In vivo MR tractography of thigh muscles using diffusion imaging: initial results
- First Online:
- Cite this article as:
- Budzik, J.F., Le Thuc, V., Demondion, X. et al. Eur Radiol (2007) 17: 3079. doi:10.1007/s00330-007-0713-z
- 348 Views
The aims of this preliminary study were (1) to demonstrate the feasibility of providing in vivo 3D architecture of human thigh muscles using tractography on a 1.5T magnet, and (2) to assess the value of tractography images to obtain averaged microstructural parameters, i.e., the fractional anisotropy (FA) and the mean apparent diffusion coefficient (ADC), over the whole thigh. Five healthy volunteers were included in this study. Their right thighs were imaged using diffusion tensor imaging and gradient-echo T2* sequences. Muscular tractography was performed on each muscle. MR tractography provided a good approach of the muscle shape and of the orientation of the muscle fibers. There was no aberration in the color-encoding scheme nor in the luminosity assigned to each fiber. In contrast, tendons were not drawn in any of the muscles studied. FA values ranged from 0.27 to 0.38. Mean ADC values ranged from 0.76 to 0.96 × 10−3 mm2/s. Our study demonstrated the feasibility of providing in vivo 3D architecture of human thigh muscles using tractography on a 1.5T magnet, and of determining muscular microstructural parameters (FA and ADC). Musculoskeletal radiologists should be aware of these new developments that may provide complementary information on muscles to the usual sequences.
KeywordsThighDiffusion tensor imagingFiber trackingMuscleMRI
Magnetic resonance (MR) measurement of an effective diffusion tensor of water in tissues can provide clinically relevant information that is not available from other imaging modalities . This information includes parameters that help to characterize physical properties of tissue constituents, tissue microstructure and architectural organization . With diffusion tensor imaging (DTI) measurements related to molecular motion of water in space, it is possible to have information about the tissue orientation. This technique is known as MR fiber tracking or tractography . DTI and tractography have found clinical application in neuroradiology , but their use in the musculoskeletal field is still emerging.
Assessment of the architecture of striated muscle by DTI has been reported in several in vitro studies  or in anesthetized mice  or rats , where high SNR sequences with long acquisition times could be used. The use of DTI in humans requires fast sequences. Echo-planar diffusion-weighted sequencing for the determination of the DT in a single slice through the calf muscle has been recently reported [7–10]. Cross-sectional area measurements of fractional anisotropy (FA) and mean apparent diffusion coefficient (ADC) values obtained by positioning a region of interest (ROI) allow the assessment of only a small part of the muscle at a fixed level. Consequently, the analysis of a whole muscle implies drawing a ROI on each section of the muscle, which is time-consuming. However, tractography may represent an interesting tool to extract the muscular fibers from a single ROI and thus easily obtain the values of FA and mean ADC of these fibers. To the best of our knowledge, tractography of the striated muscles of healthy volunteers has been reported in two studies performed on human calf muscles on a 3T magnet, but no microstructural parameters were calculated using this technique [8, 10].
The aims of this preliminary study were (1) to demonstrate the feasibility of providing in vivo 3D architecture of human thigh muscles using tractography on a 1.5T magnet, and (2) to assess the value of tractography images to obtain averaged microstructural parameters, i.e., the FA and the mean ADC, over the whole thigh.
Materials and methods
Five volunteers, three men and two women, age range 25–30 years, participated in this study, which was approved by the local Ethics Committee Review Board. Each volunteer provided informed consent. All MRI scans were performed on a 1.5T full-body scanner (Achieva, Philips Medical Systems, Best, The Netherlands). The volunteers were in supine position, with legs in a relaxed state, placed parallel to the magnetic field, feet first. A SENSE Body Coil (phased array coil with four elements) was used to scan the right thigh of each subject.
We used a spin-echo single-shot diffusion-weighted EPI sequence with fat suppression. The DTI acquisition consisted of one b0 image and DT images using a b factor of 400 s/mm2 with 32 directions. A frequency-selective fat saturation was used to suppress the fat signal (SPAIR). The image acquisition parameters for the diffusion images were as follows: TE = 60 ms, TR = 6,277 ms, FOV = 200 mm, parallel imaging technique was used with a sense factor of 2, partial Fourier acquisition (half-scan factor = 0.681), recon matrix = 128 × 128, phase encoding AP. Bandwith in frequency direction was 1,367.5 Hz per pixel. Thirty-six transversal slices were acquired without spacing. The in-plane image resolution was 1.56 × 1.56 mm with a slice thickness of 8 mm. Five independent series were performed successively. The total acquisition lasted 18 min 50 s.
Eddy currents and motion-related misalignment of the diffusion tensor MR images were corrected off-line using Automated Image Registration (Philips Pride Diffusion Registration and IDL, ITT, Boulder, CO, USA). All diffusion-weighted images were reoriented to match the b0 images. Each series was then reoriented to match the others (affine transformation). The final step was averaging of the five reoriented acquisitions. This made it possible to improve SNR and to minimize artifacts.
Fractional anisotropy maps
FA maps were calculated using Pride FiberTracking. The diffusion tensors at each voxel were diagonalized so as to obtain eigenvalues and eigenvectors for each voxel. The eigenvector associated with the largest eigenvalue was used to represent the local fiber direction. FA maps were calculated from the eigenvalues on the basis of standard formulas . Thigh color maps were created on the basis of the three vector elements for each voxel. The absolute values of the vector elements were assigned to red (right-left component), green (antero-posterior component), and blue (feet-head component). The intensity of color in each voxel was gauged by the degree of FA. This color information was combined with the results of further fiber tractography.
Muscular fiber tracking
Fiber tracking was performed using a line propagation technique from the Philips Pride FiberTracking. Tracking was launched from a seed region of interest from which a line was propagated in both the retrograde and anterograde directions according to the main eigenvector at each voxel . The ROI was drawn manually on a T2*-weighted image on the whole cross-sectional area of the muscle chosen, at a level corresponding to the mid-part of the muscle belly, where the muscular boundaries were clearly identified. We used one ROI per muscle, except in the case of the vastus medialis and lateralis, where up to five ROI were drawn, as the number of fibers obtained in this manner was not sufficient to visualize the full length of the muscles. All the DTI images were performed by the same radiologist.
Signal-to-noise ratio measurements
As signal-to-noise ratio (SNR) is a critical point in fiber tracking, we chose to evaluate it in the tendons as well as in the muscles using the technique described by Price et al. . The SNR was calculated as follows: two scans were acquired with the same parameters on the same subject (registration was also performed for each scan). The two scans were then substracted to produce a “noise image” at b = 0 and b = 400. An ROI was drawn in the vastus intermedius muscle (the size of the ROI was 142 pixels). The mean value of the signal in the ROI was calculated on the b = 0 and b = 400 images. The standard deviation of the signal in the “noise image” was calculated using the same ROI. The same procedure was carried out for the quadriceps tendon.
Anatomical data from dissection of cadavers were provided to optimize the understanding of the tractography images obtained in the volunteers. The specimens were obtained and used according to institutional guidelines. Anatomic images are only provided for pedagogic reasons.
Average FA and mean ADC values with their standard deviation for each muscle
ADC (× 10−3 mm2/s)
0.91 ± 0.26
0.35 ± 0.09
0.76 ± 0.23
0.35 ± 0.11
0.92 ± 0.20
0.29 ± 0.08
0.96 ± 0.24
0.28 ± 0.08
0.87 ± 0.21
0.27 ± 0.08
0.92 ± 0.21
0.27 ± 0.09
0.96 ± 0.26
0.28 ± 0.09
Biceps femoris short head
0.90 ± 0.29
0.38 ± 0.11
Biceps femoris long head
0.89 ± 0.31
0.29 ± 0.10
0.88 ± 0.31
0.30 ± 0.09
0.85 ± 0.30
0.34 ± 0.10
Our study showed that MR tractography results provide a rather good approach for determining the shape and fiber orientation of the thigh muscles in each volunteer. The changes in color related well to the known variations in the direction of the muscle fibers, also demonstrated in the anatomical dissections. However, one must keep in mind that fibers obtained by tractography are no more than a mathematical representation of a physical phenomenon, and they do not necessarily correspond to the histological muscle fibers.
SNR values in the tendons and in the muscles. [SNR = mean value × sqrt(2)/SD(noise)]
b = 0
b = 400
Standard deviation of the noise
Mean of the ROI in tendon
Standard deviation of the noise
Mean value of the ROI in muscle
Several recent studies performed in the mid-calf region of volunteers reported that cross-sectional area measurements of FA and/or mean ADC values can be obtained by positioning an ROI on the DTI images [7–10, 14]. However, this method allows only a small volume of the muscle at a fixed level to be considered . In this regard, tractography may represent an interesting tool for the analysis of muscular microstructure as nearly the entire length of the fibers is studied, allowing the extraction of the muscular fibers from a single ROI. Therefore, values of FA and mean ADC of these fibers can be easily obtained. Our FA values were very similar to those previously reported in human calf muscles . They were somewhat lower in the human calf muscles studied by Zaraiskaya et al. . In contrast, mean ADC values determined in our study were lower than those reported by other studies on human calf muscles [9, 10, 14]. Further studies have to be undertaken to assess whether these different mean ADC values can be explained by different muscular microstructural properties (thigh versus calf muscles) and/or differences in the techniques used to obtain the values (tractography versus ROI measures).
The feasibility of in vivo MR tractography in human muscles of the thigh using a 1.5T magnet depends on several parameters that have to be optimized.
The combination of partial Fourier acquisition and parallel imaging with SENSE factor 2 reduces the total number of phase-encoded lines, which decreases geometric distortions and susceptibility artifacts. High SNR is essential for proper tracking of the muscular fibers, especially on a 1.5T magnet. In comparison to previous studies performed on 3T [8, 10], we chose a lower b factor and a greater slice thickness. To minimize artifacts, we co-registered each DTI series, then we averaged the co-registered series together. Finally, we used 32 directions along which diffusion-sensitizing gradients were applied in order to improve the accuracy of fiber tracking. These optimized parameters allowed us to obtain good tractography images, but at the cost of a longer time. However, we had to use a larger FOV along the long axis of the thigh muscles (280 mm).
As described in the section “Results,” the ROI drawn on T2* images was consistent with the FA map. The T2* images present a chemical shift in the left-right direction (readout direction). In fact, the bandwith for T2* was fixed at 108.6 Hz per pixel in order to obtain a good SNR; this results in a water-fat shift of two pixels. This shift on T2* images is visible in the bone marrow as well as in the subcutaneous fat, but it is not an obstacle in the precise identification of the muscles, as these structures do not contain a significant amount of fat.
Fractional anisotropy threshold and direction threshold
The FA threshold determines the FA value below which voxels are excluded from the propagation algorithm. In our study, we decided to base our choice of the FA threshold on the reliability of the tracking of the sartorius muscle obtained with different thresholds. We arbitrarily chose this muscle because it is easily recognized on T2*-weighted images and has a simple strap-like shape. Consequently, a value of 0.15 as the FA threshold was used for this study.
The direction threshold determines the value of the maximum angle authorized between the main eigenvectors of two adjacent voxels. A high value excludes slanted fibers, while a low value increases the risk of including nonmuscular structures such as vessels. We decided to assess which direction threshold would be the best by using the same method as in the case of FA. A value of ±7° was chosen. This method of choosing FA and direction thresholds for muscular tractography of the thigh allowed us to obtain excellent results, but further studies should be undertaken in order to confirm whether these parameters are indeed optimal using a 1.5T magnet.
Our study had several limitations. First, tractography imaging was performed by only one radiologist. Inter- and intra-observer reproducibility was not assessed in our study, whose goal was to demonstrate the feasibility of the method and present preliminary results. To the best of our knowledge, such an assessment has not been previously reported in diffusion imaging studies on muscles. The validity of this imaging technique will require assessment of its reproducibility in further studies. Second, we optimized several parameters in order to obtain tractography of muscles on a 1.5T magnet, at the cost of a longer acquisition time. Further studies may demonstrate a better compromise between the quality of tractography and the acquisition time. Third, the use of a single seed was not always enough to obtain a sufficient number of tracked fibers, especially in the case of the vastus medialis and lateralis muscles. We have no clear explanation for this phenomenon. The use of 8-mm slice thickness results in anisotropic voxels, which may be a disadvantage to providing an accurate representation of the complex slanted orientation of some muscle fibers such as the vastus medialis and lateralis. Fourth, we studied only a limited number of volunteers, and further studies are required to confirm the normal muscular FA and mean ADC values, and to look for potential differences between muscles as well as genders and ages, as reported by others .
In conclusion, our study demonstrates the feasibility of providing in vivo 3D architecture of human thigh muscles using tractography on a 1.5T magnet, and muscular microstructural parameters (FA and mean ADC). Musculoskeletal radiologists should be aware of these new developments that may provide complementary information on muscles to the usual sequences.