Research article Received: 2 August 2014,
Revised: 3 October 2014,
Accepted: 17 November 2014,
Published online in Wiley Online Library: 18 December 2014
(wileyonlinelibrary.com) DOI: 10.1002/nbm.3246
ZTE imaging with long-T2 suppression Markus Weigera*, Mingming Wua,b, Moritz C. Wurnigc, David Kenkelc, Andreas Bossc, Gustav Andreisekc and Klaas P. Pruessmanna Three-dimensional radial zero echo time (ZTE) imaging enables efficient direct MRI of tissues with rapid transverse relaxation. Yet, the feature of capturing signals with a wide range of T2 and T2* values is accompanied by a lack of contrast between the corresponding tissues. In particular, the targeted short-T2 tissues may not be easily identified, and various approaches have been proposed to generate T2 contrast by reducing the long-T2 signal of water and/or fat. The aim of this work was to provide efficient long-T2 suppression for selective direct MRI of short-T2 tissues using the ZTE technique. For magnetization preparation, suppression pulses for water and fat were designed to provide both good T2 selectivity and off-resonance performance. To obtain high efficiency at short TRs, the pulses were applied in a segmented sequence scheme with minimized timing overhead, thus leading to a quasi-steady state of magnetization. The sequence timing was adjusted for optimal tissue contrast in musculoskeletal applications by means of simulations and experiments, incorporating both T2 and T1 of the involved tissues. The developed technique was employed for imaging of a lamb joint sample at 4.7 T. ZTE images were obtained with effective suppression of signals from tissues with long-T2 water, such as muscle or articular spaces, and fat. Hence, primarily short-T2 tissues were visible, such as bone and tendon. The MR image intensity of bone showed strong similarity with bone density imaged with micro-computed tomography. Copyright © 2014 John Wiley & Sons, Ltd. Keywords: bone; tendon; fat; short T2; zero TE; inversion pulse; signal recovery; CT
INTRODUCTION
NMR Biomed. 2015; 28: 247–254
* Correspondence to: M. Weiger, Institute for Biomedical Engineering, University and ETH Zurich, Gloriastrasse 35, CH-8092 Zurich, Switzerland. E-mail: [email protected] a M. Weiger, M. Wu, K. P. Pruessmann Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland b M. Wu Institute of Biomedical Engineering, Karlsruhe Institute of Technology, Karlsruhe, Germany c M. C. Wurnig, D. Kenkel, A. Boss, G. Andreisek Institute for Diagnostic and Interventional Radiology, University Hospital Zurich, Zurich, Switzerland Abbreviations used: 3D, three-dimensional; CT, computed tomography; RF, radiofrequency; SNR, signal-to-noise ratio; ZTE imaging, zero echo time imaging.
Copyright © 2014 John Wiley & Sons, Ltd.
247
In addition to the excellent capabilities of MR for the imaging of soft tissues, more recently direct MRI of more rigid samples and tissues with very rapid transverse relaxation has become possible with dedicated techniques offering ultra-short (1,2) or ZTE (3–7). These methods share the principle of using a highbandwidth radial center-out encoding scheme, but differ in their method of obtaining data in the center of k space with respect to the interplay of gradient application, radiofrequency (RF) excitation and initial dead time. In particular, the threedimensional (3D) zero echo time (ZTE) technique, using single short RF pulses with immediate rapid encoding, uninterrupted data acquisition with high duty cycle and very short TR (8,9), is able to capture short-T2 signals with high bandwidth, efficiency and robustness (10). Yet, the feature of the ZTE technique of detecting signals with a wide range of T2 and T2* values with equal sensitivity is accompanied by a lack of contrast between the corresponding tissues. In particular, the targeted short-T2 tissues may not be easily identified. Therefore, the aim of this work was to generate contrast in ZTE imaging by suppressing the long-T2 signal, thus enabling selective direct MRI of short-T2 tissues. Various approaches have been proposed to reduce the long-T2 signal of water and/or fat (11), including subtraction of images acquired at longer TE (12,13) or with short-T2 tissues being suppressed (14), long-T2 cancellation in balanced steady-state free precession (15) and magnetization preparation selectively targeting long-T2 tissues. The latter is based on either saturation (6,16–20) or inversion involving image combination (21) or signal recovery as part of the preparation (13).
In principle, all of these approaches are suitable to generate T2 contrast also in ZTE imaging. However, as the latter is a 3D technique with a very short TR on the order of 1 ms and below (22), this combination will be highly time consuming as a result of multiple image acquisitions, long inversion times or recovery delays, or the application of magnetization preparation before each excitation. Therefore, a more efficient magnetization-prepared approach with segmented acquisition was employed and optimized in this work. Suppression pulses for water and fat were designed to provide both good T2 selectivity and off-resonance performance. To obtain high efficiency at short TR, the pulses were applied in a segmented sequence scheme with minimized timing overhead, thus leading to a quasi-steady state of magnetization. The sequence timing was adjusted for optimal tissue contrast in
M. WEIGER ET AL. musculoskeletal applications by means of simulations and experiments. The developed technique was employed for ZTE imaging of a lamb joint sample at 4.7 T, and the results were compared with those of micro-computed tomography (micro-CT).
MATERIALS AND METHODS Sequence scheme Figure 1a shows the sequence scheme used for magnetizationprepared, segmented ZTE imaging with long-T2 suppression. The suppression pulse is followed by a gradient spoiler, an acquisition segment with a train of ZTE readouts for different
248
Figure 1. Zero echo time (ZTE) imaging with long-T2 suppression. (a) Sequence scheme. The suppression pulse of duration Tsup applied for water and/or fat suppression is followed by a spoiler gradient of length Tspoil, the ZTE acquisition block of duration Tseg, with Tseg/TR readouts for different gradient directions, and a recovery delay Trec. The scheme is repeated until all radial directions have been acquired. (b) Simulated course of longitudinal magnetization for the suppression of tissue with T1 = 300 ms. A suppression pulse is applied every 100 ms, using Tseg = 50 ms, Trec = 50 ms, TR = 1 ms and a flip angle of 10°. For a 90° saturation, the magnetization varies between zero and a positive value. Using a 180° inversion instead, after a transient situation, the magnetization alternates between negative and positive values of reduced maximum amplitude. Hence, the average amplitude is also smaller and, for neighboring radial directions, cancellation effects occur. Introducing the recovery delay results in a faster recovery of magnetization and in a shift in the zero-crossing within the acquisition window. An equivalent behavior is observed for any other T1 value, differing only in the absolute amplitudes and the time to reach quasi-steady state. It should be noted that, in this simulation, instant pulses were used and complete spoiling of transverse magnetization was assumed, making the result independent of T2.
wileyonlinelibrary.com/journal/nbm
radial directions and an optional recovery delay. The scheme is repeated until all required directions had been measured. The parameters to be determined for optimal contrast between short- and long-T2 tissues are the suppression pulse, segment duration, ZTE excitation flip angle and recovery delay. Finding these parameters was performed in two major steps. First, simulations for the magnetization were carried out by numerical integration of the Bloch equations using a fourth-order Runge–Kutta algorithm. Second, the sequence parameters found by simulation were used as initial values to further optimize the contrast empirically during the experiments. Suppression pulse With the scheme in Fig. 1a, magnetization is operated in a quasisteady state, interrupted by suppression and recovery events. Hence, not only T2, but also T1, plays an important role for the signal levels obtained for different tissues. Therefore, as a first step, the effect of the flip angle of the suppression pulse was investigated by means of simulations. Figure 1b shows the different behavior of magnetization for commonly employed saturation and, alternatively, for an inversion pulse. It can be seen that, in the segmented approach, inversion leads to a reduced signal amplitude. This behavior is not affected by T2 and is, in principle, also independent of T1 (cf. Fig. 1b). In contrast to saturation, reaching the quasi-steady state with inversion takes multiple segments, but requiring dummy scanning for only about T1. In addition to the decreased amplitude, partial signal cancellation can be expected for neighboring radial directions acquired with opposite signs in the magnetization, in particular in the highly oversampled k-space center. As a result of these observations, inversion pulses are favored in the present work. The duration, off-resonance behavior and T2 characteristics of RF pulses are closely linked (20,23). Whereas simple hard pulses provide a sinc-shaped frequency profile, filter-based pulse design using the Shinnar–Le Roux algorithm enables the choice of an improved, sharper frequency response (24,25). Such pulses have been employed for single- and dual-band saturation-based suppression (16). For the present work, this design (available from http://www.radiology.ucsf.edu/research/labs/larson/longt2supp) was adapted for inversion pulses. An effective inversion bandwidth of 400 Hz was selected after having observed the B0 inhomogeneities in the targeted samples at the field strength of 4.7 T used in this study. With this prerequisite, the residual parameters were optimized to provide a useful T2 selectivity without making the pulse too long to avoid T1 relaxation during the pulse and to limit the timing overhead. Using a duration of Tsup = 10 ms and maximum relative ripple amplitudes of 0.02 and 0.01 in the pass- and stop-band, respectively, the characteristics shown in Fig. 2 were obtained. The single-band pulse can be used for either water or fat suppression, whereas the dual-band pulse of the same duration enables simultaneous suppression at both frequencies. It can be observed that the fat suppression also affects water with short T2 because of its broad spectrum. Therefore, the dual-band pulse has a degraded T2 selectivity with respect to the water-only pulse. Sequence parameters The present work was targeted to musculoskeletal applications with the aim of directly depicting bone as demonstrated previously with ultra-short-TE MRI (26), as well as microscopic ZTE
Copyright © 2014 John Wiley & Sons, Ltd.
NMR Biomed. 2015; 28: 247–254
ZTE IMAGING WITH LONG-T2 SUPPRESSION
Figure 2. Off-resonance (a) and T2 characteristics (b) of the inversion pulses used for long-T2 suppression designed with duration Tsup = 10 ms for an effective inversion bandwidth of 400 Hz. The longitudinal magnetization MZ after one pulse is shown as obtained by the Bloch simulations. The wateronly suppression pulse shows a typical T2 selectivity. Fat suppression with an off-resonance of 700 Hz at 4.7 T also affects short-T2 water species with an extremum around 250 μs where the spectral overlap of pulse and T2 is most effective. Accordingly, with the dual-band pulse, increased suppression is obtained in the short-T2 range.
imaging for an extracted, dried sample (27). Therefore, the sequence parameters were optimized using Bloch simulations to obtain a positive contrast for bone with respect to muscle and fat, whilst keeping the increase in scan time within a reasonable range. The resulting signal, shown in Fig. 3, was calculated as the average of the transverse magnetization over one segment after the quasi steady-state had been reached (e.g. the last segment in Fig. 1b). Using Tspoil = 5 ms, a segment duration Tseg = 50 ms and no recovery delay, the scan time was increased by 30%. With this timing and a ZTE flip angle of 10°, it can be seen that the signal decreases not only with increasing T2, but
NMR Biomed. 2015; 28: 247–254
Experimental details Experiments were conducted on a 4.7-T PharmaScan animal MRI system (Bruker BioSpin MRI GmbH, Ettlingen, Germany) equipped with a linear volume resonator with an inner diameter of 60 mm and a gradient system providing a maximum amplitude of 375 mT/m and a slew rate of 3375 T/m/s. The scanner was operated with the imaging software ParaVision 5.1 (spectrometer). The ZTE sequence was modified for magnetization preparation in a segmented fashion as depicted in Fig. 1a, and the suppression pulses were implemented with the optimized parameters of Fig. 2. As a fresh tissue sample, the joint of a skinned lamb forefoot was used with a size of approximately 50 mm in each dimension. ZTE imaging was performed with 3D isotropic geometry in a field of view of 60 mm using an image matrix size of 128, thus resulting in a spatial resolution of 469 μm. A signal bandwidth of 200 kHz was chosen, thus the corresponding readout duration of 320 μs is matched to the approximately expected T2 value of bone (30,31), which enables to closely approach the nominal resolution (12). Furthermore, the associated pixel bandwidth of 2.2 times the water–fat shift of 700 Hz largely prevents off-resonance blurring of the fat signal. ZTE data were acquired with four-fold radial oversampling and a total dead time of 9.9 μs, corresponding to a central k-space gap of 2.0 nominal dwell times (8); 51 896 radial spokes, as required for full Nyquist encoding, were acquired using an order of successive directions with minimized angular differences (9). The TR was set to 1 ms and a hard pulse of 6 μs duration was used to generate flip angles of 3.2° or 9.6°. Gradient spoiling was employed by leaving the readout gradient on after the acquisition throughout TR, supported by the changing gradient directions. Either two or eight averages were acquired, resulting in scan times of 1 min 44 s or 6 min 56 s if no suppression and segmentation were employed, and 4 min 10 s or 16 min 38 s with the final protocol, using suppression and a recovery delay of Trec = 55 ms. The spoiling duration after a suppression pulse was set to Tspoil = 5 ms. For rapid empirical sequence optimization, the matrix size was reduced to 96 with a correspondingly reduced number of spokes and a shorter readout duration, and two averages were used, whereas all other
Copyright © 2014 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/nbm
249
Figure 3. Normalized transverse quasi-steady-state magnetization obtained using the zero echo time (ZTE) sequence scheme of Fig. 1a and the dual-band 180° suppression pulse of Fig. 2 with Tsup = 10 ms, Tspoil = 5 ms, Tseg = 50 ms, TR = 1 ms, flip angle of 10° and Trec = 0 ms. The signal level decreases with both increasing T2 and T1. Filled circles indicate the signal for bone (T1 = 200 ms, T2 = 0.3 ms) (28), fat (T1 = 400 ms, T2 = 50 ms) and muscle (T1 = 1300 ms, T2 = 30 ms) (29). Open circles show the corresponding signal for the sequence without contrast enhancement, i.e. with a flip angle of 3° and without suppression. The attached numbers provide the ratio of the two signal values, i.e. the relative residual signal. Hence, the bone signal is reduced by approximately one-half, whereas both fat and water are reduced below 15% of the original value, corresponding to an increased positive contrast for bone. It should be noted that differences in spin density between tissues are not included in this plot.
also for larger T1. For both reasons, the contrast for bone is clearly improved as the signal reduction for fat and muscle is considerably larger than that for bone.
M. WEIGER ET AL. parameters remained unchanged, resulting in a minimum scan time of 58 s. One-dimensional projections were reconstructed algebraically to address the central k-space gap (8,32), followed by windowing with a weak de-ringing filter (a hyperbolic secant shape with order 20 and truncation 5), 3D gridding and interpolation for image display. For comparison, additional 3D Cartesian gradient echo data were acquired with identical geometry and bandwidth as applied for the ZTE protocol, using TE = 1.43 ms for water and fat to be in phase, TR = 4 ms and a flip angle of 15°, resulting in a total scan time of 1 min 6 s. Before MRI, field map-based shimming was performed using a dualecho gradient echo sequence with identical parameters as above, but a matrix size of 64, TR = 20 ms, flip angle of 20° and a second TE = 5.72 ms, again obtaining water and fat in phase. Values of the available shims X, Y, Z and Z2 were set using least-squares minimization with the MapShim tool by taking into account the complete sample volume.
Micro-CT imaging The tissue sample was also imaged with micro-CT using a Skyscan 1176 system (Bruker-microCT, Kontich, Belgium) with the following scan parameters: tube voltage, 90 kV; tube current, 278 μA; exposure time, 65 ms; rotation step, 0.7°; covered angle, 360°; copper filter, 0.1 mm; isotropic voxel size, 35 μm; field of view, 70 × 70 × 65 mm3 covered by four overlapping scans; frame averaging, 3. The total scan duration was 19 min 33 s. The 3D image was reconstructed using a modified Feldkamp algorithm with the following parameters: beam hardening correction, 35%; ring artifact reduction, 6; Gaussian smoothing kernel with full width at half-maximum of four voxels. To reduce memory requirements, the CT data were down-sampled by a factor of four before manual co-registration with the MR images using the 3D Slicer software (www.slicer.org), whilst allowing for three rotational and translational degrees of freedom each.
Data analysis Signal values for different tissue types were determined in MR images by averaging over the selected regions of interest whilst normalizing for the number of acquisition averages. Contrast was calculated according to: C¼2
S Sref S þ Sref
[1]
where a signal value S of the tissue of interest is compared with Sref of a reference tissue. With this definition, C ranges from –2 to 2, C = 0 is obtained for equal signal intensity and C = 1 indicates a signal ratio of S/Sref = 3 : 1.
RESULTS Figure 4 illustrates the experimental optimization of the sequence parameters for contrast enhancement on the tissue sample. The ZTE image without magnetization preparation (Fig. 4a) is primarily proton density weighted and shows trabecular bone, muscle and fat. The result with optimized long-T2 suppression (Fig. 4b) shows mainly the bone, with signal from muscle and fat being largely suppressed. The only relevant difference with respect to the parameters used for the simulations in Fig. 3 is a recovery delay of Trec = 55 ms, thus leading to a correspondingly increased scan time. The importance of this parameter is demonstrated in Fig. 4e, f where no delay or a longer delay lead to strongly impaired results. In addition, a longer segment duration reduces the desired contrast (Fig. 4g), whereas a smaller flip angle clearly decreases the bone signal at similar contrast (Fig. 4c). The inferiority of applying saturation instead of inversion pulses for the suppression (cf. Fig. 1b) is demonstrated in Fig. 4d. Figures 5 and 6 show reformatted, orthogonal views of the final results with optimized timing, full MR resolution and different
250
Figure 4. Optimization of the zero echo time (ZTE) sequence parameters for bone imaging with long-T2 suppression, performed on a lamb joint at reduced spatial resolution. (a) The ZTE image without suppression and a small flip angle of 3.2° shows bone, muscle and fat with mainly proton density-weighted contrast. (b) The best contrast between bone and muscle, as well as fat, is obtained with a flip angle of 9.6°, 180° suppression pulses applied to both water and fat, a segment duration Tseg = 50 ms and recovery delay Trec = 55 ms. (c) A smaller flip angle of 3.2° leads to similar contrast, but at considerably reduced bone signal. (d) Using 90° saturation instead of inversion strongly reduces the suppression efficiency. (e) Without a recovery delay, the suppression of both water and fat is reduced considerably. (f) A longer recovery delay of 110 ms impairs fat suppression. (g) Using a longer segment duration of 100 ms also degrades the contrast. All images are displayed without windowing, showing their individual full range of intensities.
wileyonlinelibrary.com/journal/nbm
Copyright © 2014 John Wiley & Sons, Ltd.
NMR Biomed. 2015; 28: 247–254
ZTE IMAGING WITH LONG-T2 SUPPRESSION contrast stages in comparison with CT. The non-suppressed ZTE image with a small flip angle (Fig. 5a) exhibits primarily proton density contrast, providing water signal with long T2 from muscle and articular spaces, fat signal from marrow in medullary cavities and intertrabecular spaces, and short-T2 signal from both trabecular and cortical bone. The latter is void of signal in the gradient echo image (arrow, Fig. 5h). A larger excitation flip angle emphasizes the T1 contrast with a reduction in muscle signal (Fig. 5b). Suppression of fat only leaves the water signal from long-T2 tissue, such as muscle and cartilage, and short T2 in both types of bone (Fig. 5c), whereas application of only long-T2 water suppression results in a fat-dominated image (Fig. 5d). Suppression of both water and fat provides the desired contrast with the highest signal for bone (Fig. 5e). In this case, eight-fold averaging was used for sufficient signal-to-noise ratio (SNR) because of the lower baseline signal in bone. The image intensity closely resembles the density of bone structures visible in CT (Fig. 5f). It should be noted that registration between MR and CT data is slightly imperfect because of small deformations of the flexible joint between scans. In the field map (Fig. 5g), regions with considerable inhomogeneity coincide with insufficient fat suppression observable at the sample edges in Fig. 5c, e. Figure 6 shows comparable features to those in Fig. 5 for different orientations. In addition, tendons are observed with high positive contrast, which cannot be identified in the CT data with the chosen window settings as, in this tissue, short T2 is not accompanied by large absorption as in bone.
Results from the quantitative analysis of signal values determined without (cf. Fig. 5a) and with (cf. Fig. 5e) contrast enhancement are summarized in Fig. 7. As expected from Fig. 3, the signal in all tissues is decreased. However, for bone and tendon, 60% and 29% of the signal are maintained, whereas muscle and fat are reduced to about only 5% of the initial level. Normalization of the signal with respect to bone illustrates the dominance of muscle and fat without contrast manipulation, which is moved to bone and tendon when suppression is applied. This can also be seen directly from the contrast values of the short-T2 tissues, which are converted from negative or very low positive to strong positive contrast.
DISCUSSION In this work, MR images were created exhibiting primarily signal from tissues with short transverse relaxation. This was achieved by using ZTE imaging with magnetization preparation for long-T2 suppression in a quasi-steady-state scheme avoiding time-consuming full recovery of the longitudinal magnetization. By optimization of the suppression pulse and sequence parameters, the desired contrast was obtained with an increase in scan time of 140% compared with a continuous ZTE acquisition. With this approach, not only T2, but also T1, of the involved tissues was utilized to optimize contrast, thus benefitting from shorter T1 in bone and tendon relative to muscle and fat.
NMR Biomed. 2015; 28: 247–254
Copyright © 2014 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/nbm
251
Figure 5. Coronal zero echo time (ZTE) images with long-T2 suppression of water and fat, showing different stages of contrast generation. (a) A common ZTE acquisition using a small flip angle without suppression and segmentation leads to primarily proton density contrast. The indicated tissues are muscle, marrow fat in the medullary cavity, trabecular bone with marrow-filled spaces and compact, fat-free cortical bone. (b) A larger flip angle introduces increased T1 weighting, thus reducing primarily the muscle signal. (c) Suppression of fat only removes signal in large marrow-filled spaces and in trabecular structures. The highest intensity is obtained for long-T2 water in muscle and cartilage in the articular spaces, whereas cortical and trabecular bone exhibit a similar, lower signal level. (d) Suppression of long-T2 water only leads to dominating signal from fat. (e) Suppression of both water and fat results in an image exhibiting mainly bone with intensities closely agreeing with the density of bone structures as seen in the computed tomography (CT) image in (f). To obtain sufficient signal-to-noise ratio (SNR), eight-fold averaging was applied here as compared with two averages in all other ZTE images. (g) The field map shows B0 inhomogeneities up to ±500 Hz at the edges (arrows), resulting in insufficient fat suppression as observable in (c) and (e). (h) Negligible signal in the gradient echo image occurs for cortical bone (arrow). In contrast, the ZTE images show signal at these locations.
M. WEIGER ET AL.
Figure 6. Sagittal zero echo time (ZTE) images with long-T2 suppression of water and fat, in analogy with Fig. 5. In (e), a tendon is clearly depicted which exhibits low signal or low contrast with surrounding tissues in all other images. In particular, it cannot be identified with computed tomography (CT) at the chosen window settings, indicating low absorption coinciding with short T2. As an example of good correspondence between CT and ZTE with suppression, a location of high bone density is indicated (open arrow). Again, signal suppression is hampered at the edges with large B0 field deviations (see gray scale bar in Fig. 5).
One important aspect of contrast generation is the relatively large excitation flip angle of about 10° considering the short TR of 1 ms. Although this could be accomplished with a pulse duration of 6 μs and a B1 amplitude of 110 μT on an animal scanner, the application on human MR systems using hard pulses can be challenging with this requirement, thus favoring a lower base field strength and modulated excitation pulses (34). Other contrast-sensitive parameters are suppression flip angle and recovery delay. The inversion as proposed here causes an oscillation of the magnetization around zero at reduced amplitude, whereas the recovery delay determines the size of the initial part of the interval actually used for data acquisition (cf. Fig. 1b). This choice affects potential signal cancellation between nearby radial directions and was optimized empirically for a given order of directions in the current work (cf. Fig. 4). It is expected that a more detailed analysis of this aspect may lead to a more optimal combination of suppression angle, ordering and delay, and thus a further increase in suppression efficiency. Bloch simulations of the magnetization provided both insights into general principles as well as useful initial values for se-
quence parameters. However, for optimum performance, the latter still had to be adapted experimentally. Furthermore, although giving the correct qualitative behavior, the relative signal changes predicted in the simulations deviated from experimental observations. In particular, the effect of introducing the recovery delay could not be reproduced in the simulations. One likely reason for this discrepancy is the difference between the assumed and actual relaxation times in the sample. Furthermore, complete destruction of transverse magnetization was assumed before each new RF pulse in the simulation, which is not guaranteed in the ZTE sequence. Moreover, the discussed cancellation effects were addressed only partially. Therefore, a more elaborate simulation will be necessary, including these effects. Moreover, the properties of the sample used in this work are likely to differ from human musculoskeletal tissues in vivo, thus requiring a certain adaptation of the sequence parameters. The latter will also depend on the field strength and will change with the targeted short-T2 tissues. In this respect, another possible application would be direct MRI of short-T2 components in the brain, such as myelin, for example (13).
252
Figure 7. Relative signal and contrast values obtained from selected tissues in images acquired with zero echo time (ZTE) imaging without (Fig. 5a) and with (Fig. 5e) long-T2 suppression of water and fat. (a) The relative signal is the ratio of signal with and without suppression. (b) The normalized signal gives the image intensity with respect to bone, and is obtained by dividing by the signal value of bone. Contrast of cortical bone and tendon is calculated according to Equation [1] with respect to muscle (c) and fat (d).
wileyonlinelibrary.com/journal/nbm
Copyright © 2014 John Wiley & Sons, Ltd.
NMR Biomed. 2015; 28: 247–254
ZTE IMAGING WITH LONG-T2 SUPPRESSION Final long-T2 suppressed imaging required a four-fold increased averaging to achieve sufficient SNR for two reasons. First, the base magnetization of bone is smaller by a factor of three to four than that of long-T2 tissues because of the lower water proton density. Second, the short-T2 signal is also reduced by the suppression scheme according to the T2 characteristics of the applied inversion pulse (cf. Fig. 2b), leading to an experimental value of 60% of the initial signal for bone. To minimize off-resonance demands of the suppression pulse, good shimming is important. In the present study, field map-based shimming provided superior results to iteratively maximizing the integral of a free induction decay signal from a wholevolume spectroscopic acquisition. A further improvement may be achieved by minimizing the maximum norm instead of the quadratic norm, which better fits the box-shaped off-resonance behavior of the suppression pulses used (cf. Fig. 2a). In the present sample, the full observed frequency range was about ±500 Hz. However, the worst field deviations were present at the abrupt ends of the specimen caused by sample preparation. Hence, concerning shimming, the situation will be more benign in in vivo imaging. The problem associated with a wide frequency band of the suppression pulse is a degraded T2 characteristic. In this respect, the filter-based design employed offers improved properties relative to block-shaped, Gaussian-shaped (17,20) or adiabatic inversion pulses (21). In particular, the steep transition is advantageous for reliable fat suppression. A beneficial T2 selectivity can be achieved for decreased susceptibility-related off-resonance effects at lower field. However, the adverse effect of fat suppression on short-T2 water will be increased because of the smaller frequency shift between water and fat. Apart from its exceptional short-T2 capabilities, ZTE imaging is also a silent sequence (22) because of the possibility of only slightly adjusting the gradients for the next nearby radial direction. Silent MRI scanning has been identified recently as increasingly in demand and is pursued by multiple vendors. With the segmented scheme presented in Fig. 1a, acoustic noise is only slightly increased by the gradient spoiler and the requirement to ramp the gradient up and down once per segment, even at standard gradient operation. If required, the added clicking sounds can be eliminated without compromise by employing reduced slew rates for the associated gradients. The images obtained in this work provide a contrast which is opposite to that which radiologists have got used to over decades in MR. Concerning bone, a very close relationship can be visually observed between the MR signal intensity and density of bone tissue in CT data. Hence, the MR images may have the potential to directly determine bone mineral density without the use of ionizing radiation. However, more data and a quantitative analysis are required to support this hypothesis. In addition to bone, tendon could also be visualized, which may have the potential to detect injuries directly in such tissue. Other tissues in which the proposed technique may potentially improve diagnostics are ligaments, menisci, labrum or the bone–cartilage interface. In this view, the next important step will be translating the methodology to human applications, where the ZTE technique has already shown promising initial results (22,33).
Acknowledgements
NMR Biomed. 2015; 28: 247–254
1. Glover GH, Pauly JM, Bradshaw KM. B-11 imaging with a 3-dimensional reconstruction method. J. Magn. Reson. Imaging 1992; 2: 47–52. 2. Grodzki DM, Jakob PM, Heismann B. Ultrashort echo time imaging using pointwise encoding time reduction with radial acquisition (PETRA). Magn. Reson. Med. 2012; 67: 510–518. 3. Hafner S. Fast imaging in liquids and solids with the back-projection low-angle shot (BLAST) technique. Magn. Reson. Imaging 1994; 12: 1047–1051. 4. Madio DP, Lowe IJ. Ultra-fast imaging using low flip angles and FIDs. Magn. Reson. Med. 1995; 34: 525–529. 5. Kuethe DO, Caprihan A, Fukushima E, Waggoner RA. Imaging lungs using inert fluorinated gases. Magn. Reson. Med. 1998; 39: 85–88. 6. Wu YT, Ackerman JL, Chesler DA, Graham L, Wang Y, Glimcher MJ. Density of organic matrix of native mineralized bone measured by water- and fat-suppressed proton projection MRI. Magn. Reson. Med. 2003; 50: 59–68. 7. Idiyatullin D, Corum C, Park JY, Garwood M. Fast and quiet MRI using a swept radiofrequency. J. Magn. Reson. 2006; 181: 342–349. 8. Weiger M, Pruessmann KP. MRI with zero echo time. eMagRes 2012; 1: 311–322. 9. Weiger M, Pruessmann KP, Bracher AK, Kohler S, Lehmann V, Wolfram U, Hennel F, Rasche V. High-resolution ZTE imaging of human teeth. NMR Biomed. 2012; 25: 1144–1151. 10. Weiger M, Pruessmann KP, Hennel F. MRI with zero echo time: hard versus sweep pulse excitation. Magn. Reson. Med. 2011; 66: 379–389. 11. Szeverenyi NM, Carl M. Contrast manipulation in MR imaging of short T2 and T2* tissues. eMagRes 2012; 1: 581–586. 12. Rahmer J, Bornert P, Groen J, Bos C. Three-dimensional radial ultrashort echo-time imaging with T2 adapted sampling. Magn. Reson. Med. 2006; 55: 1075–1082. 13. Du J, Sheth V, Li S, Carl M, VandenBerg S, Corey-Bloom J, Bydder G. Morphological and quantitative ultrashort echo time (UTE) magnetic resonance imaging of the short T2 components in white matter of the brain. Proceedings of the 21st Annual Meeting ISMRM, Milan, Italy, 2014; 815. 14. Du J, Takahashi AM, Bydder M, Chung CB, Bydder GM. Ultrashort TE imaging with off-resonance saturation contrast (UTE-OSC). Magn. Reson. Med. 2009; 62: 527–531. 15. Peng Q. Long-T2 suppression with low flip-angle b-SSFP ultra-short TE (LA-bUTE). Proceedings of the 21st Annual Meeting ISMRM, Milan, Italy, 2014; 819. 16. Larson PEZ, Gurney PT, Nayak K, Gold GE, Pauly JM, Nishimura DG. Designing long-T2 suppression pulses for ultrashort echo time imaging. Magn. Reson. Med. 2006; 56: 94–103. 17. Wang L, Tang W, Zhen Z, Chen H, Xie J, Zhao Q. Improving detection specificity of iron oxide nanoparticles (IONPs) using the SWIFT sequence with long T(2) suppression. Magn. Reson. Imaging 2014; 32: 671–678. 18. Gold GE, Pauly JM, Macovski A, Herfkens RJ. MR spectroscopic imaging of collagen – tendons and knee menisci. Magn. Reson. Med. 1995; 34: 647–654. 19. Wang LN, Tang W, Zhen ZP, Chen HM, Xie J, Zhao Q. Improving detection specificity of iron oxide nanoparticles (IONPs) using the SWIFT sequence with long T2 suppression. Magn. Reson. Imaging 2014; 32: 671–678. 20. Rahmer J, Blume U, Bornert P. Selective 3D ultrashort TE imaging: comparison of “dual-echo” acquisition and magnetization preparation for improving short-T-2 contrast. Magn. Reson. Mater. Phys. Biol. Med. 2007; 20: 83–92. 21. Larson PEZ, Conolly SM, Pauly JM, Nishimura DG. Using adiabatic inversion pulses for long-T-2 suppression in ultrashort echo time (UTE) imaging. Magn. Reson. Med. 2007; 58: 952–961. 22. Weiger M, Brunner DO, Dietrich BE, Muller CF, Pruessmann KP. ZTE imaging in humans. Magn. Reson. Med. 2013; 70: 328–332. 23. Sussman MS, Pauly JM, Wright GA. Design of practical T2-selective RF excitation (TELEX) pulses. Magn. Reson. Med. 1998; 40: 890–899. 24. Shinnar M, Bolinger L, Leigh JS. The use of finite impulse-response filters in pulse design. Magn. Reson. Med. 1989; 12: 81–87. 25. Pauly J, Leroux P, Nishimura D, Macovski A. Parameter relations for the Shinnar–Leroux selective excitation pulse design algorithm. IEEE Trans. Med. Imaging 1991; 10: 53–65. 26. Du J, Bydder GM. Qualitative and quantitative ultrashort-TE MRI of cortical bone. NMR Biomed. 2013; 26: 489–506.
Copyright © 2014 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/nbm
253
This work was supported by Swiss National Science Foundation grants 310030_144075/1 and 3106030_139258/1.
REFERENCES
M. WEIGER ET AL. 27. Weiger M, Stampanoni M, Pruessmann KP. Direct depiction of bone microstructure using MRI with zero echo time. Bone 2013; 54: 44–47. 28. Du JA, Carl M, Bydder M, Takahashi A, Chung CB, Bydder GM. Qualitative and quantitative ultrashort echo time (UTE) imaging of cortical bone. J. Magn. Reson. 2010; 207: 304–311. 29. Jordan CD, Saranathan M, Bangerter NK, Hargreaves BA, Gold GE. Musculoskeletal MRI at 3.0 T and 7.0 T: a comparison of relaxation times and image contrast. Eur. J. Radiol. 2013; 82: 734–739. 30. Horch RA, Nyman JS, Gochberg DF, Dortch RD, Does MD. Character1 ization of H NMR signal in human cortical bone for magnetic resonance imaging. Magn. Reson. Med. 2010; 64: 680–687.
31. Rad HS, Lam SCB, Magland JF, Ong H, Li C, Song HK, Love J, Wehrli FW. Quantifying cortical bone water in vivo by three-dimensional ultra-short echo-time MRI. NMR Biomed. 2011; 24: 855–864. 32. Kuethe DO, Caprihan A, Lowe IJ, Madio DP, Gach HM. Transforming NMR data despite missing points. J. Magn. Reson. 1999; 139: 18–25. 33. Weiger M, Brunner DO, Wyss M, Dietrich BE, Wilm BJ, Pruessmann KP. ZTE imaging with T1 contrast. Proceedings of the 21st Annual Meeting ISMRM, Milan, Italy, 2014; 4262. 34. Schieban K, Weiger M, Hennel F, Boss A, Pruessman KP. ZTE imaging with enhanced flip angle using modulated excitation, Magn. Reson. Med. 2014. doi: 10.1002/mrm.25464.
254 wileyonlinelibrary.com/journal/nbm
Copyright © 2014 John Wiley & Sons, Ltd.
NMR Biomed. 2015; 28: 247–254
Revised: 3 October 2014,
Accepted: 17 November 2014,
Published online in Wiley Online Library: 18 December 2014
(wileyonlinelibrary.com) DOI: 10.1002/nbm.3246
ZTE imaging with long-T2 suppression Markus Weigera*, Mingming Wua,b, Moritz C. Wurnigc, David Kenkelc, Andreas Bossc, Gustav Andreisekc and Klaas P. Pruessmanna Three-dimensional radial zero echo time (ZTE) imaging enables efficient direct MRI of tissues with rapid transverse relaxation. Yet, the feature of capturing signals with a wide range of T2 and T2* values is accompanied by a lack of contrast between the corresponding tissues. In particular, the targeted short-T2 tissues may not be easily identified, and various approaches have been proposed to generate T2 contrast by reducing the long-T2 signal of water and/or fat. The aim of this work was to provide efficient long-T2 suppression for selective direct MRI of short-T2 tissues using the ZTE technique. For magnetization preparation, suppression pulses for water and fat were designed to provide both good T2 selectivity and off-resonance performance. To obtain high efficiency at short TRs, the pulses were applied in a segmented sequence scheme with minimized timing overhead, thus leading to a quasi-steady state of magnetization. The sequence timing was adjusted for optimal tissue contrast in musculoskeletal applications by means of simulations and experiments, incorporating both T2 and T1 of the involved tissues. The developed technique was employed for imaging of a lamb joint sample at 4.7 T. ZTE images were obtained with effective suppression of signals from tissues with long-T2 water, such as muscle or articular spaces, and fat. Hence, primarily short-T2 tissues were visible, such as bone and tendon. The MR image intensity of bone showed strong similarity with bone density imaged with micro-computed tomography. Copyright © 2014 John Wiley & Sons, Ltd. Keywords: bone; tendon; fat; short T2; zero TE; inversion pulse; signal recovery; CT
INTRODUCTION
NMR Biomed. 2015; 28: 247–254
* Correspondence to: M. Weiger, Institute for Biomedical Engineering, University and ETH Zurich, Gloriastrasse 35, CH-8092 Zurich, Switzerland. E-mail: [email protected] a M. Weiger, M. Wu, K. P. Pruessmann Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland b M. Wu Institute of Biomedical Engineering, Karlsruhe Institute of Technology, Karlsruhe, Germany c M. C. Wurnig, D. Kenkel, A. Boss, G. Andreisek Institute for Diagnostic and Interventional Radiology, University Hospital Zurich, Zurich, Switzerland Abbreviations used: 3D, three-dimensional; CT, computed tomography; RF, radiofrequency; SNR, signal-to-noise ratio; ZTE imaging, zero echo time imaging.
Copyright © 2014 John Wiley & Sons, Ltd.
247
In addition to the excellent capabilities of MR for the imaging of soft tissues, more recently direct MRI of more rigid samples and tissues with very rapid transverse relaxation has become possible with dedicated techniques offering ultra-short (1,2) or ZTE (3–7). These methods share the principle of using a highbandwidth radial center-out encoding scheme, but differ in their method of obtaining data in the center of k space with respect to the interplay of gradient application, radiofrequency (RF) excitation and initial dead time. In particular, the threedimensional (3D) zero echo time (ZTE) technique, using single short RF pulses with immediate rapid encoding, uninterrupted data acquisition with high duty cycle and very short TR (8,9), is able to capture short-T2 signals with high bandwidth, efficiency and robustness (10). Yet, the feature of the ZTE technique of detecting signals with a wide range of T2 and T2* values with equal sensitivity is accompanied by a lack of contrast between the corresponding tissues. In particular, the targeted short-T2 tissues may not be easily identified. Therefore, the aim of this work was to generate contrast in ZTE imaging by suppressing the long-T2 signal, thus enabling selective direct MRI of short-T2 tissues. Various approaches have been proposed to reduce the long-T2 signal of water and/or fat (11), including subtraction of images acquired at longer TE (12,13) or with short-T2 tissues being suppressed (14), long-T2 cancellation in balanced steady-state free precession (15) and magnetization preparation selectively targeting long-T2 tissues. The latter is based on either saturation (6,16–20) or inversion involving image combination (21) or signal recovery as part of the preparation (13).
In principle, all of these approaches are suitable to generate T2 contrast also in ZTE imaging. However, as the latter is a 3D technique with a very short TR on the order of 1 ms and below (22), this combination will be highly time consuming as a result of multiple image acquisitions, long inversion times or recovery delays, or the application of magnetization preparation before each excitation. Therefore, a more efficient magnetization-prepared approach with segmented acquisition was employed and optimized in this work. Suppression pulses for water and fat were designed to provide both good T2 selectivity and off-resonance performance. To obtain high efficiency at short TR, the pulses were applied in a segmented sequence scheme with minimized timing overhead, thus leading to a quasi-steady state of magnetization. The sequence timing was adjusted for optimal tissue contrast in
M. WEIGER ET AL. musculoskeletal applications by means of simulations and experiments. The developed technique was employed for ZTE imaging of a lamb joint sample at 4.7 T, and the results were compared with those of micro-computed tomography (micro-CT).
MATERIALS AND METHODS Sequence scheme Figure 1a shows the sequence scheme used for magnetizationprepared, segmented ZTE imaging with long-T2 suppression. The suppression pulse is followed by a gradient spoiler, an acquisition segment with a train of ZTE readouts for different
248
Figure 1. Zero echo time (ZTE) imaging with long-T2 suppression. (a) Sequence scheme. The suppression pulse of duration Tsup applied for water and/or fat suppression is followed by a spoiler gradient of length Tspoil, the ZTE acquisition block of duration Tseg, with Tseg/TR readouts for different gradient directions, and a recovery delay Trec. The scheme is repeated until all radial directions have been acquired. (b) Simulated course of longitudinal magnetization for the suppression of tissue with T1 = 300 ms. A suppression pulse is applied every 100 ms, using Tseg = 50 ms, Trec = 50 ms, TR = 1 ms and a flip angle of 10°. For a 90° saturation, the magnetization varies between zero and a positive value. Using a 180° inversion instead, after a transient situation, the magnetization alternates between negative and positive values of reduced maximum amplitude. Hence, the average amplitude is also smaller and, for neighboring radial directions, cancellation effects occur. Introducing the recovery delay results in a faster recovery of magnetization and in a shift in the zero-crossing within the acquisition window. An equivalent behavior is observed for any other T1 value, differing only in the absolute amplitudes and the time to reach quasi-steady state. It should be noted that, in this simulation, instant pulses were used and complete spoiling of transverse magnetization was assumed, making the result independent of T2.
wileyonlinelibrary.com/journal/nbm
radial directions and an optional recovery delay. The scheme is repeated until all required directions had been measured. The parameters to be determined for optimal contrast between short- and long-T2 tissues are the suppression pulse, segment duration, ZTE excitation flip angle and recovery delay. Finding these parameters was performed in two major steps. First, simulations for the magnetization were carried out by numerical integration of the Bloch equations using a fourth-order Runge–Kutta algorithm. Second, the sequence parameters found by simulation were used as initial values to further optimize the contrast empirically during the experiments. Suppression pulse With the scheme in Fig. 1a, magnetization is operated in a quasisteady state, interrupted by suppression and recovery events. Hence, not only T2, but also T1, plays an important role for the signal levels obtained for different tissues. Therefore, as a first step, the effect of the flip angle of the suppression pulse was investigated by means of simulations. Figure 1b shows the different behavior of magnetization for commonly employed saturation and, alternatively, for an inversion pulse. It can be seen that, in the segmented approach, inversion leads to a reduced signal amplitude. This behavior is not affected by T2 and is, in principle, also independent of T1 (cf. Fig. 1b). In contrast to saturation, reaching the quasi-steady state with inversion takes multiple segments, but requiring dummy scanning for only about T1. In addition to the decreased amplitude, partial signal cancellation can be expected for neighboring radial directions acquired with opposite signs in the magnetization, in particular in the highly oversampled k-space center. As a result of these observations, inversion pulses are favored in the present work. The duration, off-resonance behavior and T2 characteristics of RF pulses are closely linked (20,23). Whereas simple hard pulses provide a sinc-shaped frequency profile, filter-based pulse design using the Shinnar–Le Roux algorithm enables the choice of an improved, sharper frequency response (24,25). Such pulses have been employed for single- and dual-band saturation-based suppression (16). For the present work, this design (available from http://www.radiology.ucsf.edu/research/labs/larson/longt2supp) was adapted for inversion pulses. An effective inversion bandwidth of 400 Hz was selected after having observed the B0 inhomogeneities in the targeted samples at the field strength of 4.7 T used in this study. With this prerequisite, the residual parameters were optimized to provide a useful T2 selectivity without making the pulse too long to avoid T1 relaxation during the pulse and to limit the timing overhead. Using a duration of Tsup = 10 ms and maximum relative ripple amplitudes of 0.02 and 0.01 in the pass- and stop-band, respectively, the characteristics shown in Fig. 2 were obtained. The single-band pulse can be used for either water or fat suppression, whereas the dual-band pulse of the same duration enables simultaneous suppression at both frequencies. It can be observed that the fat suppression also affects water with short T2 because of its broad spectrum. Therefore, the dual-band pulse has a degraded T2 selectivity with respect to the water-only pulse. Sequence parameters The present work was targeted to musculoskeletal applications with the aim of directly depicting bone as demonstrated previously with ultra-short-TE MRI (26), as well as microscopic ZTE
Copyright © 2014 John Wiley & Sons, Ltd.
NMR Biomed. 2015; 28: 247–254
ZTE IMAGING WITH LONG-T2 SUPPRESSION
Figure 2. Off-resonance (a) and T2 characteristics (b) of the inversion pulses used for long-T2 suppression designed with duration Tsup = 10 ms for an effective inversion bandwidth of 400 Hz. The longitudinal magnetization MZ after one pulse is shown as obtained by the Bloch simulations. The wateronly suppression pulse shows a typical T2 selectivity. Fat suppression with an off-resonance of 700 Hz at 4.7 T also affects short-T2 water species with an extremum around 250 μs where the spectral overlap of pulse and T2 is most effective. Accordingly, with the dual-band pulse, increased suppression is obtained in the short-T2 range.
imaging for an extracted, dried sample (27). Therefore, the sequence parameters were optimized using Bloch simulations to obtain a positive contrast for bone with respect to muscle and fat, whilst keeping the increase in scan time within a reasonable range. The resulting signal, shown in Fig. 3, was calculated as the average of the transverse magnetization over one segment after the quasi steady-state had been reached (e.g. the last segment in Fig. 1b). Using Tspoil = 5 ms, a segment duration Tseg = 50 ms and no recovery delay, the scan time was increased by 30%. With this timing and a ZTE flip angle of 10°, it can be seen that the signal decreases not only with increasing T2, but
NMR Biomed. 2015; 28: 247–254
Experimental details Experiments were conducted on a 4.7-T PharmaScan animal MRI system (Bruker BioSpin MRI GmbH, Ettlingen, Germany) equipped with a linear volume resonator with an inner diameter of 60 mm and a gradient system providing a maximum amplitude of 375 mT/m and a slew rate of 3375 T/m/s. The scanner was operated with the imaging software ParaVision 5.1 (spectrometer). The ZTE sequence was modified for magnetization preparation in a segmented fashion as depicted in Fig. 1a, and the suppression pulses were implemented with the optimized parameters of Fig. 2. As a fresh tissue sample, the joint of a skinned lamb forefoot was used with a size of approximately 50 mm in each dimension. ZTE imaging was performed with 3D isotropic geometry in a field of view of 60 mm using an image matrix size of 128, thus resulting in a spatial resolution of 469 μm. A signal bandwidth of 200 kHz was chosen, thus the corresponding readout duration of 320 μs is matched to the approximately expected T2 value of bone (30,31), which enables to closely approach the nominal resolution (12). Furthermore, the associated pixel bandwidth of 2.2 times the water–fat shift of 700 Hz largely prevents off-resonance blurring of the fat signal. ZTE data were acquired with four-fold radial oversampling and a total dead time of 9.9 μs, corresponding to a central k-space gap of 2.0 nominal dwell times (8); 51 896 radial spokes, as required for full Nyquist encoding, were acquired using an order of successive directions with minimized angular differences (9). The TR was set to 1 ms and a hard pulse of 6 μs duration was used to generate flip angles of 3.2° or 9.6°. Gradient spoiling was employed by leaving the readout gradient on after the acquisition throughout TR, supported by the changing gradient directions. Either two or eight averages were acquired, resulting in scan times of 1 min 44 s or 6 min 56 s if no suppression and segmentation were employed, and 4 min 10 s or 16 min 38 s with the final protocol, using suppression and a recovery delay of Trec = 55 ms. The spoiling duration after a suppression pulse was set to Tspoil = 5 ms. For rapid empirical sequence optimization, the matrix size was reduced to 96 with a correspondingly reduced number of spokes and a shorter readout duration, and two averages were used, whereas all other
Copyright © 2014 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/nbm
249
Figure 3. Normalized transverse quasi-steady-state magnetization obtained using the zero echo time (ZTE) sequence scheme of Fig. 1a and the dual-band 180° suppression pulse of Fig. 2 with Tsup = 10 ms, Tspoil = 5 ms, Tseg = 50 ms, TR = 1 ms, flip angle of 10° and Trec = 0 ms. The signal level decreases with both increasing T2 and T1. Filled circles indicate the signal for bone (T1 = 200 ms, T2 = 0.3 ms) (28), fat (T1 = 400 ms, T2 = 50 ms) and muscle (T1 = 1300 ms, T2 = 30 ms) (29). Open circles show the corresponding signal for the sequence without contrast enhancement, i.e. with a flip angle of 3° and without suppression. The attached numbers provide the ratio of the two signal values, i.e. the relative residual signal. Hence, the bone signal is reduced by approximately one-half, whereas both fat and water are reduced below 15% of the original value, corresponding to an increased positive contrast for bone. It should be noted that differences in spin density between tissues are not included in this plot.
also for larger T1. For both reasons, the contrast for bone is clearly improved as the signal reduction for fat and muscle is considerably larger than that for bone.
M. WEIGER ET AL. parameters remained unchanged, resulting in a minimum scan time of 58 s. One-dimensional projections were reconstructed algebraically to address the central k-space gap (8,32), followed by windowing with a weak de-ringing filter (a hyperbolic secant shape with order 20 and truncation 5), 3D gridding and interpolation for image display. For comparison, additional 3D Cartesian gradient echo data were acquired with identical geometry and bandwidth as applied for the ZTE protocol, using TE = 1.43 ms for water and fat to be in phase, TR = 4 ms and a flip angle of 15°, resulting in a total scan time of 1 min 6 s. Before MRI, field map-based shimming was performed using a dualecho gradient echo sequence with identical parameters as above, but a matrix size of 64, TR = 20 ms, flip angle of 20° and a second TE = 5.72 ms, again obtaining water and fat in phase. Values of the available shims X, Y, Z and Z2 were set using least-squares minimization with the MapShim tool by taking into account the complete sample volume.
Micro-CT imaging The tissue sample was also imaged with micro-CT using a Skyscan 1176 system (Bruker-microCT, Kontich, Belgium) with the following scan parameters: tube voltage, 90 kV; tube current, 278 μA; exposure time, 65 ms; rotation step, 0.7°; covered angle, 360°; copper filter, 0.1 mm; isotropic voxel size, 35 μm; field of view, 70 × 70 × 65 mm3 covered by four overlapping scans; frame averaging, 3. The total scan duration was 19 min 33 s. The 3D image was reconstructed using a modified Feldkamp algorithm with the following parameters: beam hardening correction, 35%; ring artifact reduction, 6; Gaussian smoothing kernel with full width at half-maximum of four voxels. To reduce memory requirements, the CT data were down-sampled by a factor of four before manual co-registration with the MR images using the 3D Slicer software (www.slicer.org), whilst allowing for three rotational and translational degrees of freedom each.
Data analysis Signal values for different tissue types were determined in MR images by averaging over the selected regions of interest whilst normalizing for the number of acquisition averages. Contrast was calculated according to: C¼2
S Sref S þ Sref
[1]
where a signal value S of the tissue of interest is compared with Sref of a reference tissue. With this definition, C ranges from –2 to 2, C = 0 is obtained for equal signal intensity and C = 1 indicates a signal ratio of S/Sref = 3 : 1.
RESULTS Figure 4 illustrates the experimental optimization of the sequence parameters for contrast enhancement on the tissue sample. The ZTE image without magnetization preparation (Fig. 4a) is primarily proton density weighted and shows trabecular bone, muscle and fat. The result with optimized long-T2 suppression (Fig. 4b) shows mainly the bone, with signal from muscle and fat being largely suppressed. The only relevant difference with respect to the parameters used for the simulations in Fig. 3 is a recovery delay of Trec = 55 ms, thus leading to a correspondingly increased scan time. The importance of this parameter is demonstrated in Fig. 4e, f where no delay or a longer delay lead to strongly impaired results. In addition, a longer segment duration reduces the desired contrast (Fig. 4g), whereas a smaller flip angle clearly decreases the bone signal at similar contrast (Fig. 4c). The inferiority of applying saturation instead of inversion pulses for the suppression (cf. Fig. 1b) is demonstrated in Fig. 4d. Figures 5 and 6 show reformatted, orthogonal views of the final results with optimized timing, full MR resolution and different
250
Figure 4. Optimization of the zero echo time (ZTE) sequence parameters for bone imaging with long-T2 suppression, performed on a lamb joint at reduced spatial resolution. (a) The ZTE image without suppression and a small flip angle of 3.2° shows bone, muscle and fat with mainly proton density-weighted contrast. (b) The best contrast between bone and muscle, as well as fat, is obtained with a flip angle of 9.6°, 180° suppression pulses applied to both water and fat, a segment duration Tseg = 50 ms and recovery delay Trec = 55 ms. (c) A smaller flip angle of 3.2° leads to similar contrast, but at considerably reduced bone signal. (d) Using 90° saturation instead of inversion strongly reduces the suppression efficiency. (e) Without a recovery delay, the suppression of both water and fat is reduced considerably. (f) A longer recovery delay of 110 ms impairs fat suppression. (g) Using a longer segment duration of 100 ms also degrades the contrast. All images are displayed without windowing, showing their individual full range of intensities.
wileyonlinelibrary.com/journal/nbm
Copyright © 2014 John Wiley & Sons, Ltd.
NMR Biomed. 2015; 28: 247–254
ZTE IMAGING WITH LONG-T2 SUPPRESSION contrast stages in comparison with CT. The non-suppressed ZTE image with a small flip angle (Fig. 5a) exhibits primarily proton density contrast, providing water signal with long T2 from muscle and articular spaces, fat signal from marrow in medullary cavities and intertrabecular spaces, and short-T2 signal from both trabecular and cortical bone. The latter is void of signal in the gradient echo image (arrow, Fig. 5h). A larger excitation flip angle emphasizes the T1 contrast with a reduction in muscle signal (Fig. 5b). Suppression of fat only leaves the water signal from long-T2 tissue, such as muscle and cartilage, and short T2 in both types of bone (Fig. 5c), whereas application of only long-T2 water suppression results in a fat-dominated image (Fig. 5d). Suppression of both water and fat provides the desired contrast with the highest signal for bone (Fig. 5e). In this case, eight-fold averaging was used for sufficient signal-to-noise ratio (SNR) because of the lower baseline signal in bone. The image intensity closely resembles the density of bone structures visible in CT (Fig. 5f). It should be noted that registration between MR and CT data is slightly imperfect because of small deformations of the flexible joint between scans. In the field map (Fig. 5g), regions with considerable inhomogeneity coincide with insufficient fat suppression observable at the sample edges in Fig. 5c, e. Figure 6 shows comparable features to those in Fig. 5 for different orientations. In addition, tendons are observed with high positive contrast, which cannot be identified in the CT data with the chosen window settings as, in this tissue, short T2 is not accompanied by large absorption as in bone.
Results from the quantitative analysis of signal values determined without (cf. Fig. 5a) and with (cf. Fig. 5e) contrast enhancement are summarized in Fig. 7. As expected from Fig. 3, the signal in all tissues is decreased. However, for bone and tendon, 60% and 29% of the signal are maintained, whereas muscle and fat are reduced to about only 5% of the initial level. Normalization of the signal with respect to bone illustrates the dominance of muscle and fat without contrast manipulation, which is moved to bone and tendon when suppression is applied. This can also be seen directly from the contrast values of the short-T2 tissues, which are converted from negative or very low positive to strong positive contrast.
DISCUSSION In this work, MR images were created exhibiting primarily signal from tissues with short transverse relaxation. This was achieved by using ZTE imaging with magnetization preparation for long-T2 suppression in a quasi-steady-state scheme avoiding time-consuming full recovery of the longitudinal magnetization. By optimization of the suppression pulse and sequence parameters, the desired contrast was obtained with an increase in scan time of 140% compared with a continuous ZTE acquisition. With this approach, not only T2, but also T1, of the involved tissues was utilized to optimize contrast, thus benefitting from shorter T1 in bone and tendon relative to muscle and fat.
NMR Biomed. 2015; 28: 247–254
Copyright © 2014 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/nbm
251
Figure 5. Coronal zero echo time (ZTE) images with long-T2 suppression of water and fat, showing different stages of contrast generation. (a) A common ZTE acquisition using a small flip angle without suppression and segmentation leads to primarily proton density contrast. The indicated tissues are muscle, marrow fat in the medullary cavity, trabecular bone with marrow-filled spaces and compact, fat-free cortical bone. (b) A larger flip angle introduces increased T1 weighting, thus reducing primarily the muscle signal. (c) Suppression of fat only removes signal in large marrow-filled spaces and in trabecular structures. The highest intensity is obtained for long-T2 water in muscle and cartilage in the articular spaces, whereas cortical and trabecular bone exhibit a similar, lower signal level. (d) Suppression of long-T2 water only leads to dominating signal from fat. (e) Suppression of both water and fat results in an image exhibiting mainly bone with intensities closely agreeing with the density of bone structures as seen in the computed tomography (CT) image in (f). To obtain sufficient signal-to-noise ratio (SNR), eight-fold averaging was applied here as compared with two averages in all other ZTE images. (g) The field map shows B0 inhomogeneities up to ±500 Hz at the edges (arrows), resulting in insufficient fat suppression as observable in (c) and (e). (h) Negligible signal in the gradient echo image occurs for cortical bone (arrow). In contrast, the ZTE images show signal at these locations.
M. WEIGER ET AL.
Figure 6. Sagittal zero echo time (ZTE) images with long-T2 suppression of water and fat, in analogy with Fig. 5. In (e), a tendon is clearly depicted which exhibits low signal or low contrast with surrounding tissues in all other images. In particular, it cannot be identified with computed tomography (CT) at the chosen window settings, indicating low absorption coinciding with short T2. As an example of good correspondence between CT and ZTE with suppression, a location of high bone density is indicated (open arrow). Again, signal suppression is hampered at the edges with large B0 field deviations (see gray scale bar in Fig. 5).
One important aspect of contrast generation is the relatively large excitation flip angle of about 10° considering the short TR of 1 ms. Although this could be accomplished with a pulse duration of 6 μs and a B1 amplitude of 110 μT on an animal scanner, the application on human MR systems using hard pulses can be challenging with this requirement, thus favoring a lower base field strength and modulated excitation pulses (34). Other contrast-sensitive parameters are suppression flip angle and recovery delay. The inversion as proposed here causes an oscillation of the magnetization around zero at reduced amplitude, whereas the recovery delay determines the size of the initial part of the interval actually used for data acquisition (cf. Fig. 1b). This choice affects potential signal cancellation between nearby radial directions and was optimized empirically for a given order of directions in the current work (cf. Fig. 4). It is expected that a more detailed analysis of this aspect may lead to a more optimal combination of suppression angle, ordering and delay, and thus a further increase in suppression efficiency. Bloch simulations of the magnetization provided both insights into general principles as well as useful initial values for se-
quence parameters. However, for optimum performance, the latter still had to be adapted experimentally. Furthermore, although giving the correct qualitative behavior, the relative signal changes predicted in the simulations deviated from experimental observations. In particular, the effect of introducing the recovery delay could not be reproduced in the simulations. One likely reason for this discrepancy is the difference between the assumed and actual relaxation times in the sample. Furthermore, complete destruction of transverse magnetization was assumed before each new RF pulse in the simulation, which is not guaranteed in the ZTE sequence. Moreover, the discussed cancellation effects were addressed only partially. Therefore, a more elaborate simulation will be necessary, including these effects. Moreover, the properties of the sample used in this work are likely to differ from human musculoskeletal tissues in vivo, thus requiring a certain adaptation of the sequence parameters. The latter will also depend on the field strength and will change with the targeted short-T2 tissues. In this respect, another possible application would be direct MRI of short-T2 components in the brain, such as myelin, for example (13).
252
Figure 7. Relative signal and contrast values obtained from selected tissues in images acquired with zero echo time (ZTE) imaging without (Fig. 5a) and with (Fig. 5e) long-T2 suppression of water and fat. (a) The relative signal is the ratio of signal with and without suppression. (b) The normalized signal gives the image intensity with respect to bone, and is obtained by dividing by the signal value of bone. Contrast of cortical bone and tendon is calculated according to Equation [1] with respect to muscle (c) and fat (d).
wileyonlinelibrary.com/journal/nbm
Copyright © 2014 John Wiley & Sons, Ltd.
NMR Biomed. 2015; 28: 247–254
ZTE IMAGING WITH LONG-T2 SUPPRESSION Final long-T2 suppressed imaging required a four-fold increased averaging to achieve sufficient SNR for two reasons. First, the base magnetization of bone is smaller by a factor of three to four than that of long-T2 tissues because of the lower water proton density. Second, the short-T2 signal is also reduced by the suppression scheme according to the T2 characteristics of the applied inversion pulse (cf. Fig. 2b), leading to an experimental value of 60% of the initial signal for bone. To minimize off-resonance demands of the suppression pulse, good shimming is important. In the present study, field map-based shimming provided superior results to iteratively maximizing the integral of a free induction decay signal from a wholevolume spectroscopic acquisition. A further improvement may be achieved by minimizing the maximum norm instead of the quadratic norm, which better fits the box-shaped off-resonance behavior of the suppression pulses used (cf. Fig. 2a). In the present sample, the full observed frequency range was about ±500 Hz. However, the worst field deviations were present at the abrupt ends of the specimen caused by sample preparation. Hence, concerning shimming, the situation will be more benign in in vivo imaging. The problem associated with a wide frequency band of the suppression pulse is a degraded T2 characteristic. In this respect, the filter-based design employed offers improved properties relative to block-shaped, Gaussian-shaped (17,20) or adiabatic inversion pulses (21). In particular, the steep transition is advantageous for reliable fat suppression. A beneficial T2 selectivity can be achieved for decreased susceptibility-related off-resonance effects at lower field. However, the adverse effect of fat suppression on short-T2 water will be increased because of the smaller frequency shift between water and fat. Apart from its exceptional short-T2 capabilities, ZTE imaging is also a silent sequence (22) because of the possibility of only slightly adjusting the gradients for the next nearby radial direction. Silent MRI scanning has been identified recently as increasingly in demand and is pursued by multiple vendors. With the segmented scheme presented in Fig. 1a, acoustic noise is only slightly increased by the gradient spoiler and the requirement to ramp the gradient up and down once per segment, even at standard gradient operation. If required, the added clicking sounds can be eliminated without compromise by employing reduced slew rates for the associated gradients. The images obtained in this work provide a contrast which is opposite to that which radiologists have got used to over decades in MR. Concerning bone, a very close relationship can be visually observed between the MR signal intensity and density of bone tissue in CT data. Hence, the MR images may have the potential to directly determine bone mineral density without the use of ionizing radiation. However, more data and a quantitative analysis are required to support this hypothesis. In addition to bone, tendon could also be visualized, which may have the potential to detect injuries directly in such tissue. Other tissues in which the proposed technique may potentially improve diagnostics are ligaments, menisci, labrum or the bone–cartilage interface. In this view, the next important step will be translating the methodology to human applications, where the ZTE technique has already shown promising initial results (22,33).
Acknowledgements
NMR Biomed. 2015; 28: 247–254
1. Glover GH, Pauly JM, Bradshaw KM. B-11 imaging with a 3-dimensional reconstruction method. J. Magn. Reson. Imaging 1992; 2: 47–52. 2. Grodzki DM, Jakob PM, Heismann B. Ultrashort echo time imaging using pointwise encoding time reduction with radial acquisition (PETRA). Magn. Reson. Med. 2012; 67: 510–518. 3. Hafner S. Fast imaging in liquids and solids with the back-projection low-angle shot (BLAST) technique. Magn. Reson. Imaging 1994; 12: 1047–1051. 4. Madio DP, Lowe IJ. Ultra-fast imaging using low flip angles and FIDs. Magn. Reson. Med. 1995; 34: 525–529. 5. Kuethe DO, Caprihan A, Fukushima E, Waggoner RA. Imaging lungs using inert fluorinated gases. Magn. Reson. Med. 1998; 39: 85–88. 6. Wu YT, Ackerman JL, Chesler DA, Graham L, Wang Y, Glimcher MJ. Density of organic matrix of native mineralized bone measured by water- and fat-suppressed proton projection MRI. Magn. Reson. Med. 2003; 50: 59–68. 7. Idiyatullin D, Corum C, Park JY, Garwood M. Fast and quiet MRI using a swept radiofrequency. J. Magn. Reson. 2006; 181: 342–349. 8. Weiger M, Pruessmann KP. MRI with zero echo time. eMagRes 2012; 1: 311–322. 9. Weiger M, Pruessmann KP, Bracher AK, Kohler S, Lehmann V, Wolfram U, Hennel F, Rasche V. High-resolution ZTE imaging of human teeth. NMR Biomed. 2012; 25: 1144–1151. 10. Weiger M, Pruessmann KP, Hennel F. MRI with zero echo time: hard versus sweep pulse excitation. Magn. Reson. Med. 2011; 66: 379–389. 11. Szeverenyi NM, Carl M. Contrast manipulation in MR imaging of short T2 and T2* tissues. eMagRes 2012; 1: 581–586. 12. Rahmer J, Bornert P, Groen J, Bos C. Three-dimensional radial ultrashort echo-time imaging with T2 adapted sampling. Magn. Reson. Med. 2006; 55: 1075–1082. 13. Du J, Sheth V, Li S, Carl M, VandenBerg S, Corey-Bloom J, Bydder G. Morphological and quantitative ultrashort echo time (UTE) magnetic resonance imaging of the short T2 components in white matter of the brain. Proceedings of the 21st Annual Meeting ISMRM, Milan, Italy, 2014; 815. 14. Du J, Takahashi AM, Bydder M, Chung CB, Bydder GM. Ultrashort TE imaging with off-resonance saturation contrast (UTE-OSC). Magn. Reson. Med. 2009; 62: 527–531. 15. Peng Q. Long-T2 suppression with low flip-angle b-SSFP ultra-short TE (LA-bUTE). Proceedings of the 21st Annual Meeting ISMRM, Milan, Italy, 2014; 819. 16. Larson PEZ, Gurney PT, Nayak K, Gold GE, Pauly JM, Nishimura DG. Designing long-T2 suppression pulses for ultrashort echo time imaging. Magn. Reson. Med. 2006; 56: 94–103. 17. Wang L, Tang W, Zhen Z, Chen H, Xie J, Zhao Q. Improving detection specificity of iron oxide nanoparticles (IONPs) using the SWIFT sequence with long T(2) suppression. Magn. Reson. Imaging 2014; 32: 671–678. 18. Gold GE, Pauly JM, Macovski A, Herfkens RJ. MR spectroscopic imaging of collagen – tendons and knee menisci. Magn. Reson. Med. 1995; 34: 647–654. 19. Wang LN, Tang W, Zhen ZP, Chen HM, Xie J, Zhao Q. Improving detection specificity of iron oxide nanoparticles (IONPs) using the SWIFT sequence with long T2 suppression. Magn. Reson. Imaging 2014; 32: 671–678. 20. Rahmer J, Blume U, Bornert P. Selective 3D ultrashort TE imaging: comparison of “dual-echo” acquisition and magnetization preparation for improving short-T-2 contrast. Magn. Reson. Mater. Phys. Biol. Med. 2007; 20: 83–92. 21. Larson PEZ, Conolly SM, Pauly JM, Nishimura DG. Using adiabatic inversion pulses for long-T-2 suppression in ultrashort echo time (UTE) imaging. Magn. Reson. Med. 2007; 58: 952–961. 22. Weiger M, Brunner DO, Dietrich BE, Muller CF, Pruessmann KP. ZTE imaging in humans. Magn. Reson. Med. 2013; 70: 328–332. 23. Sussman MS, Pauly JM, Wright GA. Design of practical T2-selective RF excitation (TELEX) pulses. Magn. Reson. Med. 1998; 40: 890–899. 24. Shinnar M, Bolinger L, Leigh JS. The use of finite impulse-response filters in pulse design. Magn. Reson. Med. 1989; 12: 81–87. 25. Pauly J, Leroux P, Nishimura D, Macovski A. Parameter relations for the Shinnar–Leroux selective excitation pulse design algorithm. IEEE Trans. Med. Imaging 1991; 10: 53–65. 26. Du J, Bydder GM. Qualitative and quantitative ultrashort-TE MRI of cortical bone. NMR Biomed. 2013; 26: 489–506.
Copyright © 2014 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/nbm
253
This work was supported by Swiss National Science Foundation grants 310030_144075/1 and 3106030_139258/1.
REFERENCES
M. WEIGER ET AL. 27. Weiger M, Stampanoni M, Pruessmann KP. Direct depiction of bone microstructure using MRI with zero echo time. Bone 2013; 54: 44–47. 28. Du JA, Carl M, Bydder M, Takahashi A, Chung CB, Bydder GM. Qualitative and quantitative ultrashort echo time (UTE) imaging of cortical bone. J. Magn. Reson. 2010; 207: 304–311. 29. Jordan CD, Saranathan M, Bangerter NK, Hargreaves BA, Gold GE. Musculoskeletal MRI at 3.0 T and 7.0 T: a comparison of relaxation times and image contrast. Eur. J. Radiol. 2013; 82: 734–739. 30. Horch RA, Nyman JS, Gochberg DF, Dortch RD, Does MD. Character1 ization of H NMR signal in human cortical bone for magnetic resonance imaging. Magn. Reson. Med. 2010; 64: 680–687.
31. Rad HS, Lam SCB, Magland JF, Ong H, Li C, Song HK, Love J, Wehrli FW. Quantifying cortical bone water in vivo by three-dimensional ultra-short echo-time MRI. NMR Biomed. 2011; 24: 855–864. 32. Kuethe DO, Caprihan A, Lowe IJ, Madio DP, Gach HM. Transforming NMR data despite missing points. J. Magn. Reson. 1999; 139: 18–25. 33. Weiger M, Brunner DO, Wyss M, Dietrich BE, Wilm BJ, Pruessmann KP. ZTE imaging with T1 contrast. Proceedings of the 21st Annual Meeting ISMRM, Milan, Italy, 2014; 4262. 34. Schieban K, Weiger M, Hennel F, Boss A, Pruessman KP. ZTE imaging with enhanced flip angle using modulated excitation, Magn. Reson. Med. 2014. doi: 10.1002/mrm.25464.
254 wileyonlinelibrary.com/journal/nbm
Copyright © 2014 John Wiley & Sons, Ltd.
NMR Biomed. 2015; 28: 247–254
