Abstract
Zero echo time (ZTE) sequences capture signal from tissues with extremely short T2* and are useful for qualitative and quantitative imaging of musculoskeletal tissues’ ultrashort-T2* components. One such sequence is Pointwise Encoding Time Reduction with Radial Acquisition (PETRA). While this sequence has shown promising results, it has undergone only limited testing at 7 tesla (T). The purpose of this work was to evaluate PETRA at 7T in its standard, commercially available form and with sequence code modifications to allow extended echo times for the purpose of performing ultrashort-T2* mapping. We acquired PETRA images of MnCl2 and collagen phantoms and of the knee in eight participants (5 for optimization and 3 for ultrashort-T2* mapping assessment; 5 male/3 female, 39 ± 11 years old). Images were acquired using a 1-transmit/28-receive-channel knee coil. Artifacts, signal, signal-to-noise ratio (SNR), ultrashort-T2*, the corresponding curve fit quality, and repeatability were assessed. In knee tissues, SNR was higher at TE = 0.07 msec than in a conventional-TE sequence (Dual-Echo Steady State with TE = 2.55 msec), with values of 68–337 for PETRA versus 16–30 for the same regions in the conventional-TE series. Acquisition of series for ultrashort-T2* maps was feasible at 1.50 mm isotropic acquisition resolution and TE ≤ 0.58 msec. Strong linear correlations were observed between relaxation rates (R2*) and MnCl2 concentration, and between signal and collagen concentration. Ultrashort-T2* signal decay curve fit R2 and repeatability were high for phantom and knee ultrashort-T2* <1 msec. PETRA imaging with minimal artifacts, high SNR, and scan time < 11 minutes was achieved at 7T at high (0.34 mm isotropic) resolution at TE = 0.07 msec and lower resolution (1.52 mm isotropic) at echo times ≤ 0.58 msec. Ultrashort-T2* mapping provided acceptable curve-fitting results for substances with sub-millisecond T2*.
Introduction
Conventional magnetic resonance imaging (MRI) sequences provide limited visualization of tissues with short/ultrashort T2* values, due to their inability to reduce the echo times (TE) sufficiently to acquire signal from these tissues prior to signal decay. However, ultrashort echo-time (UTE) and zero echo-time (ZTE) MRI sequences are designed to acquire signal from these tissues, enabling both qualitative and quantitative evaluations. In the musculoskeletal system, UTE and ZTE imaging are particularly useful for studies of bone, tendon, calcified cartilage, meniscus, and other organs [1–11]. One ZTE sequence that has begun to be applied to these tissues is Pointwise Encoding Time Reduction with Radial Acquisition (PETRA) [1,4,8,12,13].
The PETRA sequence allows 3D isotropic imaging with TEs under 0.1 msec by applying the readout gradients prior to the radiofrequency pulse and combining radial half-projection filling of the outer portions of k-space with single pointwise Cartesian filling of the central portions of k-space, which are otherwise missed during the transmit/receive switching delay [1,7]. PETRA has been applied at 3T for in vivo imaging of the knee with a single TE with a focus on meniscus [12,14], cartilage [14], and the cruciate [14] and collateral ligaments [14]. The signal-to-noise ratio (SNR) benefits of UTE and ZTE sequences at 7T compared to lower field strengths have been demonstrated and it has been shown that high SNR can be achieved at high resolution (0.83 mm isotropic) for in vivo human limb/joint imaging at 7T [10].
While other ZTE imaging approaches have been applied in human joints at 7T [10,15], the reported applications of PETRA at 7T in musculoskeletal and other ultrashort-T2* tissues have been limited. PETRA has been applied to ex vivo imaging of bovine bone specimens at 7T [4] and is available as a commercial sequence at 7T, but we are unaware of previous literature describing 7T testing and optimization in human joints. Therefore, 7T testing and optimization of the PETRA sequence is needed to evaluate image appearance and potential challenges at this higher field strength.
In addition to minimizing artifacts compared to some other ZTE approaches [4], the Cartesian single point k-space filling approach that PETRA uses allows image acquisition at different k-space gap sizes, corresponding to a range of TEs, as required to perform ultrashort-T2* mapping [1,4]. In addition to single-TE PETRA assessment, it would be beneficial to expand the applications of this sequence to ultrashort-T2* mapping and to assess the performance of such mapping at 7T. As mentioned above, PETRA-based T1ρ relaxometry [13] has been performed by adding a T1ρ preparation module and performing multiple scans at a series of spin-lock times. T1ρ relaxometry provides information about proteoglycan content [13]. Similarly, ultrashort-T2* relaxometry provides quantitative information about tissue degeneration or other changes, potentially related to disruption of normal collagen organization, in ultrashort-T2* tissues [16–18]. Ultrashort-T2* mapping has been shown to be particularly sensitive to tissue changes and tissue properties in ultrashort-T2* tissues such as deep/calcified cartilage [17,18], meniscus [16,19], tendon [20,21], and ligament [22,23]. However, PETRA-based ultrashort-T2* mapping has not yet been assessed.
In this work we present ultrashort-T2* mapping results using a modified version of a commercially available PETRA sequence which allows an increased TE range of 0.07 to 0.58 msec. Using in vivo knee imaging studies, we performed an initial optimization and addressed the appearance of artifacts at the commercially available sequence settings when using the shortest achievable TE. In addition, we evaluated the feasibility of using the modified extended TE range to perform ultrashort-T2* mapping in phantom and in vivo knee images.
Materials and methods
The source code of the product PETRA sequence was modified by extending the maximum allowable TE from 0.10 msec to 1.10 msec and adjusting the related sequence parameters such as repetition time (TR), size of the cartesian section of the k-space, and sequence timing. The increase in TE values required an increase in the size of the inner k-space portion filled with single pointwise Cartesian approach and a reduction in the outer k-space filled using radial sampling. Together, these changes increased the signal acquisition time, and therefore also the TE. We limited our maximum time for a single scan to 12 minutes for participant tolerance and determined that 0.58 was the maximum TE feasible within this scan time (details provided in the next section). The modified PETRA sequence was implemented on a 7T MAGNETOM Terra (Siemens, Erlangen, Germany) to collect phantom and knee data using a 1-transmit/28-receive-channel knee coil (Quality Electrodynamics, Mayfield Village, Ohio, USA).
Multi-echo image optimization for ultrashort-T2* mapping
The multi-echo image optimization goals included achieving the necessary levels of SNR and artifact reduction to allow ultrashort-T2* mapping within clinically feasible scan times, reported as minutes:seconds, at the highest practically achievable spatial resolution within this time limit (S1 Table).
In vivo knee imaging.
Eight healthy study participants participated in the study between November 19, 2022 and June 15, 2024. The study was approved by the institutional review board, and written informed consent was obtained from all participants. For the initial optimization studies, five healthy participants (two male, three female, 42 ± 13 years old) were scanned using the modified PETRA sequence for initial optimization. Dielectric pads (7TNS Neuro Set, Multiwave Imaging, Marseille, France) were wrapped around the anterior and posterior surfaces of the knee. All PETRA scans used a single excitation and a flip angle of 6°. In a phantom, a transmit/receive reference amplitude of 175 V allowed the desired 6° flip angle to be achieved. In human participants, the system transmit/receive reference amplitude was adjusted to 170 V, which was experimentally set at the maximum value that avoided exceeding the specific absorption rate limit. Images were subsequently evaluated to confirm that SNR was at a high level. Initial imaging was performed with spatial resolutions ranging from 0.32 to 1.56 mm isotropic and TR values ranging from 3.99 to 8.28 msec. Data with and without chemical shift selective fat saturation [1] were collected to test the effect of fat suppression on data quality. Optimized parameters were selected for 1) a high-resolution series with a single short TE and for 2) a lower-resolution series that made acquisition at extended TE values up to 0.58 msec feasible, given the allowable scan time threshold of 12 minutes. The scan parameters listed in S1 Table were selected as providing the best balance between desired resolution, contrast, and scan time. For all knees, a conventional-TE DESS sequence (TR = 8.68 msec, TE = 2.55 msec, resolution 0.30 mm isotropic, fat suppressed) was also acquired for comparison and to provide a conventional-contrast anatomical reference.
Ultrashort-T2* mapping
MnCl2 phantom.
A MnCl2 phantom was used to perform initial evaluation of the ultrashort-T2* mapping protocol using scan parameters based on the initial knee optimization work. The MnCl2 phantom contained ten 5 mL centrifuge tubes having 0.02–27.07 mM MnCl2 tetrahydrate, based on target ultrashort-T2* values estimated using reported relaxivity relationships at 3T [24], and 30 mM NaCl in deionized water [25]. The tubes were arranged within a cylindrical plastic container sized to fit within the knee coil. Plastic supports were used to keep the tubes stationary in the central volume of the container.
Two scan sessions were performed. In the first session, the phantom was imaged with the PETRA sequence using: matrix = 96; FOV = 150mm isotropic; 29,000 radial views; voxel size = 1.56 mm isotropic; flip angle = 6°; receive bandwidth = 365 Hz/pixel; fat suppressed; TR = 7.13 msec; and TE values of 0.07, 0.09, 0.12, 0.16, 0.20, 0.24, 0.30, 0.40, and 0.50 msec. To test within-day repeatability, the TE = 0.07 and 0.50 msec scans were repeated. Acquisition time ranged from 4–5 minutes for TE = 0.07–0.40 msec and was 7 minutes for TE = 0.50 msec. The second session replicated the first session, apart from repeating the TE = 0.07 and 0.50 msec scans.
Collagen phantom.
The collagen phantom contained nine 5 mL centrifuge tubes filled with water-collagen solutions having different collagen (chicken hydrolyzed collagen powder, BulkSupplements, Henderson, NV, USA) weight/volume (W/V) percentages (0, 5, 10, 15, 20, 25, 30, 35, and 40%). The tubes were arranged and supported within a cylindrical container in a similar manner to the MnCl2 phantom. The container was filled with approximately 1L of 30 mM NaCl in deionized water for coil loading.
Two scan sessions, separated by six weeks, were performed. During both sessions, PETRA series were acquired using the following parameters: total scan time = 5:21; matrix = 160; FOV = 160 mm isotropic; 80,000 radial views; voxel size = 1.00 mm isotropic; flip angle = 6°; bandwidth = 365 Hz/pixel; TR = 4.00 msec; TE = 0.07 msec. During the first scan session, the phantom was imaged with the phantom in two physical orientations. The phantom was rotated about its long axis (which was aligned with the scanner’s z-axis for both scans) by 180° relative to its initial orientation for the second scan. This was done to test for spatial positioning effects on signal homogeneity. During the first scan session, scans with TE values 0.09, 0.12, 0.16, 0.20, 0.24, 0.30, 0.40, and 0.50 msec were also acquired to allow ultrashort-T2* mapping. To allow confirmation of which tubes exhibited relatively short versus long T2* times, and thus would be reasonable or unreasonable to apply ultrashort-T2* mapping to, conventional T2* mapping using an axial plane multi-echo GRE sequence was also performed during the first scan session (matrix = 132×134 (264×268 interpolated); FOV = 157×159 mm; pixel size = 0.595×0.593 mm in-plane; slice thickness = 1.20 mm; flip angle = 15°; bandwidth = 600 Hz/pixel; fat suppressed; TR = 37.00 msec; TE = 1.60, 3.81, 5.82, 7.83, 9.84, 11.85, 23.00 msec).
In vivo knee imaging.
PETRA imaging, using the parameters listed for the low resolution series in S1 Table and with seven TE values (0.07, 0.09, 0.12, 0.16, 0.22, 0.30, and 0.58 msec; total cumulative scan time for all TE values = 44:03), were performed in three male participants (33 ± 5 years old) with no history of knee injury to test the feasibility of performing ultrashort-T2* mapping in vivo. Ultrashort- T2* mapping repeatability data (two scans separated by 10 days) were acquired for one of these participants.
Image analysis procedures
General.
For all images, 3D Slicer [26] was used to perform segmentation and MATLAB (R2024a, The Mathworks, Natick, MA) was used to measure/calculate results within the desired regions of interest (ROIs).
Image segmentation.
ROIs in the knee and image background were manually drawn in 3D Slicer by a researcher with 8 years of knee MRI segmentation experience. For all tissues except cartilage, skin, and cortical bone, the ROIs included the entire tissue of interest in at least one slice where the tissue was clearly visualized. The cartilage segmentation excluded the most medial and lateral slices, where partial volume averaging or blurring occurred. The skin and cortical bone segmentations were limited to central regions within a small set of slices, to avoid thinner tissue regions relative to the scan resolution and the resulting partial volume averaging and the blurring seen in the most medial and lateral slices. For all tissues, regions near the proximal and distal borders of the image were excluded, to avoid signal inhomogeneity near the edges of the coil’s sensitive volume.
For the phantoms, segmentation was performed using the TE = 0.07 msec images. Circular ROIs of 7 mm diameter, centered within each 15 mm diameter centrifuge tube and avoiding regions with air bubbles, were drawn in on six slices within the straight section of the tube.
Analysis of Signal Behavior.
Two scans were acquired for one participant to allow for SNR measurement. The SNR was measured in the cortical bone, patellar tendon, posterior cruciate ligament, anterior cruciate ligament, cartilage, and skin in the fat-suppressed low resolution (1.52 mm isotropic) and high resolution (0.34 mm isotropic), TE = 0.07 msec PETRA series. The mean signal in the ROIs described above was calculated. PETRA image noise was measured by calculating the difference between two subsequently acquired identical series, and then calculating the standard deviation of the difference image divided by √2 to correct for the difference operation [27]. Contrast-to-noise ratio (CNR) was also calculated, with the mean signal within a ROI in the adipose tissue posterior to the femur in the most central slice through the posterior cruciate ligament being used as the reference for contrast.
A custom MATLAB script was used to analyze the signal-concentration relationship at TE = 0.07 msec (collagen phantom) and to calculate the ultrashort-T2* maps using the TE = 0.07–0.50 msec data (MnCl2 phantom and knee images).
Ultrashort T2* Fitting.
For the MnCl2 phantom studies, the ultrashort-T2* maps were calculated from 1.50 mm isotropic, fat-suppressed PETRA at nine different TE values (0.07–0.50 msec) using voxel-wise monoexponential fitting in MATLAB. The built-in lsqcurvefit function was used, which seeks the best fit based on searching for a minimum in the sum of squares of the residuals with the trust-region-reflective algorithm. Two additional fitting methods, monoexponential fitting with a noise term and log-linear least squares fitting, were also tested for comparison. Monoexponential fitting with a noise term was performed following the same method as the monoexponential fitting. Log-linear least squares fitting was performed by first log transforming the voxel signal intensity at each TE to allow voxel-wise linear least squares non-iterative fitting of the natural log of the signal-intensity versus the array of TEs. The negative of the inverse of the slope of the resulting linear function for each voxel is the estimated ultrashort-T2*.
The coefficient of determination, R2, for the ultrashort-T2* fitting was calculated for each voxel. The ROI median and interquartile range of the ultrashort-T2* for the MnCl2 phantom and signal mean ± standard deviation for the collagen phantom were calculated. The relationships between ultrashort-T2* and MnCl2 concentration and between signal intensity at TE = 0.07 msec and collagen concentration were evaluated.
For in vivo studies, ultrashort-T2* values were calculated using the same process described for the MnCl2 phantom, including monoexponential fitting without and with a noise term and log-linear least squares fitting, except that the 1.52 mm isotropic, fat-suppressed PETRA at seven TE values (0.07–0.58 msec) were used (due to the need for a shorter total scan time for in vivo imaging). In addition, the images were carefully visually inspected using semi-transparent image overlay visualization to confirm that there was no participant movement between acquisition of the image volumes with different TE values. In addition to the ultrashort-T2* fitting performed with all seven acquired image volumes, ultrashort-T2* fitting was also performed with a subset of three TE values (0.07, 0.12, and 0.30 msec). The goal of using only 3 TE values was to reduce the total scan time (16:43), thereby minimizing participant motion between series and increasing the practical feasibility. In addition, in-plane frequency-domain interpolation was performed in MATLAB prior to segmentation and ultrashort-T2* fitting to allow more precise delineation of the knee structures (resulting in a sagittal-plane series interpolated resolution of 0.76 x 0.76 x 1.52 mm). Median ultrashort-T2* values within each ROI were calculated once using all ROI voxels, and again after voxels for which R2 was < 0.5 were excluded. The ultrashort-T2* values and R2 values were measured in the cortical bone, patellar tendon, meniscus, posterior cruciate ligament, anterior cruciate ligament, cartilage, and skin. X2 was also calculated to allow more robust comparison between goodness-of-fit between the fitting using seven TE values and three TE values.
Results
Multi-echo image optimization for ultrashort-T2* mapping
In vivo knee imaging.
Fig 1 shows an example high-resolution PETRA knee image, with comparison to a dual-echo steady state (DESS) conventional TE sequence image for anatomical reference with more standard tissue contrast. The PETRA images were obtained with high SNR, even for ultrashort-T2* tissues. As listed in Table 1, the SNR in the cortical bone, patellar tendon, meniscus, posterior cruciate ligament, anterior cruciate ligament, cartilage, and skin in the TE = 0.07 msec, 1.52 mm isotropic resolution, TR = 4.21 msec series ranged from 68 to 337 (total scan time, 5:20). In the 0.34 isotropic resolution, TR = 7.07 msec series (Fig 1), the same SNR measurements ranged from 23 to 75 (total scan time, 6:55). Note that the PETRA series at the two resolutions were acquired with different TR, so the SNR does not follow the usual relationship with voxel size between the two series, and CNR was also impacted. The 1.52 mm resolution series had higher SNR efficiency (SNR per unit time) and CNR. For comparison to a more standard-contrast image series, the SNRs in these regions for the 0.30 mm isotropic resolution conventional-TE DESS series ranged from 16 to 30.
Fig 1. Example high-resolution PETRA knee image, with comparison to a conventional-TE DESS sequence image.
Sagittal images of one participant’s knee from the fat-suppressed PETRA series at 0.34 mm isotropic resolution (A) and from the fat-suppressed dual-echo steady state (DESS) at 0.30 mm isotropic resolution (B). These images illustrate the difference in signal, and resulting differences in contrast between conventional and zero echo-time sequences, in the patellar tendon, cruciate ligaments, and other ultrashort-T2* tissues in these images. Zero echo-time series can enhance visualization of certain structures, supplementing the visualization provided by series with more conventional tissue contrast.
Table 1. SNR, SNR efficiency, and CNR measurements for PETRA at TE = 0.07 msec for knee tissue ROIs at two resolutions.
| Results TR = 7.07 msec and 0.34 mm isotropic resolution | Results for TR = 4.21 msec and 1.52 mm isotropic resolution | |||||
|---|---|---|---|---|---|---|
| Knee tissue | SNR | SNR efficiency (SNR per minutes of scan time) | CNR | SNR | SNR efficiency (SNR per minutes of scan time) |
CNR |
| Cortical bone | 23 | 3.3 | 2 | 68 | 12.8 | 6 |
| Patellar tendon | 58 | 8.4 | 33 | 183 | 34.3 | 108 |
| Meniscus | 36 | 5.2 | 11 | 139 | 26.1 | 64 |
| Posterior cruciate ligament | 39 | 5.6 | 14 | 163 | 30.6 | 88 |
| Anterior cruciate ligament | 34 | 4.9 | 9 | 143 | 26.8 | 69 |
| Cartilage | 52 | 7.5 | 27 | 172 | 32.3 | 98 |
| Skin | 75 | 10.8 | 50 | 337 | 63.2 | 263 |
Comparing the images acquired with and without fat suppression revealed that fat suppression reduced the appearance of off-resonance artifacts and improved the visualization of short-T2* tissues (Fig 2). Images with 0.34 mm isotropic resolution were acquired in scan times of 6:55 when acquired with fat suppression and 6:00 when acquired without fat suppression. Fig 2 shows example low-resolution images acquired at TE = 0.07 and 0.50 msec and high-resolution images acquired at TE = 0.07 msec, with and without fat suppression.
Fig 2. Sagittal PETRA images all acquired in a single participant knee, acquired using the parameters listed in S1 Table.
Low-resolution images acquired at TE = 0.07 msec with (A) and without (B) fat suppression, at TE = 0.58 msec with (C) and without (D) fat suppression, and high-resolution images at TE = 0.07 msec with (E) and without (F) fat suppression. Note the off-resonance artifacts appearing as double lines and thick low/high signal borders around tissue interfaces (e.g., patellar tendon-infrapatellar fat pad interface, muscle-fat interfaces, etc.).
To achieve acquisitions at longer TE values without exceeding 12 minutes per scan, the resolution needed to be set to 1.52 mm isotropic. At this resolution with 70,000 radial views and TR = 4.21 msec the scan time was under 12 minutes at TE values up to 0.58 msec (scan time = 10:55, with fat suppression).
Ultrashort-T2* mapping
MnCl2 phantom.
Monoexponential fitting without a noise term generally produced more homogeneous ultrashort-T2* within each ROI and smaller inter-scan differences than the other two fitting methods, and so are presented as the main results. The monoexponential-fit without noise term results for ultrashort-T2* measurements, ultrashort-T2* curve fit R2 values, and absolute and percent change between scans for each phantom solution ROI for two sessions are shown in Table 2. Bland-Altman analysis results are shown in S10 Fig and S11 Table. The same results for the other two fitting methods are shown in S2 Table and S3 Table. Example ultrashort-T2* curve fits are shown in Fig 3A. There was a significant linear correlation between 1/T2*ultrashort values and MnCl2 concentration (r = 0.99 and p ≤ 0.01 for both scans), as shown in Fig 3B. Mean R2 dropped sharply for MnCl2 concentrations below 3.6 mM, and at the lowest concentration (0.02 mM) a large number of very high or even negative ultrashort-T2* values was calculated due to the lack of signal decay measured over the range of acquired TEs, leading to poor fit quality and the very large interquartile range for that measurement. The scan 1 and scan 2 slopes and intercepts were calculated using just the results for the solution concentrations with average R2 for ultrashort-T2* fit ≥ 0.85 (which corresponded to 5.4–27.07 mM, with ultrashort-T2* values of ≤ 1.14 msec). The scan 1 and scan 2 slopes were 108.68 and 109.41 seconds-1/mM (<1% difference between scans), and the intercepts were 413.97 and 359.81 seconds-1 (13% difference between scans).
Table 2. Calculated ultrashort-T2* values, between-scan absolute and percent change, and ultrashort-T2* fit R2 from two scans for the MnCl2 phantom. The median and interquartile range values are calculated for the sample of all voxels within each ROI.
| Phantom MnCl2 solution concentration [mM] | Scan 1 T2* median (interquartile range) [msec] | Scan 2 T2* median (interquartile range) [msec] | Between- scan T2* absolute change [msec] | Between- scan T2* percent change | Scan 1 mean R2 | Scan 2 mean R2 |
|---|---|---|---|---|---|---|
| 27.07 | 0.31 (0.03) | 0.31 (0.03) | 0.00 | 1% | 0.96 | 0.96 |
| 13.53 | 0.47 (0.04) | 0.49 (0.05) | 0.02 | 5% | 0.98 | 0.95 |
| 9.01 | 0.78 (0.10) | 0.78 (0.06) | 0.00 | 0% | 0.86 | 0.88 |
| 5.4 | 1.07 (0.20) | 1.14 (0.20) | 0.07 | 7% | 0.97 | 0.95 |
| 3.6 | 3.55 (3.90) | 2.03 (0.61) | 1.52 | 43% | 0.50 | 0.84 |
| 2.7 | 2.91 (2.87) | 2.45 (0.68) | 0.47 | 16% | 0.54 | 0.73 |
| 1.79 | 5.12 (2.17) | 6.61 (1.80) | 1.49 | 29% | 0.34 | 0.24 |
| 1.12 | 4.73 (0.73) | 6.08 (2.19) | 1.34 | 28% | 0.37 | 0.30 |
| 1.07 | 3.96 (2.23) | 4.32 (1.87) | 0.36 | 9% | 0.56 | 0.64 |
| 0.02 | 12.34 (9.61) | 9.89 (10.40) | 2.44 | 20% | 0.08 | 0.11 |
Fig 3. MnCl2 phantom plots.
Plots showing examples of median measured (round markers) and fitted signal (dashed line) based on estimated ultrashort-T2* signal versus TE for three MnCl2 phantom solutions from two (top and bottom plots for each region) scans (a), the relationship between mean 1/T2* and MnCl2 concentration, averaged over two scans (b), the relationship between mean signal and collagen concentration at TE = 0.07 msec, averaged over three scans (c). Error bars represent the interquartile range within each subregion for signal versus TE plots and the standard deviation between scans for the 1/T2* and signal versus concentration plots.
Collagen phantom.
For the collagen phantom, the mean conventional-T2* values for all W/V percentages greatly exceeded the ultrashort-T2* upper limit, being 23 msec for 40% collagen to 330 msec for 5% collagen; therefore ultrashort-T2* mapping was not evaluated. There was a significant linear correlation between collagen concentration and signal intensity in the TE = 0.07 msec images (r = 0.98 on average, p <0.0001). The individual signal measurements, linear fitting parameters, and correlation statistics are shown in Table 3. Fig 3C shows this relationship for the average of the three TE = 0.07 msec PETRA series. The between-scan and between-orientation signal and linear fitting parameter CV were all ≥ 0.05 except for 0% collagen signal (Table 3). Bland-Altman analysis results are shown in S9 Fig and S11 Table.
Table 3. Signal measurements for each collagen concentration from three scans of the collagen phantom and the resulting signal versus concentration linear fitting parameters and correlation statistics.
| Scan 1, orientation 1 | Scan 1, orientation 2 | Scan 2, orientation 1 | CV | |
|---|---|---|---|---|
| Collagen W/V percentage [%] | Mean signal (standard deviation) measurements [a.u.] | |||
| 0 | 301 (32) | 268 (16) | 355 (20) | 14% |
| 5 | 307 (16) | 309 (15) | 310 (10) | 1% |
| 10 | 389 (13) | 403 (23) | 362 (10) | 5% |
| 15 | 463 (11) | 455 (25) | 417 (10) | 5% |
| 20 | 520 (21) | 481 (25) | 496 (13) | 4% |
| 25 | 604 (19) | 564 (19) | 607 (24) | 4% |
| 30 | 637 (16) | 627 (19) | 651 (23) | 2% |
| 35 | 680 (21) | 646 (14) | 625 (9) | 4% |
| 40 | 792 (30) | 719 (43) | 752 (13) | 5% |
| Linear fit and correlation parameter values | ||||
| Slope [a.u./%] | 12.40 | 11.26 | 11.01 | 6% |
| Intercept [a.u.] | 273.34 | 271.74 | 288.32 | 3% |
| r | 0.99 | 0.99 | 0.96 | |
| P | <0.0001 | <0.0001 | <0.0001 | |
In vivo knee imaging.
Acquiring images with different TEs but other parameters held constant allowed creation of ultrashort-T2* maps. Fig 4 shows an ultrashort-T2* map and corresponding R2 map for one knee for two different scan sessions. Ultrashort-T2* measurements and fitting information for the cortical bone, patellar tendon, meniscus, posterior cruciate ligament, anterior cruciate ligament, cartilage, and skin are shown in Table 4. The percent of voxels excluded due to R2 being below 0.5 are listed for each tissue in Table 5. Scan-rescan results for one participant and interscan coefficients of variation are shown in Table 6 and Bland-Altman analysis results are shown in S10 Fig and S11 Table. Results obtained using all seven acquired TE values and using a subset of three TE values are presented for both tables. Example T2* curves are shown in Fig 5. The same sets of results obtained using monoexponential fitting with the addition of a noise term and log-linear least squares fitting are presented in S4 Table, S5 Table, S6 Table, and S7 Table.
Fig 4. Ultrashort-T2* map and corresponding R2 map for one knee for two different scan sessions.
Images and maps for one participant’s knee scanned in two sessions, grouped by scan session and labeled as Scan 1 (top two rows) and Scan 2 (bottom two rows). Images for each scan include an example PETRA image at TE = 0.07 msec, ultrashort-T2* map in msec calculated using 7 TE values and R2 fit goodness map for the 7-TE fit in the masked ROIs (upper row of each set), and the ultrashort-T2* map in msec calculated using 3 TE values and R2 fit goodness map for the 3-TE fit in the masked ROIs (bottom row of each set).
Table 4. Knee tissue calculated median and range ultrashort-T2* values, ultrashort-T2* fit R2, and ultrashort-T2* fit X2 for single scans for three participants. The results calculated using all seven TE values and the subset of three TE values are presented. The median value represents the median of the three median values (one per participant, for each ROI) calculated for each ROI. The range represents the minimum and maximum participant ROI median values.
| Knee tissue | Median ultrashort-T2* (Range) [msec] | Median R2 (Range) | Median X2 (Range) | Median ultrashort-T2* (Range) [msec] | Median R2 (Range) | Median X2 (Range) |
|---|---|---|---|---|---|---|
| Results when fitting to seven TE values | ||||||
| All voxels | Voxels with R2 ≥0.5 | |||||
| Cortical bone | 0.7 (0.35–1.34) | 0.93 (0.71–0.93) | 6.36 (3.71–9.38) | 0.7 (0.4–1.29) | 0.93 (0.82–0.93) | 4.26 (3.71–9.38) |
| Patellar tendon | 1.17 (1.06–1.57) | 0.90 (0.71–0.93) | 5.75 (3.48–18.45) | 1.07 (1.01–1.29) | 0.92 (0.81–0.94) | 5.15 (3.16–15.93) |
| Meniscus | 1.95 (1.75–2.29) | 0.59 (0.58–0.62) | 8.39 (8.3–8.95) | 1.77 (1.57–2.11) | 0.77 (0.73–0.78) | 6.12 (5.92–6.51) |
| Posterior cruciate ligament | 1.38 (1.25–2.27) | 0.79 (0.76–0.81) | 8.79 (3.54–10.37) | 1.31 (1.27–2.25) | 0.82 (0.81–0.84) | 8.21 (2.86–8.41) |
| Anterior cruciate ligament | 1.5 (0.97–1.94) | 0.71 (0.35–0.84) | 9.89 (7.85–13.33) | 1.3 (0.76–1.66) | 0.81 (0.77–0.9) | 5.77 (4.91–10.81) |
| Cartilage | 1.05 (0.9–1.22) | 0.87 (0.75–0.9) | 10.54 (7.39–19.2) | 0.97 (0.8–1.09) | 0.91 (0.82–0.93) | 7.12 (6.81–14.76) |
| Skin | 0.82 (0.76–0.98) | 0.98 (0.97–0.98) | 3.73 (2.27–4.55) | 0.82 (0.76–0.96) | 0.98 (0.97–0.98) | 3.63 (2.27–4.29) |
| Results when fitting to three TE values | ||||||
| All voxels | Voxels with R2 ≥0.5 | |||||
| Cortical bone | 0.75 (0.35–0.75) | 0.87 (0.83–0.89) | 1.2 (1.05–3.61) | 0.75 (0.35–0.75) | 0.87 (0.87–0.89) | 1.05 (0.86–3.61) |
| Patellar tendon | 1.01 (0.94–1.58) | 0.89 (0.52–0.93) | 1.03 (0.63–5.24) | 0.80 (0.80–0.96) | 0.97 (0.87–0.98) | 0.48 (0.35–2.98) |
| Meniscus | 0.91 (0.89–1.03) | 0.82 (0.76–0.88) | 1.98 (1.33–2.18) | 0.89 (0.80–1.00) | 0.88 (0.8–0.89) | 1.76 (1.22–1.80) |
| Posterior cruciate ligament | 0.76 (0.73–1.13) | 0.91 (0.49–0.93) | 1.59 (1.36–2.16) | 0.75 (0.72–1.00) | 0.92 (0.79–0.96) | 1.4 (0.71–1.54) |
| Anterior cruciate ligament | 0.79 (0.60–0.84) | 0.89 (0.56–0.98) | 1.62 (0.66–3.98) | 0.75 (0.53–0.82) | 0.92 (0.91–0.99) | 0.89 (0.29–1.23) |
| Cartilage | 0.71 (0.67–0.72) | 0.95 (0.91–0.95) | 0.97 (0.94–2.34) | 0.68 (0.66–0.70) | 0.96 (0.93–0.96) | 0.85 (0.81–1.95) |
| Skin | 0.74 (0.73–0.82) | 0.98 (0.98–0.99) | 0.42 (0.25–0.43) | 0.74 (0.72–0.82) | 0.98 (0.98–0.99) | 0.42 (0.25–0.43) |
Table 5. Percent voxels excluded for R2 being below the 0.5 threshold for single scans for three participants. The results for fitting using all seven TE values and the subset of three TE values are presented.
| Knee tissue | Median (range) of percent voxels excluded for R2 < 0.5 [%] | |
|---|---|---|
| Results when fitting to seven TE values | Results when fitting to three TE values | |
| Cortical bone | 0 (0–30) | 0 (0–17) |
| Patellar tendon | 13 (7–25) | 31 (26–49) |
| Meniscus | 39 (37–42) | 14 (9–27) |
| Posterior cruciate ligament | 18 (6–24) | 17 (3–51) |
| Anterior cruciate ligament | 30 (29–60) | 21 (16–47) |
| Cartilage | 24 (10–25) | 7 (6–11) |
| Skin | 2 (0–3) | 0 (0–1) |
Table 6. Scan 1 and scan 2 single-participant ROI median and interquartile range (IQR) ultrashort-T2* values and between-scan absolute and percent change calculated from the two scan sessions for one participant. The results calculated using all seven TE values and the subset of three TE values are presented for all voxels and for only voxels with acceptable ultrashort-T2* fit (R2 ≥ 0.5). The median and interquartile range values are calculated for the sample of all voxels within each ROI.
| Scan 1 median ultrashort-T2* (IQR) within each ROI [msec] | Scan 2 median ultrashort-T2* (IQR) within each ROI [msec] | Scan-rescan ultrashort-T2* absolute change [ms] | Scan-rescan ultrashort-T2* percent change | |||||
|---|---|---|---|---|---|---|---|---|
| Knee tissue | All voxels | Voxels with R2 ≥0.5 | All voxels | Voxels with R2 ≥0.5 | All voxels | Voxels with R2 ≥0.5 | All voxels | Voxels with R2 ≥0.5 |
| Results when fitting to seven TE values | ||||||||
| Cortical bone | 0.35 (0.05) | 0.35 (0.05) | 1.77 (1.29) | 1.69 (1.22) | 1.44 | 1.34 | 398% | 378% |
| Patellar tendon | 1.57 (1.41) | 1.29 (0.84) | 4.43 (1996.69) | 1.46 (1.10) | 2.85 | 0.17 | 182% | 13% |
| Meniscus | 1.75 (1.32) | 1.57 (1.03) | 1.95 (1.37) | 1.73 (0.85) | 0.20 | 0.16 | 12% | 10% |
| Posterior cruciate ligament | 1.25 (0.75) | 1.27 (0.77) | 1.30 (0.78) | 1.26 (0.78) | 0.04 | 0.01 | 3% | 1% |
| Anterior cruciate ligament | 1.50 (1.06) | 1.30 (0.87) | 1.56 (0.92) | 1.35 (0.72) | 0.06 | 0.05 | 4% | 4% |
| Cartilage | 1.22 (1.16) | 1.09 (1.08) | 1.40 (1.04) | 1.18 (0.94) | 0.18 | 0.09 | 15% | 9% |
| Skin | 0.76 (0.33) | 0.76 (0.33) | 0.92 (0.35) | 0.92 (0.35) | 0.16 | 0.16 | 21% | 20% |
| Results when fitting to three TE values | ||||||||
| Cortical bone | 0.35 (0.09) | 0.35 (0.09) | 0.97 (0.41) | 0.74 (0.38) | 0.62 | 0.38 | 174% | 108% |
| Patellar tendon | 1.58 (2.57) | 0.96 (0.53) | 3.03 (8.62) | 1.08 (0.45) | 1.46 | 0.12 | 92% | 13% |
| Meniscus | 0.89 (0.48) | 0.80 (0.34) | 0.94 (0.38) | 0.86 (0.26) | 0.05 | 0.06 | 5% | 8% |
| Posterior cruciate ligament | 0.73 (0.32) | 0.72 (0.31) | 0.75 (0.27) | 0.74 (0.26) | 0.02 | 0.02 | 3% | 3% |
| Anterior cruciate ligament | 0.79 (0.29) | 0.75 (0.27) | 0.82 (0.25) | 0.79 (0.24) | 0.03 | 0.04 | 4% | 5% |
| Cartilage | 0.72 (0.46) | 0.68 (0.44) | 0.77 (0.34) | 0.74 (0.32) | 0.05 | 0.06 | 7% | 8% |
| Skin | 0.73 (0.22) | 0.72 (0.22) | 0.74 (0.20) | 0.74 (0.20) | 0.02 | 0.02 | 3% | 3% |
Fig 5. Example T2* curves.
Plots of the median measured (round markers) and fitted (dashed line) signal based on estimated ultrashort-T2* signal versus TE for the measured knee tissue regions from two scans of a single participant’s knee (10 days between scans). The points and for signal decay curves fitted to all seven TE values are shown in black and the points and signal decay curves fitted to a subset of three TE values are shown in red. Error bars represent the interquartile range within each subregion.
Discussion
We modified a commercially available ZTE sequence, PETRA [1], to extend the maximum technically allowable echo time to up to 1.10 msec and demonstrate the feasibility of extended TE imaging (up to 0.58 msec) with PETRA at 7T in MnCl2 and collagen phantoms and in vivo. PETRA imaging was feasible in vivo within clinically realistic acquisition times at high resolution (0.34 mm isotropic) at short TE values and lower resolution (1.52 mm isotropic) at the high TE values. In addition to distinguishing between tissues and tissue components with comparatively short T2* values from those with long T2* values, these modifications further allow advanced applications such as ultrashort-T2* mapping from 7T ZTE data.
Image quality indicators
Higher SNR was measured in all acquired PETRA images than in a conventional TE series at similar resolution. Fat suppression was useful for reducing off-resonance artifact. As previously described [10], coil artifact was observed in a few instances. However, it was outside of the knee volume and therefore did not impact image quality. Furthermore, near-isocenter positioning of the knee allows for minimization of lateral image blurring, as reported previously [28].
Ultrashort T2* mapping
The extension of the maximum TE allows ultrashort-T2* mapping by multiple acquisitions for substances with sub-millisecond T2* values, demonstrated here using the MnCl2 phantom and in vivo knees. Ultrashort-T2* fitting R2 was ≥0.86 for MnCl2 phantom solutions with ultrashort-T2* ≤ 1.14 msec and for most of the knee tissues analyzed. For MnCl2 samples, there was a strong significant correlation between 1/ultrashort-T2* values and concentration. Although this finding is not surprising, given the predictions of fast exchange theory, the linearity of the relaxivity plot provides an important quality verification of the T2* mapping capabilities of the sequence.
We simulated collagen-rich tissues using a collagen phantom. Although the collagen concentrations were similar to the range found in typical short-T2* tissues, the lack of collagen organization limits the phantom’s similarity to actual tissue and resulted in measured conventional T2* values that were much longer than in the tissues of interest [29]; therefore, ultrashort-T2* mapping was not evaluated. However, we observed a strong correlation between collagen concentration and signal at TE = 0.07 msec, likely due to T1 differences between the collagen solutions. Previously reported collagen solution T1 values suggest that the expected T1 range for our 5%-40% collagen solutions would be around 1.6–2.6 sec at 3T [30] and somewhat longer at 7T [31–33]. The T1 values in short-T2* musculoskeletal and skin tissues with would be expected to be around 330–350 (bound water)/390–1000 (pore water) msec in cortical bone at 4.7 T [34,35], 620–700 msec in tendon at 3T [3,36,37], 819–870 msec in ligament at 3T [37], 870 msec in meniscus at 3T [37], and 1.13 sec in cartilage at 3T [37], and 820–1060 msec in dermis and epidermis of the skin at 1.5T [38], with relatively longer T1 values expected for each tissue at 7T. This range of T1 values would also be expected to create collagen-dependent signal variations in similarly configured PETRA images. While variations in other structural and compositional tissue properties would prevent across-tissue interpretation of the signal variations as reflecting collagen content, it is possible that within-tissue signal variations could be interpretable as variations in collagen content and could be associated with tissue pathology. Future work with relevant tissues is needed to determine if this application is possible, as the collagen solutions used in this work do not reflect the structure or complexity of actual tissue.
The median ultrashort-T2* values were generally similar between the three participants. When comparing ultrashort-T2* values calculated using all seven versus the subset of three TE values we found that the median ultrashort-T2* values were lower when only three TE values were used than when seven were used (maximum TE = 0.30 msec versus 0.58 msec). This may reflect higher contributions of shorter T2* components when shorter TE values are used, with longer components contributing more to the overall measurement as a greater proportion of longer TE values are included. Future work in tissue samples or patients is needed to determine if the differences in ultrashort-T2* values calculated using 7 versus 3 TE values are meaningful.
The use of three TE values for ultrashort-T2* fitting resulted in better fit and reproducibility than when seven TE values were used. Of the tested tissues, cortical bone showed the poorest reproducibility, which may be in part due to the thin ROI relative to the low scan resolution and potential for chemical shift artifact around the bone, and the relatively distal location of the bone ROI relative to the image volume center. Midshaft bone measurements, where the cortical bone is thicker and the bone ROI can be centered in the imaging volume, may be more reproducible. The shorter scan time, better decay curve fit, and higher between-scan reproducibility makes the three TE approach preferable for mapping of the ultrashort-T2* components. Regardless of the number of TE values used, the maximum TE of ≤0.58 msec used in this work prevents the capture and analysis of long- T2* components. The maximum TE value is limited in this sequence, as image quality was markedly reduced when TE values above 0.58 msec were tested. The addition of a conventional-TE T2* sequence would be necessary to capture these long-T2* components, when desired.
For patellar tendon, the median value of 1.07 msec when seven TE were used and 0.80 msec when 3 TE values were used were lower than the value of 1.57 ± 0.11 msec reported for the short T2* component in one patellar tendon sample at 7T [39]. Our values calculated using three TE were of similar magnitude to those reported for short-component T2* in the Achilles tendon of healthy volunteers at 7T, which ranged from 0.20–0.48 msec [20]. Tendon long axis angle relative to the main magnetic field was not reported in that work, limiting comparison to the patellar tendon due to the different anatomical orientation of the tendons. We did see some non-exponential behavior in the signal decay of the tendon ROI in some cases (as illustrated in Fig 5), which may be due to partial volume averaging with adjacent adipose tissue or blurring where the tendon is too far from isocenter.
The monoexponential T2* of cartilage in uninjured and ACL-reconstruction knees, measured with a mix of ultrashort and conventional TE (TE range: 0.032–16 msec) at 3T, has been reported to be around 14–20 msec [40], and the short/long-T2* components of human cadaveric patellar cartilage by biexponential fitting (TE range: 0.08–40 msec) at 3T and 1st/2nd/3rd components of bovine nasal cartilage by triexponential fitting (TE range: 0.6–614.4 msec) at 9.4T have been reported to be about 0.50/35 msec and 2.3/25.2/96.3 msec respectively [41]. Our cartilage ultrashort-T2* values of 0.97 msec (seven TE) and 0.68 msec (three TE) were measured with a sub-millisecond maximum TE, causing a strong ultrashort component weighting, and as a result are more similar to the reported short-component biexponential and triexponential T2*values than the longer monoexponential T2* values. Ultrashort-T2* components may be more sensitive to some pathology than the long components [22].
Conventional- and ultrashort-T2* mapping has been less thoroughly explored in skin than it has been in musculoskeletal tissues, but it is of interest since skin is a high-collagen-content tissue that may be impacted alongside musculoskeletal tissues when collagen structure or function is abnormal. Skin conventional-T2*, calculated from images acquired at TE values of 2.8–60 msec, has been reported to be 9.8 msec at 1.5T [42,43]. Our measured value of 0.82 msec (seven TE)/0.74 msec (three TE) likely reflects only the shorter components since we only acquired images at sub-millisecond TE values.
Our cortical bone ultrashort-T2* median value of 0.70 msec (seven TE)/0.75 msec (3 TE) was similar to the previously reported bone specimen ultrashort-T2* values of 0.4–0.7 msec measured at 7T (TE range: 0.064–2.048 msec) [6] and near the bound-water ultrashort-T2* values measured previously at 4.7T [35].
The meniscus ultrashort-T2* mapping results for both sets of TE (three and seven TE) are both lower than ultrashort-T2* at 3T (7.3 msec) [19] and short-T2* values at 7T (7.31–9.19 msec) [44] reported previously for healthy menisci. However, our values were within the range of 0.4–4 msec reported previously for short-component T2* of meniscus in vitro at 9.4T [45]. Similarly, our ACL and PCL ultrashort-T2* values were lower than the previously reported values [46]. This is likely in part due to the expected reduction in T2* at 7T compared to 3T [47]. As noted above, the extent of the field strength-dependence has been shown to vary by region within tendon [20], and this may also be true in the two cruciate ligaments.
The differences between the results here for all tissues and those reported in the literature are also likely due to the difference in TE range used, where most other studies have been performed with either a single or a few ultrashort-TE and a larger number of conventional TE values (>1 msec), in contrast to our use of only ultrashort-TE values. In addition, our use of a coefficient of determination threshold for excluding poor-fit voxels is less commonly used than approaches in which voxels are excluded for having T2* or T2 values outside of an expected measurement range. While the use of bicomponent exponential fitting and models with incorporation of Gaussian decay and oscillating off-resonance components have been found to improve the fit of tendon [48,49] and meniscus [45] T2* signal decay, these more complex models were not explored in this current work, as the upper range of TE values and number of TE points acquired were not sufficient to enable this more complex modeling. However, these components do likely contribute to the signal behavior and signal change between TE values. These factors, along with partial volume averaging and potential issues due to movement between series, artifact, or noise, may contribute to the poor monoexponential fit in some voxels.
Limitations
This work was limited to imaging of MnCl2 and collagen phantoms and a small number of healthy volunteers. Additional image optimization and parameter adjustment may be needed for optimized imaging of patients, which could be performed based on quantitative measures such as SNR and CNR and reader-based image quality and structure/pathology visualization assessments as described in previous work performed by others [50–52]. The sensitivity and specificity of PETRA-based short-TE signal measurements and ultrashort-T2* mapping for quantifying tissue health remains to be assessed. The small sample size and the strong impact on ultrashort-T2* of the TE values used here limits the direct comparison between the reported values and our measurements.
PETRA’s combination of pointwise and radial sampling varies depending on TE, and this is expected to have some impact on the measured signal. Future work, using a phantom with known ultrashort-T2* values or with tissue samples or participants being imaged with both PETRA and a reference ultrashort-T2* measurement sequence, such as FID-based ultrashort-T2* calculations [34], may be needed to assess the impact on the accuracy of PETRA ultrashort-T2* measurements.
Ultrashort-T2* mapping using PETRA does require relatively long cumulative scan time and low resolution to avoid unrealistic scan times. We found that the maximum TE needed to be restricted to 0.58 msec to avoid extremely high scan times for single scans. Bicomponent or multicomponent T2* mapping for musculoskeletal tissues would require a wider range of TE values, since long-T2* tissue suppression by subtraction imaging and quantitative ultrashort-T2* mapping require at least one TE long enough, generally >1 msec, for signal of the short-T2* tissue of interest to significantly decrease compared to the long-T2* tissue values [3,20,41]. A potential solution to the cumulative scan time with long TE values may be to fix the central portions of k-space filled by the single pointwise approach to an upper limit as function of TE. In this manner, scan time would be shortened while maintaining image resolution and SNR. Other methods, such as iterative reconstruction or deep learning algorithms could also be used for improved efficiency [53–55].
Conclusion
In this work we present the results of phantom and in vivo knee imaging using a modified PETRA sequence with an increased TE range. The results obtained are encouraging, demonstrating that following suitable image optimization, this updated PETRA sequence can provide high quality in vivo images. Furthermore, ultrashort-T2* mapping was feasible using the modified extended TE range in the phantoms and in vivo in a small sample of healthy knees. Work is in progress to reduce scan time at long TEs by setting the Cartesian central region of the k-space to an upper limit while preserving image information.
Supporting information
(DOCX)
The median and interquartile range values are calculated for the sample of all voxels within each ROI.
(DOCX)
The median and interquartile range values are calculated for the sample of all voxels within each ROI.
(DOCX)
The results calculated using all seven TE values and the subset of three TE values are presented. The median value represents the median of the three median values (one per participant, for each ROI) calculated for each ROI. The range represents the minimum and maximum participant ROI median values.
(DOCX)
The results calculated using all seven TE values and the subset of three TE values are presented. The median value represents the median of the three median values (one per participant, for each ROI) calculated for each ROI. The range represents the minimum and maximum participant ROI median values.
(DOCX)
The results calculated using all seven TE values and the subset of three TE values are presented for all voxels and for only voxels with acceptable ultrashort-T2* fit (R2 ≥ 0.5). The median value represents the median of the three median values (one per participant, for each ROI) calculated for each ROI. The range represents the minimum and maximum participant ROI median values.
(DOCX)
The results calculated using all seven TE values and the subset of three TE values are presented for all voxels and for only voxels with acceptable ultrashort-T2* fit (R2 ≥ 0.5). The median and interquartile range values are calculated for the sample of all voxels within each ROI.
(DOCX)
(TIF)
(TIF)
Plots are shown for the values based on 7-TE and 3-TE and include all ultrashort-T2* values (all R2).
(TIF)
Ultrashort-T2* results are for all ultrashort-T2* values (all R2).
(DOCX)
Acknowledgments
We thank Dr. Dean Hoffmeister, MD, for providing input on image interpretation and clinical applications, and the Carle Illinois Advanced Imaging Center 7T MRI technologists who acquired the MR imaging data.
Data Availability
The data cannot be shared publicly in an unrestricted manner because the Institutional Review Board protocol under which participants were consented only permits sharing of the data with qualified investigators, who have current human subjects research ethics training, at academic institutions or private research organizations. The PETRA image data are available to any qualified investigator at an academic institution or private research organization who has current human subjects research ethics training certification and can describe a scientific use for the data. Interested parties can use the following link (https://redcap.link/AR073831_Data_Request) to initiate a Data Use Agreement.
Funding Statement
BMD received financial support for this work from NIH/NIAMS 1 R01 AR073831 and from the Stephens Family Clinical Research Institute Innovations in Medical Research Award. The sponsors/funders did not play any role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.
References
- 1.Grodzki DM, Jakob PM, Heismann B. Ultrashort echo time imaging using pointwise encoding time reduction with radial acquisition (PETRA). Magn Reson Med. 2012;67(2):510–8. doi: 10.1002/mrm.23017 [DOI] [PubMed] [Google Scholar]
- 2.Breighner RE, Endo Y, Konin GP, Gulotta LV, Koff MF, Potter HG. Technical developments: zero echo time imaging of the shoulder: enhanced osseous detail by using mr imaging. Radiology. 2018;286(3):960–6. doi: 10.1148/radiol.2017170906 [DOI] [PubMed] [Google Scholar]
- 3.Chang EY, Du J, Chung CB. UTE imaging in the musculoskeletal system. J Magn Reson Imaging. 2015;41(4):870–83. doi: 10.1002/jmri.24713 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Froidevaux R, Weiger M, Brunner DO, Dietrich BE, Wilm BJ, Pruessmann KP. Filling the dead-time gap in zero echo time MRI: Principles compared. Magn Reson Med. 2018;79(4):2036–45. doi: 10.1002/mrm.26875 [DOI] [PubMed] [Google Scholar]
- 5.Han M, Larson PEZ, Liu J, Krug R. Depiction of achilles tendon microstructure in vivo using high-resolution 3-dimensional ultrashort echo-time magnetic resonance imaging at 7 T. Invest Radiol. 2014;49(5):339–45. doi: 10.1097/RLI.0000000000000025 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Krug R, Larson PEZ, Wang C, Burghardt AJ, Kelley DAC, Link TM, et al. Ultrashort echo time MRI of cortical bone at 7 tesla field strength: a feasibility study. J Magn Reson Imaging. 2011;34(3):691–5. doi: 10.1002/jmri.22648 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Larson PEZ, Han M, Krug R, Jakary A, Nelson SJ, Vigneron DB, et al. Ultrashort echo time and zero echo time MRI at 7T. MAGMA. 2016;29(3):359–70. doi: 10.1007/s10334-015-0509-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Li C, Magland JF, Zhao X, Seifert AC, Wehrli FW. Selective in vivo bone imaging with long-T2 suppressed PETRA MRI. Magn Reson Med. 2017;77(3):989–97. doi: 10.1002/mrm.26178 [DOI] [PubMed] [Google Scholar]
- 9.Schieban K, Weiger M, Hennel F, Boss A, Pruessmann KP. ZTE imaging with enhanced flip angle using modulated excitation. Magn Reson Med. 2015;74(3):684–93. doi: 10.1002/mrm.25464 [DOI] [PubMed] [Google Scholar]
- 10.Weiger M, Brunner DO, Dietrich BE, Müller CF, Pruessmann KP. ZTE imaging in humans. Magn Reson Med. 2013;70(2):328–32. doi: 10.1002/mrm.24816 [DOI] [PubMed] [Google Scholar]
- 11.Zibetti MVW, Johnson PM, Sharafi A, Hammernik K, Knoll F, Regatte RR. Rapid mono and biexponential 3D-T1ρ mapping of knee cartilage using variational networks. Sci Rep. 2020;10(1):19144. doi: 10.1038/s41598-020-76126-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Lee YH, Suh J-S, Grodzki D. Ultrashort echo (UTE) versus pointwise encoding time reduction with radial acquisition (PETRA) sequences at 3 Tesla for knee meniscus: a comparative study. Magn Reson Imaging. 2016;34(2):75–80. doi: 10.1016/j.mri.2015.09.003 [DOI] [PubMed] [Google Scholar]
- 13.Sharafi A, Baboli R, Chang G, Regatte RR. 3D-T1ρ prepared zero echo time-based PETRA sequence for in vivo biexponential relaxation mapping of semisolid short-T2 tissues at 3 T. J Magn Reson Imaging. 2019;50(4):1207–18. doi: 10.1002/jmri.26664 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Kim SK, Kim D, Lee SJ, Choo HJ, Oh M, Son Y, et al. Clinical value of pointwise encoding time reduction with radial acquisition (PETRA) MR sequence in assessing internal derangement of knee. Clin Imaging. 2018;51:260–5. doi: 10.1016/j.clinimag.2018.05.022 [DOI] [PubMed] [Google Scholar]
- 15.Lee HM, Weiger M, Giehr C, Froidevaux R, Brunner DO, Rösler MB, et al. Long-T2 -suppressed zero echo time imaging with weighted echo subtraction and gradient error correction. Magn Reson Med. 2020;83(2):412–26. doi: 10.1002/mrm.27925 [DOI] [PubMed] [Google Scholar]
- 16.Hager B, Walzer SM, Deligianni X, Bieri O, Berg A, Schreiner MM, et al. Orientation dependence and decay characteristics of T2 * relaxation in the human meniscus studied with 7 Tesla MR microscopy and compared to histology. Magn Reson Med. 2019;81(2):921–33. doi: 10.1002/mrm.27443 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Williams A, Qian Y, Bear D, Chu CR. Assessing degeneration of human articular cartilage with ultra-short echo time (UTE) T2* mapping. Osteoarthritis Cartilage. 2010;18(4):539–46. doi: 10.1016/j.joca.2010.02.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Imamura R, Teramoto A, Murahashi Y, Okada Y, Okimura S, Akatsuka Y, et al. Ultra-Short Echo Time-MRI T2* mapping of articular cartilage layers is associated with histological early degeneration. Cartilage. 2025;16(1):118–24. doi: 10.1177/19476035231205685 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Williams A, Qian Y, Golla S, Chu CR. UTE-T2* mapping detects sub-clinical meniscus injury after anterior cruciate ligament tear. Osteoarthritis Cartilage. 2012;20(6):486–94. doi: 10.1016/j.joca.2012.01.009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Juras V, Zbyn S, Pressl C, Valkovic L, Szomolanyi P, Frollo I, et al. Regional variations of T₂* in healthy and pathologic achilles tendon in vivo at 7 Tesla: preliminary results. Magn Reson Med. 2012;68(5):1607–13. doi: 10.1002/mrm.24136 [DOI] [PubMed] [Google Scholar]
- 21.Robson MD, Benjamin M, Gishen P, Bydder GM. Magnetic resonance imaging of the Achilles tendon using ultrashort TE (UTE) pulse sequences. Clin Radiol. 2004;59(8):727–35. doi: 10.1016/j.crad.2003.11.021 [DOI] [PubMed] [Google Scholar]
- 22.Wilms LM, Radke KL, Latz D, Thiel TA, Frenken M, Kamp B, et al. UTE-T2* versus conventional T2* mapping to assess posterior cruciate ligament ultrastructure and integrity-an in-situ study. Quant Imaging Med Surg. 2022;12(8):4190–201. doi: 10.21037/qims-22-251 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Jerban S, Hananouchi T, Ma Y, Namiranian B, Dorthe EW, Wong JH, et al. Correlation between the elastic modulus of anterior cruciate ligament (ACL) and quantitative ultrashort echo time (UTE) magnetic resonance imaging. J Orthop Res. 2022;40(10):2330–9. doi: 10.1002/jor.25266 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Thangavel K, Saritaş EÜ. Aqueous paramagnetic solutions for MRI phantoms at 3 T:A detailed study on relaxivities. Turk J Elec Eng & Comp Sci. 2017;25:2108–21. doi: 10.3906/elk-1602-123 [DOI] [Google Scholar]
- 25.Keenan KE, Ainslie M, Barker AJ, Boss MA, Cecil KM, Charles C, et al. Quantitative magnetic resonance imaging phantoms: a review and the need for a system phantom. Magn Reson Med. 2018;79(1):48–61. doi: 10.1002/mrm.26982 [DOI] [PubMed] [Google Scholar]
- 26.Fedorov A, Beichel R, Kalpathy-Cramer J, Finet J, Fillion-Robin J-C, Pujol S, et al. 3D Slicer as an image computing platform for the Quantitative Imaging Network. Magn Reson Imaging. 2012;30(9):1323–41. doi: 10.1016/j.mri.2012.05.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.National Electrical Manufacturers Association. Determination of signal-to-noise ratio (SNR) in diagnostic magnetic resonance imaging. 2008. https://www.nema.org/docs/default-source/standards-document-library/ms1-2008-r2014-watermarked.pdf?sfvrsn=2101f7b9_2 [Google Scholar]
- 28.Ilbey S, Jung M, Emir U, Bock M, Özen AC. Characterizing Off-center MRI with ZTE. Z Med Phys. 2024;34(3):446–55. doi: 10.1016/j.zemedi.2022.09.002 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Beveridge JE, Machan JT, Walsh EG, Kiapour AM, Karamchedu NP, Chin KE, et al. Magnetic resonance measurements of tissue quantity and quality using T2 * relaxometry predict temporal changes in the biomechanical properties of the healing ACL. J Orthop Res. 2018;36(6):1701–9. doi: 10.1002/jor.23830 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Gao J, Xu X, Yu X, Fu Y, Zhang H, Gu S, et al. Quantitatively relating magnetic resonance T 1 and T 2 to glycosaminoglycan and collagen concentrations mediated by penetrated contrast agents and biomacromolecule-bound water. Regen Biomater. 2023;10: rbad035. doi: 10.1093/rb/rbad035 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Aro O-P. Comparison of relaxation properties of collagen gel MRI phantoms at multiple magnetic fields. University of Oulu. 2020. Available: https://urn.fi/URN:NBN:fi:oulu-202008152795 [Google Scholar]
- 32.Martin MN, Jordanova KV, Kos AB, Russek SE, Keenan KE, Stupic KF. Relaxation measurements of an MRI system phantom at low magnetic field strengths. MAGMA. 2023;36(3):477–85. doi: 10.1007/s10334-023-01086-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Stanisz GJ, Odrobina EE, Pun J, Escaravage M, Graham SJ, Bronskill MJ, et al. T1, T2 relaxation and magnetization transfer in tissue at 3T. Magn Reson Med. 2005;54(3):507–12. doi: 10.1002/mrm.20605 [DOI] [PubMed] [Google Scholar]
- 34.Horch RA, Gore JC, Does MD. Origins of the ultrashort-T2 1H NMR signals in myelinated nerve: a direct measure of myelin content?. Magn Reson Med. 2011;66(1):24–31. doi: 10.1002/mrm.22980 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Ketsiri T, Uppuganti S, Harkins KD, Gochberg DF, Nyman JS, Does MD. T1 relaxation of bound and pore water in cortical bone. NMR Biomed. 2023;36(5):e4878. doi: 10.1002/nbm.4878 [DOI] [PubMed] [Google Scholar]
- 36.Filho GH, Du J, Pak BC, Statum S, Znamorowski R, Haghighi P, et al. Quantitative characterization of the Achilles tendon in cadaveric specimens: T1 and T2* measurements using ultrashort-TE MRI at 3 T. AJR Am J Roentgenol. 2009;192(3):W117-24. doi: 10.2214/AJR.07.3990 [DOI] [PubMed] [Google Scholar]
- 37.Ma Y, Zhao W, Wan L, Guo T, Searleman A, Jang H, et al. Whole knee joint T1 values measured in vivo at 3T by combined 3D ultrashort echo time cones actual flip angle and variable flip angle methods. Magnet Resonance Med. 2018;81(3):1634–44. doi: 10.1002/mrm.27510 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Richard S, Querleux B, Bittoun J, Jolivet O, Idy-Peretti I, de Lacharriere O, et al. Characterization of the skin in vivo by high resolution magnetic resonance imaging: water behavior and age-related effects. J Invest Dermatol. 1993;100(5):705–9. doi: 10.1111/1523-1747.ep12472356 [DOI] [PubMed] [Google Scholar]
- 39.Hager B, Schreiner MM, Walzer SM, Hirtler L, Mlynarik V, Berg A, et al. transverse relaxation anisotropy of the achilles and patellar tendon studied by MR Microscopy. J Magn Reson Imaging. 2022;56(4):1091–103. doi: 10.1002/jmri.28095 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Williams AA, Titchenal MR, Do BH, Guha A, Chu CR. MRI UTE-T2* shows high incidence of cartilage subsurface matrix changes 2 years after ACL reconstruction. J Orthop Res. 2019;37(2):370–7. doi: 10.1002/jor.24110 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Du J, Diaz E, Carl M, Bae W, Chung CB, Bydder GM. Ultrashort echo time imaging with bicomponent analysis. Magn Reson Med. 2012;67(3):645–9. doi: 10.1002/mrm.23047 [DOI] [PubMed] [Google Scholar]
- 42.Song HK, Wehrli FW, Ma J. In vivo MR microscopy of the human skin. Magn Reson Med. 1997;37(2):185–91. doi: 10.1002/mrm.1910370207 [DOI] [PubMed] [Google Scholar]
- 43.Mirrashed F, Sharp JC. In vivo morphological characterisation of skin by MRI micro-imaging methods. Skin Res Technol. 2004;10(3):149–60. doi: 10.1111/j.1600-0846.2004.00071.x [DOI] [PubMed] [Google Scholar]
- 44.Kajabi AW, Zbýň Š, Smith JS, Hedayati E, Knutsen K, Tollefson LV, et al. Seven tesla knee MRI T2*-mapping detects intrasubstance meniscus degeneration in patients with posterior root tears. Radiol Adv. 2024;1(1):umae005. doi: 10.1093/radadv/umae005 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Kirsch S, Kreinest M, Reisig G, Schwarz MLR, Ströbel P, Schad LR. In vitro mapping of 1H ultrashort T2* and T2 of porcine menisci. NMR Biomed. 2013;26(9):1167–75. doi: 10.1002/nbm.2931 [DOI] [PubMed] [Google Scholar]
- 46.Liu J, Nazaran A, Ma Y, Chen H, Zhu Y, Du J, et al. Single- and Bicomponent Analyses of T2* relaxation in knee tendon and ligament by using 3d ultrashort echo time cones (UTE Cones) Magnetic Resonance Imaging. Biomed Res Int. 2019;2019:8597423. doi: 10.1155/2019/8597423 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Juras V, Schreiner M, Laurent D, Zbýň Š, Mlynarik V, Szomolanyi P, et al. The comparison of the performance of 3 T and 7 T T2 mapping for untreated low-grade cartilage lesions. Magn Reson Imaging. 2019;55:86–92. doi: 10.1016/j.mri.2018.09.021 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Anjum MAR, Gonzalez FM, Swain A, Leisen J, Hosseini Z, Singer A, et al. Multi-component T2* relaxation modelling in human Achilles tendon: quantifying chemical shift information in ultra-short echo time imaging. Magn Reson Med. 2021;86(1):415–28. doi: 10.1002/mrm.28686 [DOI] [PubMed] [Google Scholar]
- 49.Liu F, Kijowski R. Assessment of different fitting methods for in-vivo bi-component T2* analysis of human patellar tendon in magnetic resonance imaging. Muscles Ligaments Tendons J. 2017;7(1):163–72. doi: 10.11138/mltj/2017.7.1.163 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Do HP, Lockard CA, Berkeley D, Tymkiw B, Dulude N, Tashman S, et al. improved resolution and image quality of musculoskeletal magnetic resonance imaging using deep learning-based denoising reconstruction: a prospective clinical study. Skeletal Radiol. 2024;53(12):2585–96. doi: 10.1007/s00256-024-04679-3 [DOI] [PubMed] [Google Scholar]
- 51.Getzmann JM, Huber FA, Nakhostin D, Deininger-Czermak E, Schumann P, Finkenstaedt T, et al. Impact of acceleration on bone depiction quality by ultrashort echo time magnetic resonance bone imaging sequences in medication-related osteonecrosis of the jaw. Eur J Radiol Open. 2022;9:100421. doi: 10.1016/j.ejro.2022.100421 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.Van Dyck P, Vanhevel F, Vanhoenacker FM, Wouters K, Grodzki DM, Gielen JL, et al. Morphological MR imaging of the articular cartilage of the knee at 3 T—comparison of standard and novel 3D sequences. Insights Imaging. 2015;6: 285–293. doi: 10.1007/s13244-015-0405-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53.Ensle F, Kaniewska M, Lohezic M, Guggenberger R. Enhanced bone assessment of the shoulder using zero-echo time MRI with deep-learning image reconstruction. Skeletal Radiol. 2024;53(12):2597–606. doi: 10.1007/s00256-024-04690-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54.Carretero-Gómez L, Fung M, Wiesinger F, Carl M, McKinnon G, de Arcos J, et al. Deep learning-enhanced zero echo time MRI for glenohumeral assessment in shoulder instability: a comparative study with CT. Skeletal Radiol. 2024:10.1007/s00256-024-04830–0. doi: 10.1007/s00256-024-04830-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55.A fast algorithm to correct excitation profile in zero echo time (ZTE) imaging. Proc 23rd Ann Meet ISMRM. 2015:3733. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
(DOCX)
The median and interquartile range values are calculated for the sample of all voxels within each ROI.
(DOCX)
The median and interquartile range values are calculated for the sample of all voxels within each ROI.
(DOCX)
The results calculated using all seven TE values and the subset of three TE values are presented. The median value represents the median of the three median values (one per participant, for each ROI) calculated for each ROI. The range represents the minimum and maximum participant ROI median values.
(DOCX)
The results calculated using all seven TE values and the subset of three TE values are presented. The median value represents the median of the three median values (one per participant, for each ROI) calculated for each ROI. The range represents the minimum and maximum participant ROI median values.
(DOCX)
The results calculated using all seven TE values and the subset of three TE values are presented for all voxels and for only voxels with acceptable ultrashort-T2* fit (R2 ≥ 0.5). The median value represents the median of the three median values (one per participant, for each ROI) calculated for each ROI. The range represents the minimum and maximum participant ROI median values.
(DOCX)
The results calculated using all seven TE values and the subset of three TE values are presented for all voxels and for only voxels with acceptable ultrashort-T2* fit (R2 ≥ 0.5). The median and interquartile range values are calculated for the sample of all voxels within each ROI.
(DOCX)
(TIF)
(TIF)
Plots are shown for the values based on 7-TE and 3-TE and include all ultrashort-T2* values (all R2).
(TIF)
Ultrashort-T2* results are for all ultrashort-T2* values (all R2).
(DOCX)
Data Availability Statement
The data cannot be shared publicly in an unrestricted manner because the Institutional Review Board protocol under which participants were consented only permits sharing of the data with qualified investigators, who have current human subjects research ethics training, at academic institutions or private research organizations. The PETRA image data are available to any qualified investigator at an academic institution or private research organization who has current human subjects research ethics training certification and can describe a scientific use for the data. Interested parties can use the following link (https://redcap.link/AR073831_Data_Request) to initiate a Data Use Agreement.