Fast low-angle shot diffusion tensor imaging with stimulated echo encoding in the muscle of rabbit shank.

Research article Received: 31 January 2013,

Revised: 25 September 2013,

Accepted: 29 September 2013,

Published online in Wiley Online Library: 21 October 2013

(wileyonlinelibrary.com) DOI: 10.1002/nbm.3046

Fast low-angle shot diffusion tensor imaging with stimulated echo encoding in the muscle of rabbit shank Patrick Hiepea*, Karl-Heinz Herrmanna, Daniel Güllmara, Christian Rosa, Tobias Siebertb†, Reinhard Blickhanb, Klaus Hahnc and Jürgen R. Reichenbacha In the past, spin-echo (SE) echo planar imaging(EPI)-based diffusion tensor imaging (DTI) has been widely used to study the fiber structure of skeletal muscles in vivo. However, this sequence has several shortcomings when measuring restricted diffusion in small animals, such as its sensitivity to susceptibility-related distortions and a relatively short applicable diffusion time. To address these limitations, in the current work, a stimulated echo acquisition mode (STEAM) MRI technique, in combination with fast low-angle shot (FLASH) readout (turbo-STEAM MRI), was implemented and adjusted for DTI in skeletal muscles. Signal preparation using stimulated echoes enables longer effective diffusion times, and thus the detection of restricted diffusion within muscular tissue with intracellular distances up to 100 μm. Furthermore, it has a reduced penalty for fast T2 muscle signal decay, but at the expense of 50% signal loss compared with a SE preparation. Turbo-STEAM MRI facilitates high-resolution DTI of skeletal muscle without introducing susceptibility-related distortions. To demonstrate its applicability, we carried out rabbit in vivo measurements on a human whole-body 3 T scanner. DTI parameters of the shank muscles were extracted, including the apparent diffusion coefficient, fractional anisotropy, eigenvalues and eigenvectors. Eigenvectors were used to calculate maps of structural parameters, such as the planar index and the polar coordinates θ and φ of the largest eigenvector. These parameters were compared between three muscles. θ and φ showed clear differences between the three muscles, reflecting different pennation angles of the underlying fiber structures. Fiber tractography was performed to visualize and analyze the architecture of skeletal pennate muscles. Optimization of tracking parameters and utilization of T2-weighted images for improved muscle boundary detection enabled the determination of additional parameters, such as the mean fiber length. The presented results support the applicability of turbo-STEAM MRI as a promising method for quantitative DTI analysis and fiber tractography in skeletal muscles. Copyright © 2013 John Wiley & Sons, Ltd. Keywords: MRI; stimulated echo; STEAM; turbo-FLASH; diffusion tensor imaging; fiber tractography; muscles; rabbit shank

INTRODUCTION The field of biomechanics uses sophisticated models to investigate and describe the relationship between architectural and functional features of complex muscle structures. The development and generation of realistic models require a knowledge of the structural arrangement, which is ideally obtained in vivo

* Correspondence to: P. Hiepe, Medical Physics Group, Department of Diagnostic and Interventional Radiology I, Jena University Hospital–Friedrich Schiller University Jena, ‘MRT am Steiger’, Philosophenweg 3, 07743 Jena, Germany. E-mail: [email protected] a P. Hiepe, K.-H. Herrmann, D. Güllmar, C. Ros, J. R. Reichenbach Medical Physics Group, Institute of Diagnostic and Interventional Radiology I, Jena University Hospital–Friedrich Schiller University Jena, Jena, Germany b T. Siebert, R. Blickhan Science of Motion, Institute of Sport Science, Friedrich Schiller University Jena, Jena, Germany

146

c K. Hahn Institute of Computational Biology, Helmholtz Centrum München, Neuherberg, Germany

NMR Biomed. 2014; 27: 146–157

using noninvasive high-resolution imaging techniques, such as MRI (1). In particular, diffusion tensor imaging (DTI) provides valuable information about the directional dependence of diffusion as a result of the underlying microstructure of tissues in vivo (2,3). Anisotropic diffusion in skeletal muscles arises from partly restricted diffusion caused by barriers, such as cell membranes of adjacent muscle fibers and intracellular structures such as protein chains, which also exhibit highly aligned structures (4–6). Thus, the measurement of diffusion anisotropy allows inferences to be made about microstructural tissue ordering and could, †

Present address: Institute of Sport and Motion Science, University of Stuttgart, Stuttgart, Germany Abbreviations used: 3D, three-dimensional; ADC, apparent diffusion coefficient; CP, planar index; CPC, clothes pin coil; DTI, diffusion tensor imaging; DW, diffusion weighted; DWI, diffusion-weighted imaging; EPI, echo planar imaging; FA, fractional anisotropy; FACT, Fiber Assignment by Continuous Tracking; FID, free induction decay; FLASH, fast low-angle shot; FoV, field of view; GRAPPA, generalized autocalibrating partially parallel acquisition; PE, phase encoding; PSF, point spread function; ROI, region of interest; SD, standard deviation; SE, spin echo; SNR, signal-to-noise ratio; STE, stimulated echo; STEAM, stimulated echo acquisition mode; TM, mixing time; TSE, turbo spin echo.

Copyright © 2013 John Wiley & Sons, Ltd.

TURBO-STEAM DTI IN THE MUSCLE OF RABBIT SHANK

NMR Biomed. 2014; 27: 146–157

sequences (with typical values of TE = 80 ms and Δ = 50 ms) are sensitive to structure sizes of up to 24.5 μm. Early implementations of combined STEAM DTI preparation and EPI, which are still prone to image distortions, were applied with prolonged TM values in the range 100–300 ms to maximize SNR (1,25). Recently, Sigmund et al. (29) demonstrated a STEAM DTI sequence based on EPI using TM = 1 s, and thus diffusion times of 1.02 s, which resulted in higher microstructural contrast compared with short diffusion times. However, imaging of small objects in a human whole-body scanner with high spatial resolution requires a sequence which is insensitive to B0 inhomogeneities. Therefore, the aim of this work was to combine STEAM preparation and diffusion times of approximately 1 s with a single-shot, fast low-angle shot (FLASH) readout to generate highresolution distortion-free DTI images of small fields of view (FoVs). The developed turbo-STEAM sequence was demonstrated by acquiring DTI data of a rabbit shank, which was analyzed for motion-induced image artifacts and for different diffusion parameters, such as ADC, FA and pennation angle. In addition, we proposed an improvement of fiber tracking accuracy, which may contribute to the generation of improved myo-fiber models.

MATERIALS AND METHODS Pulse sequence We modified a previously described STEAM-based DWI sequence (30) using the vendor-specific sequence development environment (IDEA, version VB17, Siemens Healthcare, Erlangen, Germany). The proposed sequence consists of two modules that generate DW stimulated echoes (STEs) and provide image formation, respectively (see Fig. 1). High b values are achieved by applying diffusion sensitizing gradients in the short τ 1 and τ 2 intervals in combination with long TM. The corresponding b value is approximated by (27): δ [1] b ¼ γ2 δ2 G2 tRF þ δ þ TM 3 where the DW gradient duration and strength are given by δ and G, respectively. A pulse duration of the three RF pulses of tRF = 2.56 ms was applied in this work. As shown in Fig. 1, TE is defined by TESTE = (τ 1 + TM + τ 2). The STEAM preparation module consists of a slice-selective 90° RF excitation, diffusion-weighting gradient pulses in predefined directions and a second slice-selective 90° RF pulse. During the time interval τ 1, loss of transverse magnetization as a result of spin dephasing is induced by field inhomogeneities (T2* relaxation) as well as by the diffusion-weighting gradient pulse. To ensure complete dephasing, which is required for appropriate generation of a STE, small diffusion gradients (5 mT/m) are also applied for the acquisition of the b0 images (26). The second 90° RF pulse rotates the disc of dephased spins and splits the initial magnetization into two components. One component is stored in the z direction and decays exponentially during TM with the time constant T1. The other component remains in the x–y plane and forms an SE, which is suppressed by crusher gradients (checkered gradients in Fig. 1). The acquisition module consists of a series of slice-selective refocusing pulses (α pulses) that fractionally tip the stored magnetization back into the transverse plane in order to generate a series of STEs. Similar to the FLASH readout technique, spatial


wileyonlinelibrary.com/journal/nbm

147

consequently, be applied to characterize the severity of injured or degenerated skeletal muscles by means of the apparent diffusion coefficient (ADC) and fractional anisotropy (FA) (7–10). The local muscle fiber orientation is generally assumed to correspond to the direction of the eigenvector associated with the largest eigenvalue and can be derived voxel-by-voxel from the measured diffusion tensor (11). The orientation of the first eigenvector provides information about the pennation angle, which denotes the angle between the load axis and muscle fibers of skeletal muscles. Diffusion tensor metric mapping, tensor orientation and fiber tractography studies have already been performed in several reports investigating the human calf (1,5,12–17), thigh (18,19) and forearm (20) muscles. Most of these DTI studies have employed single-shot, diffusion-weighted (DW), spin-echo echo planar imaging (SE-EPI) sequences with image acquisition times of less than 0.5 s (6,7,12,13,16,18–20). The rapid single-shot acquisition mode makes SE-EPI insensitive to motioninduced image artifacts. Although, SE-EPI DTI is widely established, it is sensitive to B0 inhomogeneities, which cause geometric distortions and signal loss. Furthermore, single-shot SE-EPI DTI has only limited spatial resolution because larger matrices require longer readout trains, which results in increasing blurring artifacts caused by T2* decay and decreasing signal-to-noise ratio (SNR) caused by T2 relaxation. SNR is further reduced compared with applications in brain, as muscle tissue has a distinctly smaller T2/T1 ratio at 3 T (31.7 ms/ 1420 ms) (21) compared with brain tissue (e.g. 69 ms/1084 ms for white matter) (22). To compensate for this loss in SNR, recent muscle DTI studies performed at 3 T applied b values in the range 400–500 s/mm2 with increased slice thicknesses of 3-6 mm (7,13,16,18). This range of b values is used to optimize the noise performance in the estimation of the diffusion coefficients of approximately D = (1.5–2.0) × 10–3 mm2/s (first approximation based on the expression b = 1/D for Gaussian diffusion processes). However, such b values in an SE-DW preparation use a short diffusion time to maintain TE short, which results in less sensitivity to hindered and restricted diffusion in geometries with large pore spacing, as occurs in muscle tissue (23). To improve the sensitivity to hindered diffusion effects in muscle DTI, i.e. increasing structural contrast (24) and thus FA (23), prolonged diffusion times on the order of 1 s can be used. One attempt to overcome the limitation of SE-EPI DTI in this matter is the use of a stimulated echo acquisition mode (STEAM) DWI preparation (5,24). Although STEAM results in a 50% signal reduction compared with SE preparation, it allows an increase in the applicable diffusion time by extending the mixing time TM (T1 decay) rather than TE (T2 decay), which reduces the relaxation-dependent signal loss (1,25,26). As an example, the signal of skeletal muscle tissue at 3 T acquired with TM = 1000 ms decreases to approximately one-half of the signal corresponding to TM = 10 ms. In comparison, the signal of an SE sequence would have decayed completely within 300 ms because of the short T2 relaxation time of muscle tissue. Thus, the STEAM-based approach permits diffusion times of up to Δ ≈ 1 s, which further enables the detection of restricted diffusion processes within structures possessing intracellular distances of up to 109.5 μm (three-dimensional (3D) rootmean-square displacement based on Einstein’s equation [x(Δ) – x(0)]2 = 6DΔ (27), with D = 2 × 10–3 mm2/s). These long diffusion times are required even in small animal muscle tissue, such as the shank muscles of a rabbit, which show myo-fiber diameters in the range 50–100 μm (28). In contrast, conventional SE-EPI

P. HIEPE ET AL.

Figure 1. RF pulse and gradient diagram of the implemented turbo-stimulated echo acquisition mode (STEAM) diffusion tensor imaging (DTI) sequence containing the diffusion-weighted (DW) STEAM preparation module and turbo-fast low-angle shot (FLASH) readout module. DW gradient pairs are applied in the three orthogonal directions and consist of two gradient slopes with strength G, duration δ and the same polarity. Each pulse is applied twice, once during the τ 1 interval and once during the τ 2 interval. The time between the DW gradient-related de- and rephasing of the spin evolution is defined as the diffusion time Δ. The square brackets on the right define the FLASH readout interval, which is repeated n times for n differently phase-encoded stimulated echoes (STEs). Dark gray, diffusion-encoding gradients; checkered areas, crusher gradients to suppress spin echo.

encoding is achieved by the application of phase and readout gradients prior to and during STE acquisition, respectively (31). Diffusion gradient pulses are applied directly after the α pulses in the τ 2 interval to reverse the spin dephasing previously induced by the diffusion gradients of the first module. In addition, the rephasing DW gradient slopes crush the free induction decay (FID) signals from each α pulse. The high b values are mostly self-spoiling, but to suppress unwanted signal coherence pathways during FLASH-based turbo-STEAM readout, RF and gradient spoiling is applied, as well as weak diffusion weighting for b0 images to dephase the FID signal of the α pulses. As a result of the long TM, imaging gradients can cause motion artifacts and additional undesirable diffusion weightings (26). In order to minimize these effects, all dephasing imaging gradient lobes were placed close to their rephasing counterpart, preferably within the same τ interval whenever possible. To render negligible cross-term effects on the diffusion weighting, the following strategy was applied. After application of the slice-selection gradient during the first 90° RF excitation pulse, the resulting gradient moment is overcompensated by the rewinder in the slice-selection direction (see Gz axis in Fig. 1), which pre-dephases the spins in the x–y plane prior to the application of the second, slice-selective 90° RF pulse in the τ 1 interval. During readout in the τ 2 interval, spins are rephased by the slice-selection rewinder, which is applied simultaneously with the phase encoding (PE) gradient for each phase-encoded STE, which, in turn, is refocused by the third slice-selective RF pulse. Furthermore, the readout dephasing gradient is placed in the τ 2 interval. The turbo-FLASH-based spatial encoding suffers from blurring artifacts in the PE direction as a result of a broadened point spread function (PSF) (30). This blurring occurs because the signal for each α pulse depends not only on the T1 decay, but also on the continuous fractional consumption proportional to cos(α) of the stored longitudinal magnetization, thereby leaving less

magnetization for subsequent α pulses. In addition, the transverse magnetization is further reduced by progressive diffusion weighting as a result of the extended diffusion time Δ for each incrementally acquired STE. Overall, the ratio of subsequently available longitudinal magnetization is given by: 2 2 2 Mnþ1 ¼ cosðαn ÞeðTR=T 1 Þ eDðγ δ G TRÞ Mn

[2]

Thus, by choosing the minimal TR, the decay of the amplitudes of successively produced STEs can be reduced, which results in minimal T2 signal loss during 2τ time. To further enhance SNR, a centric reordering scheme of the PE gradient table was implemented and flip angles of α = 20° were used (30,32). In addition, generalized autocalibrating partially parallel acquisition (GRAPPA)-based k space sampling (33) was applied, and rectangular FoVs were used to reduce the number of PE steps and thus minimize blurring effects. MRI measurements In vivo MRI protocol optimization was performed with four rabbits on a 3 T human whole-body MR scanner (MAGNETOM Trio TIM, Siemens Medical Solutions, Erlangen, Germany) with a maximum gradient amplitude of 28 mT/m and a maximum gradient slew rate of 178 mT/m/ms. Data were acquired with an eightchannel, receive-only, multipurpose coil, which consists of two elements, each containing four small loop coils [clothes pin coil (CPC), Noras MRI Products GmbH, Höchberg, Germany; see Fig. 2]. All imaging experiments were approved by the local animal care committee. In this paper, we present the MRI data obtained with an optimized sequence protocol, which were collected in the left shank of a female rabbit (Oryctolagus cuniculus, m = 3.1 kg).

148

Figure 2. Experimental set-up for in vivo diffusion tensor imaging (DTI) measurements in the left hind shank of a rabbit using a clinical human wholebody MR scanner and a two-element multipurpose coil (clothes pin coil, CPC).



NMR Biomed. 2014; 27: 146–157

TURBO-STEAM DTI IN THE MUSCLE OF RABBIT SHANK The rabbit was anesthetized using a mixture of ketamine (10 mg/kg/h) and xylazine (1 mg/kg/h). Fresh oxygen (0.5 L/min) was applied to the animal through a nasal tube during the measurements. A custom-made framework immobilized the examined leg of the animal. The shanks of the animal were aligned parallel to the magnetic field of the scanner. MR images were acquired from the mid-calf region with the rabbit’s leg in a relaxed position. A T2-weighted, high-resolution 3D scan with sagittal orientation was performed to image the anatomy of the leg [SPACE sequence (3D turbo spin echo variant with variable flip angles): TR/TE = 2500/ 333 ms; FoV, 128 × 116 mm2; matrix, 256 × 231 pixels; bandwidth, 145 Hz/pixel; slice thickness, 0.5 mm; 192 slices per 3D block; TA = 69 min]. Oversampling in the PE (96 %) and slice (8.3 %) direction was applied to avoid aliasing artifacts. Images were intensity corrected to account for coil sensitivity profiles using routines included in the FSL image processing package (34). Overall, four turbo-STEAM DTI datasets were acquired (NEX = 4), whereby each dataset consisted of five low b value images (b = 25 s/mm2, δ = 3.7 ms, G = 5 mT/m) and 30 DW images with diffusion weighting in 30 different spatial directions (b = 600 mm2/s, δ = 3.7 ms, G = 25 mT/m; DW amplitude was limited because of stimulation restrictions). Further sequence parameters were: TR/ TRall/TM = 14.8/1678/980 ms; τ 1 = τ 2 = 10 ms; FoV, 256 × 72 mm2; matrix, 256 × 72 pixels; in-plane resolution, 1.0 mm × 1.0 mm; bandwidth, 310 Hz/pixel. A fat saturation pulse was applied prior to the STEAM preparation to avoid off-resonance artifacts caused by the low bandwidth. Thirty-one slices, each 2 mm thick, were acquired in the sagittal orientation and parallel to the orientation of the tibia bone. Using a negative slice distance factor of – 50% during interleaved two-dimensional acquisitions, an effective isotropic resolution of 1.0 mm3 was achieved. A GRAPPA acceleration factor of two was used, which results in a turbo factor of 48 (number of STEs after single excitation). The total acquisition time for turbo-STEAM DTI data was 128 min.

coefficients ρ were averaged (ρmean) and ranges (ρmax–ρmin) for each DW image index were determined. The SNR of turbo-STEAM images was calculated for single and arithmetically averaged raw data and for denoised data using the mean value of a 3D region of interest (ROI) in the soleus muscle (see description of ROI determination below) and the SD of signal intensities in a box of 20 × 72 × 31 pixels that contained only background noise. For tensor reconstruction and DTI analyses, the software Diffusion Toolkit (R. Wang, Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Boston, MA, USA) v1 , → v2 was used. Eigenvalues λ1, λ2 and λ3 and eigenvectors → → and v 3 of the diffusion tensors were determined in each voxel. ADC, FA and maps of the first eigenvector orientation (expressed in polar coordinates θ and φ) were derived from these data. Differences between the second and third eigenvalues were analyzed by calculating the planar index CP, which is a measure of cross-sectional fiber ellipticity (15): CP ¼ 2

ðλ2 λ3 Þ λ2 þ λ2 þ λ3

[3]

In a given frame of reference, the orientation of the primary eigenvector can be described with two angles: zenith and azimuth (1,5,13,15). In this study, the zenith angle was defined as the angle between the first eigenvector → v 1 and the readout direction of the sequence → r , which corresponded to the proximal–distal direction on the sagittal diffusion tensor images, and thus to the direction of the tibia (Fig. 3). The zenith angle θ was calculated as follows: ! →→ r v1 θ ¼ arccos → → r jj v 1

[4]

Denoising of DTI data To increase the SNR of the high-resolution DTI data obtained, a denoising procedure was applied, which reduces background noise in single DW images by combining voxel-wise averaging, nonlinear filtering (edge preserving smoothing) and Rician bias correction (35). Monte Carlo validations of the method can be found in refs. (35,36). This denoising approach achieves an SNR improvement by a factor of 3–6, and blurring effects between voxels are minimized because of the nonlinearity of the filter, keeping the effective resolution of the DWIs practically unchanged. The filter was implemented in IDL (ITT Visual Information Solutions, Boulder, CO, USA).

MR image analyses

NMR Biomed. 2014; 27: 146–157

and varies between 0° and 180°. Maps of CP, θ and φ were calculated with custom-written MATLAB routines. For ROI-based DTI analyses, the soleus, tibialis and gastrocnemius medialis muscles were manually segmented slice-by-slice on the reformatted transverse images of the 3D high-resolution TSE dataset. After applying a closing procedure (including eroding



149

To address the issue of motion-induced image artifacts, as a result of the diffusion time of 1 s and the long duration of the image readout of 700 ms, we analyzed the variance across multiple acquisitions. Motion effects, including blood pulsation and spontaneous muscle twitching, in low- and high-DW images were investigated by pixel-wise calculation of the standard deviation (SD) and by image-based two-dimensional correlation analyses of the multiple DTI datasets DWINEX = 1,2,3,4 in MATLAB (The Mathworks, Inc., Natick, MA, USA). The latter was performed to measure the similarity between images of DWINEX = 1 and DWINEX = 2,3,4 for every slice position and DW direction (DW image index). The calculated correlation

The zenith angle varies between 0º and 90º and is an estimate of the pennation angle for the majority of shank muscles. In general, the pennation angle denotes the angle between the reconstructed fiber orientation and the line of action. The latter was determined for the soleus, tibialis and gastrocnemius medialis muscles based on the T2-weighted 3D scan in more detail. This analysis was performed by an experienced biomechanics scientist (TS), and was used to correct the ROI-based θ values for the determination of corrected pennation angles θcorr. The azimuth angle φ corresponds to the angle between the → projection of the first eigenvector v 1 onto the transverse plane and the anterior–posterior direction, and is defined as follows: 8 ! → > v 1z > → > > arctan → v 1z > 0 > < v 1x φ¼ [5] ! → > > v 1z > → > > arctan → þ π v 1z < 0 : v 1x

P. HIEPE ET AL.

Figure 3. T2-weighted high-resolution images of a rabbit’s leg before (a) and after (b–d) intensity correction. In (b), the joint angles are illustrated and, in (c), the three lines of action of the soleus (red, 1), tibialis (blue, 2) and gastrocnemius medialis (orange, 3) muscles are indicated. The lines of action of the gastrocnemius medialis and tibialis muscles are aligned parallel to the tibia (thick white line), whereas the soleus muscle is slightly deflected relative to the tibial axis (ε = 2°). The dotted white lines indicate the planes of the transverse slices shown in (d), where overlaid regions of interest (ROIs) mark the three shank muscles.

and dilating to smooth the boundaries), the segmented muscle ROIs were registered and transformed to match the turbo-STEAM DTI data. This transformation was performed by applying the affine transformation algorithm supplied with the FSL package (37). To assess statistical differences in ADC, FA, eigenvalues, CP and orientation angles θ and φ between the three segmented muscles, ROI-specific values were compared using a two-sample t-test with the null hypothesis that the data of each muscle are independent random samples from normal distributions with equal means using MATLAB. As a result of the large number of analyzed ROI voxels, a Bonferroni correction was performed and the statistical significance level was defined as p < 0.001. Fiber tracking was performed using the deterministic Fiber Assignment by Continuous Tracking (FACT) algorithm (38). Thresholds of 0.3 for FA and 7° per iteration step (0.1 voxel) for the maximum change in direction angle were selected as stopping criteria for all fiber tracking procedures. As artificially prolonged tracts were observed during prior pilot fiber tracking, which occasionally jumped erroneously to adjacent muscles, a mask derived from the intensity-corrected T2-weighted images was employed, in addition, to optimize the tracking results. The threshold for T2 masking was successively increased until no prolonged, nonanatomical fiber tracts were detected. To assess the tracked fiber paths visually, they were overlaid onto the manually segmented muscles. All reconstructed fiber tracts were transformed to the 3D highresolution T2-weighted MRI dataset for improved visualization. The reconstructed tracts were displayed and evaluated using TrackVis (R. Wang, Athinoula A. Martinos Center for Biomedical Imaging). In accordance with the general definition of color coding of fiber directions, fibers aligned parallel to right–left, anterior–posterior and proximal–distal directions were visualized in red, green and blue, respectively. Mean fiber lengths of the soleus, tibialis and gastrocnemius medialis muscles were determined by performing fiber tracking in each segmented muscle ROI separately. Voxels outside the corresponding ROI were excluded from tracking and no T2 masking was applied. ROIbased fiber tracking results were further used to evaluate fiber tracts which were obtained with T2 masking.

RESULTS MR image analyses

150

Figure 3 displays high-resolution sagittal T2-weighted images of the left shank of one examined rabbit (Fig. 3a–c), together with


three reformatted transverse slices of the original 3D dataset (Fig. 3d) to visualize cross-sections at different positions. As a result of the nonuniform sensitivity profile of the RF coil, the original images (in Fig. 3a) demonstrate inhomogeneous intensity distributions, which were corrected by applying an intensity correction method (in Fig. 3b). Figure 3b contains information about the position and angulations of the knee and ankle joint (deflection angles of 127° and 94°, respectively). The biomechanics scientist (TS) confirmed that the muscles were relaxed without significant passive shortening or stretching. The muscle tendons at both joints were determined (see Fig. 3c, in which the tendons are schematically marked by white circles) and used to define the lines of action of the three examined muscles: tibialis (blue region, Fig. 3d), gastrocnemius medialis (orange region, Fig. 3d) and soleus (red region, Fig. 3d). In contrast with the tibialis and gastrocnemius medialis muscles, whose lines of action (blue and orange lines in Fig. 3c) are aligned parallel to the tibia, the soleus muscle reveals a slight deflection of the line of action (ε = 2°, red line in Fig. 3c) relative to the tibial axis. The results shown in Fig. 4a, b demonstrate native turbo-STEAM DW images in the sagittal orientation (b = 25 and 600 s/mm2), which are free of motion-induced phase errors caused by nonrigid body motion (no destructive phase interferences which lead to signal losses) and have sufficient SNR (SNRb25 = 19, SNRb600 = 13). The corresponding SD maps of multiple low b value images (NEX = 4) show only a slight and local increase in SD in the anterior region of the shank, and thus suggest motion artifacts caused by arterial pulsation in this area. The mean correlation coefficients between the multiple low-DW images revealed ρmean in the range 0.9–0.95 for central sagittal slices, whereas high-DW images yielded distinctly decreased ρmean of 0.8–0.9. Slices that contained less signal information of the object revealed decreased ρmean values. Thus, the major effect in ρmean is linked directly to the SNR of the particular images. However, the differences in the determined correlation coefficients are very small (ρmax–ρmin = 0–0.02 for central sagittal slices) and systematic (SNR dependence). Motion-induced phase errors, which should appear as a random deviation, could not be observed. Figure 5 displays transverse slices of the turbo-STEAM MRI dataset and parameter maps for ADC, FA, CP and polar coordinates (θ, φ) of the first eigenvector. The first row contains images and parameter maps derived from the original noisy data; the second row shows the corresponding results after denoising. The slice positions in Fig. 5 were selected similarly to the


NMR Biomed. 2014; 27: 146–157


Figure 4. Native sagittal turbo-stimulated echo acquisition mode (STEAM) diffusion tensor imaging (DTI) of a rabbit shank with low (a) and high (b) diffusion weighting. The images have sufficient signal-to-noise ratio (SNR) (a, SNR = 19; b, SNR = 12) and are free from motion-induced phase errors. Corresponding standard deviation (SD) maps of multiple DTI acquisitions (c, d) show only small variations in the anterior region of the rabbit shank (marked with arrows). Variations in the color-coded mean correlation values ρmean (e) and maximum coefficient changes ρmax–ρmin (f), which were calculated on the basis of the two-dimensional image correlation analysis of multiple DTI datasets, are directly linked to the SNR of the particular slice and diffusionweighted image index. Central slices show very small variations of ρmax–ρmin, which indicates the absence of random motion-induced phase errors.

NMR Biomed. 2014; 27: 146–157

(in the transverse direction) for the three different muscles (see Fig. 6, top row). In addition, θ and φ of the three segmented muscles were mapped in a representative sagittal slice (see Fig. 6, bottom row). These plots demonstrate spatial variations of θ and φ for all examined muscles at muscular boundaries. The zenith angle θ yielded prominent variations at the superior and inferior end of each muscle, whereas φ revealed regularly ordered variations over the whole muscle (e.g. see gastrocnemius medialis muscle in Fig. 6).

Fiber tracking Figure 7 shows muscle fiber tracts in the left hind leg of a rabbit and provides an overview of all fiber tracts that pass through either the proximal or distal transverse slice (see transparent planes in Fig. 7 and the top and bottom transverse images in Fig. 3c, d). Fiber tracts with different orientations can be distinguished. In particular, Fig. 7 emphasizes the tracts of the tibialis muscle, which are aligned almost parallel to the tibia, in contrast with the tracts of, for example, the gastrocnemius muscle, which clearly show a greater deflection relative to the leg axis (see arrows in Fig. 7a). This observation indicates different pennation angles in the different skeletal muscles and corresponds to the results for the zenith angle θ (Table 1), where the tibialis muscle reveals a smaller θ angle compared with the other two muscles. Of special note is the bi-pennate structure of the gastrocnemius muscle, including the medial and lateral head, which is clearly visible in the posterior view (thick arrows in Fig. 7b). Tracts continue until they merge in the Achilles tendon (see thin arrow in Fig. 7b).



151

positions of the three transverse slices shown in Fig. 3d. Compared with arithmetically averaged raw data (SNRb25 = 29, SNRb600 = 18), the denoised turbo-STEAM DW images yield a markedly increased SNR (SNRb25 = 202 and SNRb600 = 102). The improvement by denoising is particularly prominent for the CP, θ and φ maps, for which the denoised data (in contrast with the uncorrected data) provide superior delineation of the different muscles, such as the soleus (1 in the T2-weighted image, which is displayed in the bottom row of Fig. 5) and gastrocnemius medialis (3) and adjacent muscles. The variations in θ and φ are most probably caused by differences in the orientation of muscle fibers with respect to the tibia. Table 1 summarizes the ROI-based mean values and SDs of the DTI parameters of the three shank muscles (cf. Fig. 3d). SDs were reduced after denoising, which also increased ADC and the eigenvalues compared with the unfiltered data. Bonferroni-corrected pair-wise t-test comparisons partially revealed significant differences of the determined mean DTI measures between specific muscles (p < 0.001), whereby denoising helped to detect significant intermuscular differences (e.g. compare FA and CP of raw and denoised data in Table 1). The highest intermuscular parameter variations were obtained for tensor orientation angles θ and φ. As a result of the deflected line of action of the soleus muscle, the corresponding zenith angle must be corrected for ε = 2° which leads to θcorr = (11.3 ± 6.1). However, the tibialis muscle yielded significantly decreased angle values (p < 0.001) compared with the soleus and gastrocnemius medialis muscles. In the 3D ROIs, relatively high standard deviations of the calculated θ and φ values were found (see Table 1). In a more detailed analysis, ROI-specific θ and φ values were averaged slice-by-slice

P. HIEPE ET AL.

Figure 5. Reformatted transverse turbo-stimulated echo acquisition mode (STEAM) diffusion tensor imaging (DTI) reconstructions of raw (first row) 2 and denoised (second row) datasets. The columns contain from left to right: diffusion-weighted (DW) images (b = 600 s/mm , DW in a single direction), apparent diffusion coefficient (ADC), fractional anisotropy (FA), planar index (CP), θ and φ maps. Raw and denoised DW images mainly show blurring artifacts in the phase-encoding (PE) direction (left–right direction, encircled in red). After denoising, particularly the maps of CP, θ and φ show improved intermuscular contrasts. The bottom row shows a T2-weighted image and the corresponding θ and φ maps, whereby the soleus (1 in T2) and gastrocnemius medialis (3 in T2) can be clearly distinguished from the surrounding muscles (see arrows).

Table 1. Mean values and standard deviations (SDs) of the apparent diffusion coefficient (ADC), fractional anisotropy (FA), three eigenvalues (λ1–3), planar index (CP), zenith angle θ and azimuth angle φ, calculated using raw and denoised turbo-stimulated echo acquisition mode (STEAM) data, for manually segmented soleus, tibialis and gastrocnemius medialis muscles Data ADC (× 10–3 mm2/s) FA (a.u.) λ1 (× 10–3 mm2/s) λ2 (× 10–3 mm2/s) λ3 (× 10–3 mm2/s) CP (a.u.) Zenith angle θ (deg) Azimuth angle φ (deg)

Raw Denoised Raw Denoised Raw Denoised Raw Denoised Raw Denoised Raw Denoised Raw Denoised Raw Denoised

Soleus muscle

Tibialis muscle

Gastrocnemius medialis muscle

0.96 ± 0.12b 0.96 ± 0.12ab 0.44 ± 0.08 0.43 ± 0.07ab 1.46 ± 0.16ab 1.46 ± 0.14ab 0.79 ± 0.14ab 0.78 ± 0.12ab 0.62 ± 0.14b 0.65 ± 0.13 0.13 ± 0.08b 0.09 ± 0.06ab 14.7 ± 7.2ab 13.3 ± 6.1a 98.8 ± 59.9ab 104.3 ± 59.8ab

0.96 ± 0.09c 0.98 ± 0.09 0.43 ± 0.08c 0.42 ± 0.06c 1.45 ± 0.15c 1.48 ± 0.11c 0.81 ± 0.12c 0.81 ± 0.12c 0.62 ± 0.11c 0.66 ± 0.10 0.14 ± 0.08c 0.10 ± 0.06c 10.7 ± 7.3c 8.3 ± 4.6c 78.3 ± 53.6 65.3 ± 51.6c

0.94 ± 0.18 0.99 ± 0.17 0.44 ± 0.13 0.40 ± 0.09 1.40 ± 0.26 1.44 ± 0.23 0.83 ± 0.21 0.85 ± 0.21 0.56 ± 0.22 0.66 ± 0.19 0.22 ± 0.48 0.14 ± 0.21 20.6 ± 16.4 14.4 ± 13.0 83.8 ± 51.1 81.2 ± 45.8

Significant difference (p < 0.001) between soleus and tibialis muscles. Significant difference (p < 0.001) between soleus and gastrocnemius medialis muscles. c Significant difference (p < 0.001) between tibialis and gastrocnemius medialis muscles. a

b

152



NMR Biomed. 2014; 27: 146–157


Figure 6. Detailed tensor orientation analyses based on denoised datasets showing the zenith θ (left) and azimuth φ (right) angles (both in degrees). The polar coordinates of the three examined muscles are averaged in the transverse plane of the corresponding region of interest (ROI) (top row) and mapped in a representative sagittal slice (bottom row). High deviations of the mapped zenith angle can be observed at muscular boundaries, whereas the azimuth varies over the entire muscles, especially for the gastrocnemius medialis (M. g.m.).

NMR Biomed. 2014; 27: 146–157



153

Figure 7. Fiber tracking results based on turbo-stimulated echo acquisition mode (STEAM) diffusion tensor imaging (DTI) data in sagittal (a) and coronal (b) views of a rabbit’s left hind leg. Illustrated tracts represent all reconstructed fiber paths that pass through one of the two illustrated transparent transverse slices. Directional color-encoded tracts give a good impression of the pennate architecture of the rabbit’s shank muscles, which lead to variously deflected fiber tracts [compare fibers tracts of the tibialis and gastrocnemius muscles marked with arrows in (a)]. In particular, the posterior view shows the bi-pennate build-up of the gastrocnemius muscle, including the medial and lateral head [see thick arrows in (b)]. Tract paths continue until attaching a corresponding tendon, such as the fibers of the gastrocnemius muscle flowing in the Achilles tendon [marked with thin arrow in (b)].

P. HIEPE ET AL. Reconstructed fiber tracts of the isolated soleus and gastrocnemius medialis muscles are shown in greater detail in Fig. 8. For each of the two muscles, a single transverse slice (yellow) of the manually segmented 3D ROIs was selected to display only those fiber tracts which pass through this transverse slice (see Fig. 8a–e). Figure 8a shows the resulting tracts using FA > 0.3 and a maximum tracking angle difference of 7° as thresholds for the tracking procedure. Figure 8b shows the same tracts, together with the segmented 3D muscle ROI volume, for the soleus (in red) and gastrocnemius medialis (in orange) muscles. As visible in Fig. 8b, there are some tracts that falsely cross boundaries between different muscles (see white arrow in Fig. 8b for the soleus muscle). Application of an additional mask obtained from thresholding the high-resolution anatomic images, as described in the Material and Methods section, removed the false tracts (see white arrows in Fig. 8b, d). This procedure reduced the number of fiber tracts, which is demonstrated in Fig. 8a, c (see white arrows). Both fiber bundles passing through the indicated transverse slice are less compact than in Fig. 8a, and the tracts do not cover the muscle. To extract parameters, such as the mean fiber length, as accurately as possible and to evaluate T2-masked tracking results, tracking is confined to the manually segmented 3D ROIs instead of using additional T2 masking. The ROI-based tracking results for the soleus and gastrocnemius medialis muscles are shown in Fig. 8e, f, whereas Fig. 8e shows, analogously to Fig. 8a–d, slicefiltered fiber tracts. T2- and ROI-masked fiber tracts reveal comparable results. Threshold-specific T2 masking provides complete boundary detection (compare arrows in Fig. 8b and Fig. 8d), but at the cost of true fiber tracts within the gastrocnemius medialis muscle (compare arrows in Fig. 8a and Fig. 8c).

Figure 8f shows all reconstructed ROI-masked fiber tracts within the two investigated muscles. Based on these results, mean fiber tract lengths of the soleus and gastrocnemius medialis muscles were determined and yielded 11.2 ± 6.2 mm and 16.2 ± 9.1 mm, respectively. The fiber tracking results for the tibialis muscle (not shown) yielded a mean fiber tract length of 19.2 ± 11.5 mm.

DISCUSSION Turbo-STEAM sequence In this work, a DTI sequence with STEAM preparation and FLASH readout module is presented. The presented single-shot images of the proposed turbo-STEAM DTI sequence are free of geometric distortions, motion-induced phase errors and have sufficient SNR (see Fig. 4). Compared with previously reported STEAM EPI acquisition techniques, which were applied in skeletal muscles (1,25), the proposed sequence allows improved spatial resolution for DT images without increased off-resonance-induced geometric distortions or T2*-related PSF blurring. We obtained DTI datasets with an isotropic resolution of 1 mm3 using a negative slice distance of –50% for 2-mm-thick slices and acquisition of a 256 × 72 pixel matrix. However, these images are affected by partial volume effects in the slice and PE direction, with this effect more prominent in the PE direction (compare DW images in Fig. 5). Thus, a major limitation to effective spatial resolution in turbo-STEAM DTI is the limited available longitudinal magnetization that is used to collect n-different phase-encoded echoes.

154

Figure 8. Reconstructed tracts for the soleus (red) and gastrocnemius medialis (orange) muscles, which pass through the selected transverse slices (yellow), without (a, c) and with (b, d) overlaid three-dimensional (3D) regions of interest (ROIs). Tracking was performed without (a, b) and with (c, d) additional masking using anatomic image information [tracking stop criteria: fractional anisotropy (FA) = 0.3; maximum angle deviation, 7°]. In (e, f), ROI-based fiber tracking results are displayed which can be used to quantify the mean fiber lengths. For the soleus and gastrocnemius medialis muscles, mean lengths of 11.2 ± 6.2 mm and 16.2 ± 9.1 mm, respectively, were determined. The tibialis muscle (not shown here) revealed a mean fiber tract length of 19.2 ± 11.5 mm.



NMR Biomed. 2014; 27: 146–157

TURBO-STEAM DTI IN THE MUSCLE OF RABBIT SHANK As each refocusing pulse successively expends the prepared STEAM magnetization within the repetitive FLASH readout interval, the actual signal distribution in k space strongly depends on the applied RF flip angle, affecting the PSF in the PE direction, and thus limiting the feasible number of PE steps (30,32). Furthermore, intrinsic T1 relaxation effects and diffusion-related signal losses contribute to blurring. Consequently, we minimized the number of PE steps using parallel imaging and by acquiring a rectangular FoV. Although the number of acquired STEs was reduced to n = 48, a distinct signal decay over the echo train, and thus a reduced effective phase resolution, was observed in this study, which was indicated by blurring in the PE direction (see Figs 4 and 5). Simulations based on Equation [2] (not shown) demonstrated a reduction factor of PE resolution of approximately 50%. Hence, major applications of the presented approach may be seen in small-FoV acquisitions, such as animal muscle DTI, where the number of PE steps is low, and thus high-resolution turbo-STEAM DTI achieves the maximum benefit compared with SE-EPI. Segmented FLASH readout minimizes PSF blurring, but results in extended acquisition times, and thus is difficult to put into practice. The application of variable flip angles throughout the echo train is another possibility to reduce blurring, whereby the variation of the flip angle is used to shape the PSF to achieve optimal image quality (31). However, the disadvantage of this approach is the reduced SNR because of the typically small starting flip angle needed for a high resolution (e.g. for n = 48 echoes as in our study: α1 = 8.6°). Compared with a constant readout flip angle, as applied in the present work (α = 20°), the variable flip angle approach would reduce the resulting SNR by a factor of approximately three. Future sequence developments should explore alternative sampling schemes, such as undersampled radial trajectories (39), which consume the limited available magnetization more efficiently. PROPELLER trajectories would allow the segmentation of the data acquisition into multiple shots (40,41), and may thus overcome the limited number of PE lines and spatial resolution. At the same time, the repeated scanning of the k space center with PROPELLER acquisition could improve SNR without increasing PSF blur, albeit at the cost of longer acquisition times. In certain in vivo experiments, where minimum macroscopic motion is guaranteed, multishot acquisition based on EPI may be another alternative (42).

diffusion time of 986 ms. Kim et al. (23) have demonstrated with measurements ex vivo that, with increased diffusion time, diffusion in the plane perpendicular to the mean fiber orientation becomes more restricted and the second and third eigenvalues decrease, whereas FA increases. It has also been proposed recently that very long diffusion times (on the order of 1 s) can be used to increase the FA values and to decrease the cone of uncertainty in skeletal muscle DTI (24). However, increasing TM results in a decreased SNR, which can lead to artificially increased FA values (44), but can be compensated by averaging (without respect to acquisition time) and application of denoising procedures. Based on the root-meansquare displacement formalism, we assumed an optimal TM for the presented type of application in the region of 1 s (as described in the Introduction), which results in high structural contrast of the muscle tissue, as reported previously (29). However, in future studies, simulations of SNR and FA as functions of TM are required to minimize the SD of FA for a constant total scan time. Of note, the corresponding long acquisition times of approximately 1.7 s for one turbo-STEAM image did not yield motioninduced phase errors (see Fig. 4). Residual motion artifacts as a result of arterial pulsation in low-DW images, which were detected in this study (marked with arrows in SD maps in Fig. 4), may affect the determination of diffusion parameters (5,17,45), but can be eliminated by electrocardiogram triggering or by applying higher b values for low-DW images. However, because of nonrigid body motion during long scan times, the application of the proposed sequence in nonanesthetized humans can be problematic. The denoising procedure which was applied includes Rician bias correction and edge-preserving nonlinear image smoothing (36). It significantly improves the SNR of DW turbo-STEAM images (see Table 1 and Fig. 4) and enables accurate DTI parameter mapping (see Fig. 5). However, the accuracy of ADC measurements was limited because of an incorrect estimation of the applied b value, which was calculated for the first STE and was underestimated for subsequent PE echoes. An underestimation of the applied b value results in overestimated ADC values. However, the first STE represents the innermost k space line with centric PE reordering, and thus the proposed b value estimation is a sufficient first approximation of the effective b value (46). DTI quantities

Turbo-STEAM DTI in skeletal muscles

NMR Biomed. 2014; 27: 146–157

The determined ADC values of the three examined rabbit shank muscles are in the range of previously reported in vivo studies, which were acquired in a mouse hind leg (4) and human thigh muscles (19). Compared with ADC maps, the denoised FA maps yield increased intermuscular contrast that corresponds to architectural differences (compare Fig. 5). In addition to the local FA reductions caused by different fiber orientation patterns of adjacent muscles, slight FA changes for different shank muscles were observed. In a cadaveric DTI study of thigh muscles, Budzik et al. (19) reported increased FA values for nonpennate muscles (e.g. sartorius) compared with pennate muscles (e.g. vastus medialis). This is in line with the results presented in this study, where the tibialis (nonpennate) muscle revealed a significantly (p < 0.001) increased FA compared with the gastrocnemius medialis (pennate) muscle. High intermuscular contrast was also observed in maps of CP, θ and φ values (demonstrated in Fig. 5). CP mapping assesses



155

Compared with applications in the brain, DTI in skeletal muscles is technically more challenging because of the lower water proton density, reduced B0 homogeneity and shorter transverse relaxation times T2 (10,25). Obtaining sufficient SNR, which is critical for tensor reconstruction and fiber tracking, is therefore crucial, and requires higher magnetic field strengths and optimized acquisition parameters. In particular, the b value needs to be adjusted carefully as it is the primary parameter determining the signal attenuation in a DWI sequence. As a result of the limited SNR in DW images of skeletal muscles, the application of high b values, measuring at long diffusion times, is impaired with SE-EPI sequences. Compared with previously reported in vivo DTI studies based on stimulated echoes, in which b values of 700 s/mm2 and diffusion times in the range 150–480 ms were used (1,5,25,43), in this study a STEAM-based DTI preparation was applied with a

P. HIEPE ET AL. myo-fiber cross-sectional ellipticity (6,16,47), and thus can provide, for example, the detection of older, atrophied or injured muscles which yield decreased fiber cross-sectional ellipticities (48). As reported previously by Karampinos et al. (15), highly pennate muscles, such as the gastrocnemius, reveal the highest CP values. In this study, the CP values of the gastrocnemius medialis were increased significantly compared with those of the soleus and tibialis muscles. The investigated pennate shank muscles in this study, the soleus and gastrocnemius medialis, yielded increased zenith angles θcorr compared with the nonpennate tibialis muscle. The determined pennation angles θcorr of the rabbit shank muscles were 8.3 ± 4.6°, 11.3 ± 6.1° and 14.4 ± 13.0° for the tibialis, soleus and gastrocnemius medialis muscles, respectively, which are in accordance with previous results. Lieber and Blevins (49) reported similar angles of 3.0 ± 1.0°, 8.5 ± 1.5° and 13.8 ± 1.7° on the muscle surface by means of a dissecting microscope and a goniometer. Slice-by-slice analyses of θ and φ revealed variations along the muscle (shown in Fig. 6, top row), whereas mapped θ and φ variations were prominent in regions of muscular boundaries consisting of aponeuroses and tendons (see Fig. 6, bottom row). Previous human DTI and fiber tracking studies have also demonstrated pennation angle heterogeneities caused by variations in the orientation of aponeuroses (14,16). Regularly ordered variations of φ in the inner muscular regions may arise as a result of a physiologically twisted fiber pattern along the leg axis. Hence, the achieved effective spatial resolution of 1 × 2 × 1 mm3, which, to our knowledge, is higher than any previously published DTI results obtained in skeletal muscles at 3 T, provides highly resolved mapping of the lead eigenvector direction, and thus improved characterization of complex fiber architectures. Fiber tractography

156

Fiber tracking reveals an estimation of fiber direction (see Fig. 7); however, boundaries between different muscles are not reliably identified based solely on DTI information, as demonstrated in Fig. 8. This has also been reported in previous studies (4,10,16,20). One major limitation of correct fiber reconstruction is the limited effective DTI resolution, as mentioned above. However, for detailed delineation of fiber tract patterns and contractioninduced deformations of different muscle compartments, it is necessary to track muscle compartments individually. Therefore, in this work, directional constraints, FA thresholding and additional T2 masking based on anatomical T2-weighted data were employed as stop criteria, where tracking results in a sufficiently reduced number of erroneously prolonged fiber tracts. As turbo-FLASH acquisition leads to geometrically undistorted diffusion tensor images, and thus provides a very good match with the T2-weighted data, T2 masks could easily be applied to the DTI data. Thus, compared with ROI filtering based on automatic muscle segmentation, which is, to our knowledge, still an unsolved problem, T2 masking provides a simple empirical approach to cut off erroneously prolonged fibers. For the soleus, gastrocnemius medialis and tibialis muscles, mean fiber lengths were determined as 11.2 ± 6.2 mm, 16.2 ± 9.1 mm and 19.2 ± 11.5 mm, respectively. These results are in good agreement with those of Lieber and Blevins (49), who reported mean lengths of 13.8 ± 0.8 mm, 14.7 ± 0.7 mm and 38.1 ± 3.0 mm, respectively, in post-mortem studies.


In the future, DTI and tractography based on the turbo-STEAM sequence and improved fiber tracking, as presented in this work, might become an important tool for the evaluation of muscle function (5,6,13,50). Functional DTI studies can assess information on the deformation of individual fiber bundles in different skeletal muscles during contraction, which may serve to develop realistic biomechanical finite element models.

CONCLUSION The present study has demonstrated the feasibility of extracting the 3D architecture of skeletal muscles in vivo with high spatial resolution on a clinical 3 T scanner by combining a STEAM-based DTI preparation and a FLASH-based readout module. The STEAM preparation allows extended diffusion times for the improved detection of restrictive diffusion in muscle tissues. The FLASHbased single-shot readout module is less sensitive to field inhomogeneity, which is a common problem with EPI sequences. Thus, turbo-STEAM DTI is a promising approach for high-resolution DTI of skeletal muscles without being compromised by motion artifacts and which allows the resolution of muscles with small cross-sections and the delineation of fiber architecture within multipennate muscles. It is anticipated that further improvements will enhance its value in clinical applications to detect degenerative changes of muscle architecture.

Acknowledgements This study was supported by the Center for Interdisciplinary Prevention of Diseases related to Professional Activities (KIP) founded by the Friedrich Schiller University Jena and the Accident Prevention and Insurance Association for Food and Restaurants (Berufsgenossenschaft Nahrungsmittel und Gaststätten, BGN, Germany). The project was also supported by the German Research Foundation (Deutsche Forschungsgemeinschaft, DFG, Germany), grant number: RE 1123/12-1.

REFERENCES 1. Karampinos DC, Banerjee S, King KF, Link TM, Majumdar S. Considerations in high-resolution skeletal muscle diffusion tensor imaging using single-shot echo planar imaging with stimulated-echo preparation and sensitivity encoding. NMR Biomed. 2012; 25(5): 766–778. 2. Basser PJ. Inferring microstructural features and the physiological state of tissues from diffusion-weighted images. NMR Biomed. 1995; 8(7–8): 333–344. 3. Le Bihan D, Mangin JF, Poupon C, Clark CA, Pappata S, Molko N, Chabriat H. Diffusion tensor imaging: concepts and applications. J. Magn. Reson. Imaging 2001; 13(4): 534–546. 4. Heemskerk AM, Strijkers GJ, Vilanova A, Drost MR, Nicolay K. Determination of mouse skeletal muscle architecture using threedimensional diffusion tensor imaging. Magn. Reson. Med. 2005; 53(6): 1333–1340. 5. Schwenzer NF, Steidle G, Martirosian P, Schraml C, Springer F, Claussen CD, Schick F. Diffusion tensor imaging of the human calf muscle: distinct changes in fractional anisotropy and mean diffusion due to passive muscle shortening and stretching. NMR Biomed. 2009; 22(10): 1047–1053. 6. Deux JF, Malzy P, Paragios N, Bassez G, Luciani A, Zerbib P, RoudotThoraval F, Vignaud A, Kobeiter H, Rahmouni A. Assessment of calf muscle contraction by diffusion tensor imaging. Eur. Radiol. 2008; 18(10): 2303–2310. 7. Zaraiskaya T, Kumbhare D, Noseworthy MD. Diffusion tensor imaging in evaluation of human skeletal muscle injury. J. Magn. Reson. Imaging 2006; 24(2): 402–408.


NMR Biomed. 2014; 27: 146–157

TURBO-STEAM DTI IN THE MUSCLE OF RABBIT SHANK 8. Heemskerk AM, Drost MR, van Bochove GS, van Oosterhout MF, Nicolay K, Strijkers GJ. DTI-based assessment of ischemia-reperfusion in mouse skeletal muscle. Magn. Reson. Med. 2006; 56(2): 272–281. 9. Heemskerk AM, Strijkers GJ, Drost MR, van Bochove GS, Nicolay K. Skeletal muscle degeneration and regeneration after femoral artery ligation in mice: monitoring with diffusion MR imaging. Radiology 2007; 243(2): 413–421. 10. Khalil C, Budzik JF, Kermarrec E, Balbi V, Le Thuc V, Cotten A. Tractography of peripheral nerves and skeletal muscles. Eur. J. Radiol. 2010; 76(3): 391–397. 11. Damon BM, Ding Z, Anderson AW, Freyer AS, Gore JC. Validation of diffusion tensor MRI-based muscle fiber tracking. Magn. Reson. Med. 2002; 48(1): 97–104. 12. Sinha S, Sinha U, Edgerton VR. In vivo diffusion tensor imaging of the human calf muscle. J. Magn. Reson. Imaging 2006; 24(1): 182–190. 13. Sinha S, Sinha U. Reproducibility analysis of diffusion tensor indices and fiber architecture of human calf muscles in vivo at 1.5 Tesla in neutral and plantarflexed ankle positions at rest. J. Magn. Reson. Imaging 2011; 34(1): 107–119. 14. Lansdown DA, Ding Z, Wadington M, Hornberger JL, Damon BM. Quantitative diffusion tensor MRI-based fiber tracking of human skeletal muscle. J. Appl. Physiol. 2007; 103(2): 673–681. 15. Karampinos DC, King KF, Sutton BP, Georgiadis JG. Myofiber ellipticity as an explanation for transverse asymmetry of skeletal muscle diffusion MRI in vivo signal. Ann. Biomed. Eng. 2009; 37(12): 2532–2546. 16. Heemskerk AM, Sinha TK, Wilson KJ, Ding Z, Damon BM. Repeatability of DTI-based skeletal muscle fiber tracking. NMR Biomed. 2010; 23 (3): 294–303. 17. Hiepe P, Reichenbach JR. Functional muscle MRI in human calf muscle using IVIM. Proceedings of the 19th Annual Meeting ISMRM, Montreal, QC, Canada, 2011; 2013. 18. Kan JH, Heemskerk AM, Ding Z, Gregory A, Mencio G, Spindler K, Damon BM. DTI-based muscle fiber tracking of the quadriceps mechanism in lateral patellar dislocation. J. Magn. Reson. Imaging 2009; 29(3): 663–670. 19. Budzik JF, Le Thuc V, Demondion X, Morel M, Chechin D, Cotten A. In vivo MR tractography of thigh muscles using diffusion imaging: initial results. Eur. Radiol. 2007; 17(12): 3079–3085. 20. Froeling M, Nederveen AJ, Heijtel DF, Lataster A, Bos C, Nicolay K, Maas M, Drost MR, Strijkers GJ. Diffusion-tensor MRI reveals the complex muscle architecture of the human forearm. J. Magn. Reson. Imaging 2012; 36(1): 237–248. 21. Gold GE, Suh B, Sawyer-Glover A, Beaulieu C. Musculoskeletal MRI at 3.0 T: initial clinical experience. Am. J. Roentgenol. 2004; 183(5): 1479–1486. 22. Stanisz GJ, Odrobina EE, Pun J, Escaravage M, Graham SJ, Bronskill MJ, Henkelman RM. T1, T2 relaxation and magnetization transfer in tissue at 3 T. Magn. Reson. Med. 2005; 54(3): 507–512. 23. Kim S, Chi-Fishman G, Barnett AS, Pierpaoli C. Dependence on diffusion time of apparent diffusion tensor of ex vivo calf tongue and heart. Magn. Reson. Med. 2005; 54(6): 1387–1396. 24. Sigmund EESD, Hodnett PA, Liu K, McGorty KA, Mechlin M, Bencardino J. T2-weighted imaging and stimulated echo diffusion tensor imaging in chronic exertional compartment syndrome calf muscle. Proceedings of the 19th Annual Meeting ISMRM, Montreal, QC, Canada, 2011; 3258. 25. Steidle G, Schick F. Echoplanar diffusion tensor imaging of the lower leg musculature using eddy current nulled stimulated echo preparation. Magn. Reson. Med. 2006; 55(3): 541–548. 26. Burstein D. Stimulated echoes: description, applications, practical hints. Concepts Magn. Reson. 1996; 8: 269–278. 27. Merboldt KD, Hänicke W, Frahm J. Diffusion measurements using stimulated echoes. In: Le Bihan D, (ed.). Diffusion and Perfusion Magnetic Resonance Imaging – Applications to Functional MRI. Raven Press: New York, 1995, pp. 28–41. 28. Hatakenaka M, Soeda H, Okafuji T, Yabuuchi H, Shiokawa S, Nishimura J, Honda H. Steroid myopathy: evaluation of fiber atrophy with T2 relaxation time--rabbit and human study. Radiology 2006; 238(2): 650–657.

29. Sigmund EE, Sui D, Ukpebor O, Baete S, Fieremans E, Babb JS, Mechlin M, Liu K, Kwon J, McGorty K, Hodnett PA, Bencardino J. Stimulated echo diffusion tensor imaging and SPAIR T(2)-weighted imaging in chronic exertional compartment syndrome of the lower leg muscles. J. Magn. Reson. Imaging 2013. DOI: 10.1002/jmri.24060 [Epub ahead of print]. 30. Hiepe P, Herrmann KH, Ros C, Reichenbach JR. Diffusion weighted inner volume imaging of lumbar disks based on turbo-STEAM acquisition. Z. Med. Phys. 2011; 21(3): 216–227. 31. Frahm J, Haase A, Matthaei D. Rapid NMR imaging of dynamic processes using the FLASH technique. Magn. Reson. Med. 1986; 3(2): 321–327. 32. Finsterbusch J, Frahm J. Half-Fourier single-shot STEAM MRI. Magn. Reson. Med. 2002; 47(3): 611–615. 33. Griswold MA, Jakob PM, Heidemann RM, Nittka M, Jellus V, Wang J, Kiefer B, Haase A. Generalized autocalibrating partially parallel acquisitions (GRAPPA). Magn. Reson. Med. 2002; 47(6): 1202–1210. 34. Sled JG, Zijdenbos AP, Evans AC. A nonparametric method for automatic correction of intensity nonuniformity in MRI data. IEEE Trans. Med. Imaging 1998; 17(1): 87–97. 35. Hahn K, Prigarin S, Hasan KM. Fitting of two-tensor models without ad hoc assumptions to detect crossing fibers using clinical DWI data. Magn. Reson. Imaging 2012; 31(4): 585–595. DOI: 10.1016/j. mri.2012.10.016. 36. Hahn K, Prigarin S, Rodenacker K, Hasan KM. Denoising for diffusion tensor imaging with low signal to noise ratios: method and Monte Carlo validation. Int. J. Biomath. Biostat. 2010; 1: 63–81. 37. Jenkinson M, Smith S. A global optimisation method for robust affine registration of brain images. Med. Image Anal. 2001; 5(2): 143–156. 38. Mori S, Crain BJ, Chacko VP, van Zijl PC. Three-dimensional tracking of axonal projections in the brain by magnetic resonance imaging. Ann. Neurol. 1999; 45(2): 265–269. 39. Block KT, Frahm J. Radial single-shot STEAM MRI. Magn. Reson. Med. 2008; 59(4): 686–691. 40. Rossbach M, Jochimsen TH, Herrmann KH, Ros C, Güllmar D, Reichenbach JR. Comparison of long-axis and short-axis PROPELLER-EPI for diffusion-weighted magnetic resonance imaging. Z. Med. Phys. 2011; 21(1): 33–41. 41. Pipe JG. Motion correction with PROPELLER MRI: application to head motion and free-breathing cardiac imaging. Magn. Reson. Med. 1999; 42(5): 963–969. 42. Jiang H, Golay X, van Zijl PC, Mori S. Origin and minimization of residual motion-related artifacts in navigator-corrected segmented diffusion-weighted EPI of the human brain. Magn. Reson. Med. 2002; 47(4): 818–822. 43. Gilbert RJ, Napadow VJ. Three-dimensional muscular architecture of the human tongue determined in vivo with diffusion tensor magnetic resonance imaging. Dysphagia 2005; 20(1): 1–7. 44. Damon BM. Effects of image noise in muscle diffusion tensor (DT)MRI assessed using numerical simulations. Magn. Reson. Med. 2008; 60(4): 934–944. 45. Karampinos DC, King KF, Sutton BP, Georgiadis JG. Intravoxel partially coherent motion technique: characterization of the anisotropy of skeletal muscle microvasculature. J. Magn. Reson. Imaging 2010; 31(4): 942–953. 46. Stocker T, Kaffanke J, Shah NJ. Whole-brain single-shot STEAM DTI at 4 Tesla utilizing transverse coherences for enhanced SNR. Magn. Reson. Med. 2009; 61(2): 372–380. 47. Galban CJ, Maderwald S, Uffmann K, de Greiff A, Ladd ME. Diffusive sensitivity to muscle architecture: a magnetic resonance diffusion tensor imaging study of the human calf. Eur. J. Appl. Physiol. 2004; 93(3): 253–262. 48. Andersen JL. Muscle fibre type adaptation in the elderly human muscle. Scand. J. Med. Sci. Sports 2003; 13(1): 40–47. 49. Lieber RL, Blevins FT. Skeletal muscle architecture of the rabbit hindlimb: functional implications of muscle design. J. Morphol. 1989; 199(1): 93–101. 50. Hiepe PGD, Ros C, Siebert T, Gussew A, Rzanny R, Reichenbach JR. Functional DTI in voluntary contracted human calf muscles using an MR compatible ergometer. Proceedings of the 20th Annual Meeting ISMRM, Melbourne, Australia, 2012; 1860.

157

NMR Biomed. 2014; 27: 146–157



Comparison of twice refocused spin echo versus stimulated echo diffusion tensor imaging for tracking muscle fibers.

Spin echo versus stimulated echo diffusion tensor imaging of the in vivo human heart.

Diffusion tensor imaging focusing on lower cervical spinal cord using 2D reduced FOV interleaved multislice single-shot diffusion-weighted echo-planar imaging: comparison with conventional single-shot diffusion-weighted echo-planar imaging.

Multiple echo multi-shot diffusion sequence.

Differential diagnosis of pancreatic cancer by single-shot echo-planar imaging diffusion-weighted imaging.

Readout-segmented echo-planar imaging in diffusion-weighted MR imaging of acute infarction of the brainstem and posterior fossa: comparison of single-shot echo-planar diffusion-weighted sequences.

Reduced Field-of-View Diffusion-Weighted Magnetic Resonance Imaging of the Pancreas: Comparison with Conventional Single-Shot Echo-Planar Imaging.

Subcortical White Matter Changes with Normal Aging Detected by Multi-Shot High Resolution Diffusion Tensor Imaging.

Simultaneous Multislice Echo Planar Imaging With Blipped Controlled Aliasing in Parallel Imaging Results in Higher Acceleration: A Promising Technique for Accelerated Diffusion Tensor Imaging of Skeletal Muscle.

Oscillating gradient spin-echo (OGSE) diffusion tensor imaging of the human brain.

Second-order motion-compensated spin echo diffusion tensor imaging of the human heart.

Image quality assessment of single-shot turbo spin echo diffusion-weighted imaging with parallel imaging technique: a phantom study.

Breath-hold black blood quantitative T1rho imaging of liver using single shot fast spin echo acquisition.

Diffusion-weighted imaging of the sellar region: a comparison study of BLADE and single-shot echo planar imaging sequences.

Peripheral neuropathy: detection with diffusion-tensor imaging.

Correction for Eddy Current-Induced Echo-Shifting Effect in Partial-Fourier Diffusion Tensor Imaging.

Diffusion tensor imaging of the brainstem in children with achondroplasia.

Feasibility of Diffusion Tensor Imaging at 1.5T Using Multi-Band Echo Planar Acquisition.

Anisotropic fractional diffusion tensor imaging.

Radiation injury to the normal brain measured by 3D-echo-planar spectroscopic imaging and diffusion tensor imaging: initial experience.

Visualization of kidney fibrosis in diabetic nephropathy by long diffusion tensor imaging MRI with spin-echo sequence.

Longitudinal diffusion tensor imaging in frontotemporal dementia.

Diffusion tensor imaging in hemorrhagic stroke.

Diffusion tensor imaging in musculoskeletal disorders.