inversion recovery (IR) pulses. With single IR-UTE, a relatively long single adiabatic inversion recovery pulse (8.6 ms in duration) is used to invert the longitudinal magnetization of long T2 Ã water. 8,23,24 UTE data acquisitions are then begun at an inversion time (TI) designed to allow the inverted longitudinal magnetization of free water to be close to the null point 8 (►Figs. 3a, b). Magnetization of fat is also inverted along with free water, and > 80% nulling of the signal from both fat and long-T2 Ã components can be achieved with a single inversion pulse. Alternatively, two inversion pulses can be used, such as with the double IR-UTE technique, allowing more complete nulling of both fat and long T2 Ã water components 25,26 (►Figs. 3c, d). Inversion-based approaches are less signal-to-noise efficient compared with saturation-based approaches but are much less sensitive to B 0 and B 1 inhomogeneities.
Differences in phase between tissues and tissue components can also be used for morphological evaluation. During the RF pulse and readout periods of the UTE sequence, differences in phase manifest and can be used as a contrast mechanism. 27 In some instances, phase images may provide more contrast between structures compared with magnitude images, 18 and the combination of the two can yield very high contrast susceptibility-weighted images. 28

Quantitative Evaluation
Using the UTE techniques, quantitative measurements of nuclear magnetic resonance relaxation times can also be performed. One of the most frequently used techniques is measurement of transverse decay times. Using a constant repetition time (TR) and variable TE, beginning with an ultrashort TE, T2 Ã can be quantified. Because the UTE technique is a gradient-echo type sequence without a refocusing pulse, T2 Ã rather than T2 is measured, and relaxation time is dependent on spin-spin interaction, tissue hydration, and susceptibility effects. 29 However, the use of the UTE does not allow for much true T2 decay to occur, 30 and for fast relaxing components, T2 Ã approaches T2. 6 From signal intensity measurements generated from the multiple TE images, T2 Ã can be fit using a monoexponential decay model or with a multiexponential decay model. A monoexponential (or single-component) model measures the mean transverse relaxation time within a voxel or region of interest. However, in actuality, all biological tissues are heterogeneous and contain a combination of tissue components. With the UTE technique, tissues of the musculoskeletal system can be divided into components with "short" and "long" relaxation times, representing water bound to macromolecules (such as collagen and proteoglycan/glycosaminoglycan [GAG]) and free water, respectively. 31 Models that contain more than two components have not been used routinely on clinical imaging systems. In particular, multiexponential T2 Ã analyses are very sensitive to spatial resolution, image signal-to-noise ratio, the number and spacing of echoes, the number of fitting components, and the differences between T2 Ã values of individual components. 32 Much of the data on the total number and relaxation times of measureable components comes from studies using highperformance spectroscopic systems and Carr-Purcell-Meiboom-Gill (CPMG) sequences that contain refocusing pulses and therefore measure T2. Although measurements made with the UTE sequence will incorporate susceptibility effects resulting in T2 Ã values that are shorter than T2, it is assumed that the number and relationship between components is similar between measurements made with CPMG sequences on spectroscopic systems and measurements made with UTE sequences on clinical MRI systems. With regard to the reproducibility of T2 Ã quantification, similar results have been obtained when comparing 2D-UTE, 3D-UTE, and 3D-UTE-Cones sequences. 33 T2 Ã measurements are complicated by the magic angle effect. 34 It can cause a 10-fold difference in signal intensity 34 and significant variations in mono-and biexponential T2 Ã measurements. For anisotropic tissues, Wang and Xia demonstrated that both the total number and component T2 relaxation times vary depending on orientation with respect to B 0 . 35 Despite this, previous studies have shown that data from UTE sequences with bicomponent T2 Ã analyses (specifically percentage of short and long components) is less sensitive to magic angle effects compared with monoexponential T2 Ã . 36 T1ρ has also been measured with the UTE sequence. In T1ρ imaging, a 90-degree pulse is delivered and a spin-lock pulse is applied. During the locking pulse, spin relaxation occurs via the time constant T1ρ. Thereafter, a À 90-degree pulse flips the magnetization back into the longitudinal plane, a crusher gradient spoils residual transverse magnetization, and the UTE sequence is used for acquisition. Using various spin-lock times, T1ρ is quantified using either a mono-or multiexponential decay model. 37 Similar to T2, T1ρ as measured with the UTE technique has been shown to be strongly affected by the magic angle effect. 38 Techniques that are less sensitive to magic angle effects include T1 relaxation and magnetization transfer (MT) techniques. T1 relaxation time depends on tissue water content as well as the macromolecular environment. 29 T1 quantification of short T2 Ã tissues has been performed using saturation recovery 18,39 and variable flip angle 40 approaches combined with UTE acquisition, for 2D and 3D imaging, respectively. MT refers to transfer of longitudinal magnetization from the bound proton pool to the free proton pool. When combined with the UTE sequence, MT imaging can be performed on short T2 tissues. The UTE-MT sequence uses an off-resonance saturation pulse for MT preparation. The off-resonance pulse can be Gaussian 41 or Fermi-shaped. 19,42 The 2D or 3D UTE sequence is used for acquisition. Using the two source images of the UTE-MT sequence, which includes a regular UTE image without the saturation pulse and one after the saturation pulse, a quantitative off-resonance saturation ratio (OSR) can be derived by subtracting the signal intensity from the two images and dividing by the signal intensity of the UTE image without saturation pulse. This same ratio has also been termed the MT ratio, but OSR is a more correct term because magnetization transfer as well as direct saturation effects are reflected in the ratio. [42][43][44][45] Importantly, OSR is highly dependent on the specific technical variables used and varies according to field strength as well as details of the off-resonance saturation pulse (including specific offresonance frequency and power values). 46  UTE (a, b), a single adiabatic IR pulse is used to invert and null the free water magnetization. The bound water magnetization is not inverted and is detected using a UTE data acquisition. In double IR-UTE (c, d), two IR pulses are used to invert and null the longitudinal magnetization of long T2 water and fat, respectively. Recovered bound water signal is detected with a UTE data acquisition. BW, bound water; DAW, digital acquisition window; FID, free induction decay; FW, free water; RF, radiofrequency.

Applications Cortical Bone
Osteoporosis is a metabolic bone disease that affects millions of people each year. Insufficiency fractures result in serious long-term disability and death in a large number of patients. 47 About 80% of the skeleton is cortical bone, and $ 80% of all fractures associated with advanced age arise at sites that are composed of mainly cortical bone. 48 Imaging of bone is important, and dual-energy X-ray absorptiometry (DEXA) represents the most commonly used technique. Unfortunately, measurements of bone mineral density with DEXA only predict fractures with a 30 to 50% success rate. [49][50][51][52] The major missing factors in accounting for the risk may be the contribution of bone organic matrix and water to the biomechanical properties of bone. UTE imaging is uniquely suited to address this need, and many qualitative and quantitative UTE techniques were initially developed for use with cortical bone.

Anatomy
Cortical bone is a composite material consisting of mineral ($ 43% by volume), organic matrix ($ 35%), and water ($ 22%). 53,54 Bone mineral provides stiffness and strength; collagen provides ductility and the ability to absorb energy before fracturing. Bone water contributes to viscoelasticity and poroelasticity. 54 Both the material composition and highly complex and hierarchical physical structure contribute to the unique strength of bone. 55 Water in cortical bone is present at various locations and in different states. 53,54 In normal bone a small fraction of this water exists in a relatively unrestricted (free) form in Haversian canals, with typical diameters > 30 µm, as well as in lacunae and canaliculi, with typical diameters of $ 10 µm to $ 0.5 µm, respectively. A larger portion of cortical bone water exists in a restricted (bound) form, either tightly bound to the crystals of the apatite-like mineral or loosely bound to the organic matrix. [56][57][58][59][60] Bound and free water make different contributions to the mechanical properties of bone. 61,62 Bound water is directly related to bone strength and toughness, whereas free water is inversely related to modulus of elasticity. Thus it is important to separate the two in studies of bone quality.

Magnetic Resonance Imaging
Using spectroscopic systems, studies have found three pools of protons including collagen methylene protons, collagenbound water, and a broad peak consisting of pore water and lipid. 61 With UTE sequences, the collagen-bound water and broad peak containing pore water and lipid can be directly detected, but collagen methylene protons cannot because the T2 Ã values are considered supershort (< 50 µs). 63 However, using MT techniques, all pools may be detected, either directly or indirectly.
Qualitative (►Fig. 4) and quantitative UTE techniques have been successfully implemented for the evaluation of the individual pools of water in cortical bone. 18 Single-component T2 Ã values have been reported to be < 1 ms. 39,64 Clinically compatible UTE techniques that can distinguish between bound and free water include T2 Ã with bicomponent analysis, single adiabatic inversion, and double adiabatic inversion 23 (►Fig. 5). Recently, two clinically compatible UTE sequence-derived indexes were introduced that significantly correlate with µCT porosity: the porosity index 65,66  and suppression ratio. 67 The porosity index is calculated from two acquisitions, one with a minimum TE (such as 50 µs), capturing signal from pore water and collagen-bound water, and a second with a longer TE, such as 1.2 ms 65 or 2.0 ms 66 capturing only signal from pore water. The porosity index is defined as the ratio of mean intensity in the longer TE image to that of the minimum TE image. The suppression ratio is also calculated from two acquisitions, one image with long T2 Ã suppression and an unsuppressed image. It is defined as the ratio of mean intensity between the unsuppressed and suppressed images. The concept is that suppression will increase with increasing pore sizes, which are associated with longer T2 Ã values.

Articular Cartilage
Osteoarthritis (OA) is a disease that affects many millions and has a substantial impact on the health care system. Conventional MRI is routinely used for the diagnosis of advanced OA where articular cartilage is significantly damaged or lost. However, conventional MRI is much less sensitive for the detection of OA in the early stages and limited with regard to quantification. Many UTE techniques are well suited for the qualitative and quantitative evaluation of articular cartilage.

Anatomy
Articular hyaline cartilage composition varies by depth from the surface as well as by location in the body. However, it is composed of $ 65 to 80% water, 68 and the dry weight is composed of $ 60% collagen (predominantly type II) and 12% sulfated proteoglycan. 69,70 Cartilage at weightbearing locations contains more GAG and water compared with nonweightbearing locations. [71][72][73] Articular hyaline cartilage demonstrates zonal variability. Four zones have been described including the most superficial zone, tangential zone, middle zone, and deep zone. 74,75 At the superficial regions, collagen fibrils are fine and parallel in orientation with respect to the articular surface. At the deeper regions, collagen fibrils are thicker and perpendicular in orientation with respect to the subchondral bone. 76 In addition, proteoglycan content increases in concentration with increasing depth from the surface. 77 A metabolically active region of calcified cartilage is present at the junction between articular cartilage and subchondral bone, containing type X collagen. 78 This layer of cartilage is $ 200 µm in thickness, with decreasing thickness seen with advancing age. 79

Magnetic Resonance Imaging
Using spectroscopic systems and fresh cartilage plugs obtained from bovine knee joints, studies have found four pools of protons including those associated with collagen (T2 $0.02 ms), those associated with mobile proteoglycan (T2 $ 1 ms), water molecules trapped within collagen fibrils (T2 $ 4 ms), and bulk water (T2 $ 20 ms). 80 Conventional MRI techniques with longer TEs capture signal from the abundant bulk water component but are unable to directly visualize any other components.
UTE sequences can be used to detect and highlight the short T2 Ã components (►Fig. 6). UTE sequences can also be used to selectively visualize the calcified layer of cartilage (►Fig. 2) that may be involved in the pathogenesis of OA. 81,82 Quantitative T2 Ã analysis obtained with the UTE technique has been shown to be more sensitive to matrix degeneration compared with T2 values obtained with conventional methods. 83 With worsening degeneration, monoexponential UTE-T2 Ã values have been found to decrease. 83 This is an opposite trend compared with conventional T2 values that increase with worsening degeneration. Using bicomponent analysis, a significant correlation between increasing short T2 Ã fractions and worsening degrees of degeneration has been shown. 36

Menisci of the Knee
The progression of OA is now viewed as a whole-joint process that can result from abnormalities in articular cartilage, subchondral bone, synovium, capsule, and meniscus. 84 Located between the femoral condyles and tibial plateaus, the menisci play important roles in load transmission, shock absorption, and maintenance of joint stability. 85 Mechanical impairment of the meniscus alters the weightbearing capacities of the knee joint, leading to damage to the adjacent articular cartilage and subchondral bone, eventually contributing to the progression of OA. 86,87 Menisci are predominantly hypointense on conventional imaging, which reflects the predominance of short T2/T2 Ã components. UTE imaging is well suited for the qualitative and quantitative evaluation of meniscal tissue.

Anatomy
The menisci of the knee are composed of $ 70% water, and the dry weight is composed of $ 60 to 70% collagen and 2 to 8% proteoglycan. 88 The hierarchical fiber structure of the knee menisci is complex, with circumferential bundles representing the largest group. 89,90 Radial fibers are another prominent fiber group and can appear as large bundles, referred to as radial ties or sheets. 91 Similar to cartilage, meniscal composition also shows zonal heterogeneity, with the peripheral portion containing higher vascularity 92 and a smaller concentration of proteoglycan. 93 Magnetic Resonance Imaging UTE imaging has been used to demonstrate the fibrocartilage and fibrillary network of the menisci. 94 After intravenous contrast administration with UTE acquisition, the vascularized portions of the meniscus and meniscal root ligaments can be readily visualized 95 (►Fig. 7). Meniscal calcifications are also readily identified and can be quantified using UTE imaging. 96 Quantitative UTE imaging with monoexponential T2 Ã analysis has been used to demonstrate subclinical meniscus degeneration. 97 With bicomponent T2 Ã analysis, quantification of normal versus abnormal meniscal regions can be made. 4,31 Changes in meniscal proteoglycan and GAG content can be detected with the UTE-T1ρ sequence. 37

Spine (Intervertebral Disks)
Low back pain is a debilitating condition. Although multifactorial, intervertebral disk (IVD) degeneration is one cause of chronic back pain, 98,99 and disk degeneration can lead to disk herniation and radicular pain. Although the exact pathogenesis is unknown, disk degeneration may involve alterations in tissues of the diskovertebral junction (DVJ), which provides mechanical and nutritional support. Conventional MR imaging of the lumbar spine has focused on morphological evaluation of the disk. UTE techniques enable imaging of the DVJ with high contrast, as well as new ways of evaluating the disk proper.

Anatomy
The human spine comprises 23 IVDs positioned between vertebrae that facilitate back movement. The IVDs are bounded peripherally by the anterior and posterior longitudinal ligaments of the spine and axially by the cartilaginous end plates 100 at the DVJ. The IVDs consist of a gelatinous nucleus pulposus (NP) in the center, surrounded by annulus fibrosus (AF). Healthy NPs are well hydrated (70-90% water) and rich in GAGs (65% of dry weight). The AF has markedly less water content (60-70%) and is rich in collagen (50-60% of dry weight) that form concentric lamellae that alternate in orientation. 101,102 At the DVJ, cartilaginous end plates (CEPs) form an interface between the IVD proper and vertebral body. The CEP in adults consists of layers of uncalcified and calcified cartilage 103 that is $ 1 mm thick. 104 The vertebral body provides a bed of capillaries that adhere directly onto the CEP, providing metabolites needed for the cell of the IVD.

Magnetic Resonance Imaging
The IVD is generally considered a tissue with fairly long T2 values ranging from $ 40 ms in the AF to > 80 ms in the NP. Conventional MR imaging typically utilizes fast spin-echo sequences in the sagittal plane, with long TEs up to 100 ms that are capable of revealing the morphologic characteristics of the NP. 105 However, the region of the DVJ is seen with low signal intensity with conventional imaging (►Fig. 8a). This makes it difficult for morphological evaluation and impossible for quantitative analysis. Using UTE techniques, the region of the DVJ is seen with great contrast in vivo (►Fig. 8b). The CEP is seen as a continuous, linear, and high signal intensity line in normal cases; regions with the Schmorl node formation and bony end-plate irregularity exhibit loss of the normal characteristics (►Fig. 8b).

Conclusion
Ultrashort TE sequences provide a method to image tissues and tissue components that have not previously been visible using conventional clinical sequences. Continuing modifications to this technique have allowed for faster and more robust sequences that allow for morphological imaging and quantitative evaluation of short T2/T2 Ã tissues in vivo. The combination of vendor incorporation of UTE sequences into their update packages and continuing validation from research studies will lead to increasing acceptance and utilization by musculoskeletal radiologists.