Single- and Bi-component T2* Analysis of Tendon Before and During Tensile Loading, Using UTE Sequences

Background: To determine if the application of tensile force alters the single- or bi-component T2* values of human tendons as measured on a clinical MRI scanner with ultrashort echo time (UTE sequences and if single-or bi-component T2* values differ when measured with 2D-UTE, 3D-UTE, or 3D-UTE-Cones sequences. Methods: Ten tendons were imaged before and during the application of tension using various UTE sequences at 3 Tesla. Single and bi-component T2* analysis was performed pre- and posttension and compared with Bonferroni-corrected paired Wilcoxon tests. Results: Range of mean pre- and posttension T2* analy- sis values were: short T2* fraction (78.6–79.7% and 77.3– 79.7%, respectively; P ¼ 1.0 for all sequences), long T2* fraction (20.3–21.4% and 20.3–22.7%, respectively; P ¼ 1.0 for all sequences), short T2* (0.9–1.0 ms and 0.9 ms, respectively; P ¼ 1.0 for all sequences), long T2* (19.9–20.4 ms and 21.9–24.0 ms, respectively; P ¼ 0.9 for 2D-UTE and P ¼ 1.0 for 3D-UTE and 3D-UTE-Cones), and single-component T2* (2.3–2.5 ms and 2.5–3.2 ms, respectively; P ¼ 1.0 for all sequences). Conclusion: No signiﬁcant difference in single- or bi- component results was found after the application of tension to tendons. Results are similar regardless of UTE sequence used for

TENDONS TRANSMIT THE force generated by muscles to bone to move the skeleton. The composition of human tendons varies, but those that are subject to primarily tensile stresses such as the Achilles, are composed of approximately 66% water (1) and the dry weight is composed of approximately 87% collagen and 0.2% glycosaminoglycans (2). The ultrastructure of tendon is complex and collagen fibrils not only course in the longitudinal direction, but are also oriented transversely and can form spirals (3). A unique combination of structure and composition confers great strength and a tendon with a cross-section area of 1 cm 2 is capable of supporting a weight up to 1000 kg (4).
MRI is a useful method for studying tendons and with the advent of ultrashort echo time (UTE) and similar sequences, signal can be detected in tissues with short mean T2*s (5). With UTE sequences, quantitative MR imaging can be readily performed on tendons. Studies have shown that T2* analysis using single and bi-component analysis may be useful in distinguishing between normal and pathologic tendons (6)(7)(8). This has been hypothesized to reflect the known histologic alterations of disruption and disorganization of collagen fibers as well as an increase in total water content (1,9). Specifically, bi-component T2* analysis may reflect the water bound to macromolecules (short component) and free water (long component) in highly organized biological tissues (7,10). Compared with single-component T2* analysis which linearly combines the short and long components, bicomponent analysis provides more information which has been shown to be beneficial in evaluating disease states (7).
However, structural and compositional alteration does not only occur in pathologic states. During muscle contraction, tendons are capable of a small amount of elongation (11), which may correspond to straightening of tendon crimps on the microscopic scale (12,13). It is also known from experimental data that tendons under tension decrease their total water content compared with a slack condition (14,15). Despite these changes, statistically significant differences in transverse relaxation times between tendons under different loading states have not been shown in a previous study (16). However, this previous study used a minimum echo time (TE) of 5 ms and analyzed T2 with single component analysis.
The purpose of this study was to determine if the application of tensile force alters the single-or bicomponent T2* values of human tendons as measured on a clinical MRI scanner with UTE sequences and if single-or bi-component T2* values differ when measured with 2D-UTE, 3D-UTE, or 3D-UTE-Cones sequences.

Sample Preparation
This anonymized cadaveric study was exempted from the Institutional Review Board. 10 tendon samples were harvested from two donor ankles (2 females, ages 80 and 86 years old), including the tibialis anterior, tibialis posterior, flexor digitorum longus, flexor hallicus longus, and peroneus longus tendons. Specimens underwent a single freeze-thaw cycle, which occurred before dissection.
Tendon samples were cut to a length between 4 and 6 cm. A Krackow stitch (17) was tied to both ends using 2-0 nylon suture (Fig. 1). The Krackow stitch is significantly stronger than simple suture fixation (18) and when both limbs are equalized, the core and periphery of the tendon can be tensioned. Samples were immersed in normal saline for 12 h at 4 C, gently blotted dry, and placed into sealed plastic syringes filled with Fomblin to minimize dehydration and susceptibility artifact (19). Sutures were fixed to one end of the syringe by a red Luer Lock cap and long suture tails were extended out from the opposite plunger end. The rubber stopper maintained the seal and care was taken to ensure tendons were entirely parallel to the long axis of the syringe. At the time of imaging, tendon specimens had been at room temperature for at least 4 h.

Tensile Loading and MR Imaging Protocol
Specimens were placed parallel to the B 0 field and imaged in the axial plane on a clinical 3 Tesla (T) MR scanner (Signa HDx, GE Healthcare Technologies, Milwaukee, WI) which had gradients capable of a slew rate of 150 T/m/s and amplitude of 40 mT/m on each axis. Hardware modifications included the addition of a custom transmit-receive switch to the receiver preamplifiers for rapid switching at the end of a radiofrequency excitation pulse to allow ultrashort echo time (UTE) imaging. A wrist coil (BC-10, Medspira, Minneapolis, MN) was used.
For each tendon, one suture end was fixed to the table and the other suture end was extended out of the bore of the magnet. The entire imaging protocol was performed under two conditions, the first with no tensile load applied to the tendon and the second with a 1 liter water bottle (1 kilogram) tied to the mobile suture end at the edge of the table (Fig. 2). A 1 kilogram weight was chosen as a maximum load to fall safely within the tensile strength boundaries of the suture (knotted pull) as rated by the manufacturer (Ethicon, Johnson & Johnson, Cincinnati, OH). The first sequence in the protocol was started 10 min after the load was applied. The majority of tendon lengthening (elastic response) was expected to occur by the time scanning began. However, as other studies have shown (20,21), viscoelastic creep would be expected to continue to elongate our tendon samples to a small degree during scanning. To ensure comparison of similar regions, the midportion of each tendon was used in the analysis (see the Image Analysis section below). A rubber eraser was placed in the field of view of each scan as an external control.
The tendons were imaged in the axial plane with a quantitative imaging protocol consisting of 2D-UTE, 3D-UTE, and 3D-UTE-Cones (22,23) with sequence parameters as shown in Table 1. For each technique, the repetition time (TR) was constant and each  acquisition provided four echoes. In total, 16 echo times (TEs) were acquired ranging from 30 ms to 30 ms to cover the expected T2* values of tendon.

Image Analysis
The images were qualitatively evaluated for the presence of saline-Fomblin fluid levels before and after the application of tension (E.Y.C., 4 years of experience in musculoskeletal radiology). Specifically, a positive fluid level was denoted if saline exuded from the tendon and dependently layered in the Fomblin-filled syringe. All images were reconstructed in multiple planes. The approximate longitudinal tendon lengths before and after tension were recorded on reformatted 3D-UTE images (0.03 ms TE) with electronic calipers because the isotropic voxels would provide the most accurate measurements. The 3D-UTE sequence began approximately 70 min after the onset of tensile loading. Percent change was calculated by: (L 1 -L 0 )/L 1 , where L 0 represents the pretension length and L 1 represents the posttension length on 3D-UTE images.
Using axial images at the midportion of each tendon in the longitudinal axis, regions of interest (ROIs) were carefully placed within the boundaries of the epitenon on the shortest TE image and copied to the corresponding position on subsequent TE images. The mean intensity within each ROI was used for quantitative curve fitting. Single and bi-component T2* analysis was performed using a semiautomated MATLAB (The Mathworks Inc., Natick, MA) code developed inhouse as previously described (24,25). In singlecomponent analysis, the UTE signals S N (t) were fitted with the following equation: S N (t) ¼ A Â exp(Àt/ T2*) þ noise. In bi-component analysis, the UTE signals S N (t) were fitted with the following commonly used model: where A S is the amplitude of the short component, A L is the amplitude of the long component, T2* S is the short component T2*, and T2* L is the long component T2*. Apparent short component fraction was defined as A S /(A S þ A L ). Background noise was estimated using a maximum likelihood estimation distribution fitting of a partial histogram and nonnegative least square curve fitting was used for both single and bi-component models. Fit curves along with their 95% confidence intervals and residual signal curves were created (25). Additionally, single component T2* maps were reconstructed on a pixel-by-pixel basis using various color scales which included all identifiable structures in the field of view, including the tendon, structures attached to the outer surface of the tendon (including paratenon (26) or tendon sheath), and surrounding fluid.

Statistical Analysis
Statistical analyses were performed using the SPSS software package (version 21; SPSS, Chicago, IL). Descriptive statistics were performed for each condition and imaging sequence. The paired Wilcoxon Rank-Sum test was used to compare data before and after the application of tension. Additionally, the paired Wilcoxon Rank-Sum test was used to determine if there were differences between sequences for both the pretension and posttension conditions. Bonferroni correction was performed for multiple comparisons and corrected P < 0.05 was considered significant.

RESULTS
The presence of dependent saline-Fomblin levels in the syringe was noted after the application of tension for each sample (Fig. 3). Each tendon lengthened after the application of tensile load with a mean increase in length of approximately 4.1% (median 3.7%, range 1.4-7.4%).
Curve fitting using the bi-component algorithm was superior compared with the single-component algorithm for all tendons with less than 2% systematic residual signal for each TE for all sequences (Fig. 4). The rubber eraser, included as a reference on all scans, showed mono-exponential decay and demonstrated a single-component T2* value of exactly 0.3 ms on all sequences for both the pretension and posttension conditions (P ¼ 1.0), confirming perfect reproducibility. Single component T2* maps showed no appreciable difference between pre-and posttension conditions for each specimen (Fig. 5).
Quantitative results are summarized in Table 2. No significant difference was found for any measurement when compared between the pretension and  posttension conditions (P > 0.05). Furthermore, no significant difference was found between the 2D-UTE and the 3D-UTE derived measurements, 2D-UTE and 3D-UTE-Cones derived measurements, and 3D-UTE and 3D-UTE-Cones derived measurements for either the pretension (Table 3) or posttension (Table 4) conditions (all P values > 0.05).

DISCUSSION
In this pilot study we imaged tendons before and after the application of tension using various UTE techniques and were unable to find a statistically significant difference between single and bi-component T2* analysis values obtained without and with tension. We were able to visualize the dependent layering of saline which was exuded from the tendon under the loaded conditions, which is consistent with results obtained from prior studies using tendon wet weights (14,15). It is also well known that tendons under load demonstrate loss of crimping and straightening of collagen fibers (13,(27)(28)(29). The combination of water loss and collagen fiber straightening with tension would have been expected to shorten single-component T2* relaxation times (28) or decrease long T2* bi-component fractions, but was not observed in our samples.
There are a limited number of studies involving MRI measures of tendons under varying static load. However our results are consistent with those by Wellen et al (16), who studied rabbit Achilles tendons under two states of tension including loading with 40 g and 500 g masses. Wellen et al found no significant difference in T2 values between the two loading conditions (16). Of note, Wellen included separate "core" and "rim" measurements and did find nonstatistical trends of T2 differences between the two locations which suggested water transport from tendon core to rim with loading. The regions of interest used in the quantitative portion of our study correspond to the "core" measurements in the study by Wellen et al. We believe that precise segmentation of the outer epitenon rim would be difficult in clinical practice due to surrounding macroscopic fat and other long T2* structures, such as the paratenon and synovial tendon sheaths. Inadvertent inclusion of these structures would confound the results of tendon analysis. However, in our study we did qualitatively evaluate T2* maps which included the "rim" and did not find any discernible difference between loading states.
A secondary objective of our study was to determine the reproducibility of various UTE techniques for the analysis of T2* in tendon. Our study shows that reproducible quantitative measurements can be made  with the 2D-UTE, 3D-UTE, or 3D-UTE-Cones techniques. Our results are similar to some previous studies using UTE techniques with center-out k-space acquisitions and single-component (30) and bi-component analysis (10). Other studies have shown slightly different results, including a study by Juras et al who used a variable echo time (vTE) sequence with bicomponent analysis on in vivo Achilles tendons and found a range of mean values including: 56-62% short fraction, 38-44% long fraction, 0.6-1.0 ms short T2*, and 13.5-23.4 ms long T2* (7). In comparison, range of mean values of our samples was: 77-80% short fraction, 20-23% long fraction, 0.9-1.0 ms short T2*, and 19.9-24.0 ms long T2*. Possible explanations for why our samples showed higher mean short fraction include differences in technique (radial versus Cartesian k-space trajectories), differences between living and postmortem tissue, structural and functional differences between tendons, and subject variability.
Zheng and Xia (31) and Wang and Xia (32) have previously highlighted the importance of experimental details for the measurement of multi-component transverse relaxation, including signal-to-noise ratio (SNR), echo spacing, shimming, pulse accuracy, and specimen details. Comparison of the results in this current study with those from two recent studies by Chang et al (19,33), reinforce the important concept that sequence parameters influence bi-component results. With the 2D-UTE parameters used in the previous studies (19,33), bi-component analyses was either unable to be reliably performed or yielded long T2* fractions of less than 5%. In comparison, our current study used the same MRI machine (with the same field strength), specimen orientation, range of echo times, bi-component fitting algorithm, and similar resolution (0.12 mm 3 voxel size in the previous studies compared with 0.06-0.27 mm 3 voxel size in our current study). Additionally, when we performed retrospective analysis of our data and discarded 4 echoes to approximate the previously used 12 echo times, our bi-component results were unchanged. Other possible explanations for differences in results include differences in SNR, usage of different tendons, or specimen variation. The effect of these and other variables on bi-component quantification using clinical magnets clearly deserves additional study.
Our results should not be compared with other MR measures of tendon under different loading conditions. For instance, Wellen et al (16) found differences in diffusion measurements between tendons under different loading conditions and Helmer et al (34) found differences in proton density. Additionally, recently Syha et al (35) have shown that short-term exercise (consisting of 15 min of high-intensity rope skipping or a cross-country run of 6.6 km) can result in MR-detectable differences in tendons of healthy volunteers using 3D-UTE off-saturation ratio. A decrease in tendon volume was also observed and the combined findings were consistent with a loss of free water molecules within the tendon (35). These results suggest that magnetization transfer techniques may be more sensitive to the short-term water changes in tendon.
Our study has several limitations. First, we included only a small number of samples. As is the nature of pilot studies, our results require external validation with larger sample sizes. Second, tendons can vary widely in structure and function throughout the body (36) and our study results may not be applicable to other tendons such as the Achilles tendon. Third, we did not precondition our tendons or standardize tendon strain despite having a fixed load. Additionally, despite the use of a fixed weight, the force transmitted to the tendon may have been lowered due to friction caused by the suture running between the syringe and plunger. Also, we did not quantify viscoelastic creep which would have caused continuing lengthening of the tendon during the scan protocol. However, to ensure comparable regions, we performed our measurements in the midportion of each tendon (in the longitudinal axis). Fourth, we did not measure the total amount of water in the tendon or water loss with tension. Finally, the imaging time of the sequences used in this study are clearly prohibitively long for clinical translation. However, our protocols were not optimized for time and future studies will be performed to determine how many different TEs and the SNR that is necessary to obtain reliable bi-component analysis results. Additionally, with the use of single slice 2D-UTE or with faster trajectories such as with the use of 3D-UTE-Cones with anisotropic voxels, imaging time can be considerably reduced. The effect of decreased resolution on bi-component analysis also requires additional study.
In conclusion, we have found that the application of tensile force did not significantly alter the single-or bi-component T2* values of a small sample of human tendons as measured on a clinical MRI scanner with UTE sequences. Additionally, single and bicomponent T2* analysis results are similar regardless of whether a 2D-UTE, 3D-UTE, or 3D-UTE-Cones sequence is used for acquisition.