On the Fat Saturation Effect in Quantitative Ultrashort Echo Time MR Imaging

Abstract

Purpose:

To investigate the effect of fat saturation (FatSat) on quantitative ultrashort echo time (UTE) imaging of variable knee tissues on a 3T scanner.

Materials and Methods:

Three quantitative UTE imaging techniques, including the UTE multi-echo sequence for T₂* measurement, the adiabatic T_1ρ prepared UTE sequence for T_1ρ measurement, and the magnetization transfer (MT)-prepared UTE sequence for MT ratio (MTR) and macromolecular proton fraction (MMF) measurements were employed in this study. Twelve samples of cartilage and twelve samples of meniscus, as well as six whole knee cadaveric specimens, were imaged with the three abovementioned UTE sequences with and without FatSat. The difference, correlation, and agreement between the UTE measurements with and without FatSat were calculated to investigate the effects of FatSat on quantification.

Results:

Fat was well-suppressed using all three UTE sequences when FatSat was deployed. For the small sample study, the quantification difference ratio (QDR) values of all the measured biomarkers ranged from 0.7% to 12.6%, whereas for the whole knee joint specimen study, the QDR values ranged from 0.2% to 12.0%. Except for T_1ρ in muscle and MMF in meniscus (P>0.05), most of the measurements showed statistical differences for T_1ρ, MTR, and MMF (P<0.05) between FatSat and non-FatSat scans. Most of the measurements for T₂* showed no significant differences (P>0.05). Strong correlations were found for all the biomarkers between measurements with and without FatSat.

Conclusion:

The UTE biomarkers showed good correlation and agreement with some slight differences between the scans with and without FatSat.

Introduction

Knee osteoarthritis (OA) is a whole-organ disease, with multiple studies that have reported that the deterioration or misalignment of any one of the constituent knee joint tissues has the potential to accelerate overall disease progression (1–3). For example, meniscal tears or collateral ligament damage can lead to loss of cartilage (2). As such, a truly comprehensive assessment of OA requires imaging of all the knee joint’s major components, including the articular cartilage, menisci, ligaments, tendons, and muscle because they all play an essential role in joint health and pathology. However, conventional MRI techniques, such as gradient recalled echo (GRE) or fast spin echo (FSE) sequences, cannot effectively detect signals from tissues in the knee joint with short T₂ relaxation times, such as deep and calcified cartilage, menisci, ligaments, and tendons (4), limiting them in their utility for whole-knee assessment.

Fortunately, specialized pulse sequences, namely ultrashort echo time (UTE) sequences with echo times (TEs) less than 100μs have been successfully developed for optimal imaging of those tissues with short T₂ relaxation times (5,6). Fat suppression is a crucial component in the effective application of these non-Cartesian UTE signal acquisitions. Fat produces a high signal due to its high proton density and short T₁ relaxation time that can shift into the adjacent tissues radially (i.e., off-resonance artifacts) (7). This contributes to the partial volume effect and introduces significant errors in quantitative UTE imaging. Currently, one of the most widely used fat suppression approaches available on MRI systems is chemical shift-based fat saturation (FatSat), which employs a spectrally selective radiofrequency (RF) pulse (around 90°) followed by a gradient to dephase all the transverse magnetizations. However, because tissues with short T₂ relaxation times have a broad frequency spectrum that may be close to or overlap with the signal peaks from fat, the signals from tissues with short T₂ relaxation times can be attenuated as a result of inadvertent direct saturation by the FatSat pulse (8). Furthermore, indirect saturation of collagen-rich tissues with either short or long T₂ relaxation times can also be induced by the FatSat pulse as a result of the MT effects.

Recently, quantitative 3D UTE imaging techniques have been developed to allow for the quantification of tissues with both short and long T₂ relaxation times in musculoskeletal system. These techniques include the UTE multi-echo sequence for T₂* measurement (9), the adiabatic T_1ρ-prepared UTE (UTE-Adiab-T_1ρ) sequence for T_1ρ measurement (10), and the magnetization transfer-prepared UTE (UTE-MT) sequence for MT ratio (MTR) and macromolecular proton fraction (MMF) measurements (11). These UTE sequences can be combined into a single protocol for comprehensive quantitative evaluation of all the major tissues in the knee joint. It is worth noting that previous studies have reported that Adiab-T_1ρ and MMF are much less sensitive to the magic angle effect than conventional continuous wave (CW) T_1ρ and T₂ sequences (12–15), an important consideration when studying collagen-rich tissues.

In this study, we compared the measurements from three quantitative UTE imaging sequences with and without FatSat to investigate the effects of fat suppression on UTE quantification of biomarkers including T₂*, T_1ρ, MTR, and MMF. Small osteochondral and meniscus samples, as well as whole knee cadaveric specimens, were studied on a clinical 3T scanner. The difference, correlation, and agreement between the UTE measurements with and without FatSat were calculated to investigate the effects of FatSat on quantification.

Methods

Sample preparation

Local institutional review board approval was received for this study. A total of eight fresh-frozen human cadaveric knee joint specimens (38 to 67-years-old, 4 males) were obtained from a nonprofit whole-body donation company (United Tissue Network, Phoenix, AZ). Twelve osteochondral and twelve meniscus samples were harvested from two joints (51-year-old male; 47-year-old female). The osteochondral samples were cut from the patella and tibia using a Delta ShopMaster band saw (Delta Machinery, Jackson, TN) with a thickness of about 5 mm. To investigate the effectiveness of fat suppression, marrow fat that was present in the trabecular bone was preserved in the osteochondral samples. The meniscal samples were sectioned in the sagittal plane with a thickness of about 4 mm using a scalpel. Adjacent synovial and fatty tissues were preserved in the meniscal samples. All twenty-four of the small osteochondral and meniscus samples were soaked in phosphate-buffered saline for about 2 hours, then placed in a plastic syringe filled with Fomblin to minimize dehydration and susceptibility before scanning. The remaining six whole knee joints were thawed in a water bath for 20–24 hours before scanning. All scans were performed at room temperature.

UTE MR imaging

All imaging was performed on a 3T clinical MRI scanner (MR750, GE Healthcare Technologies, Milwaukee, WI). The small osteochondral and meniscus samples were scanned using a homemade 30-ml birdcage coil with a high signal-to-noise ratio (SNR) performance, while the whole knee samples were scanned using an eight-channel transmit and receive knee coil.

Figure 1 shows a diagram of the quantitative 3D UTE sequences. A multi-echo UTE sequence was used for T₂* measurement (Figure 1A) (16,17). As can be seen in Figure 1B, a train of adiabatic full passage (AFP) pulses (hyperbolic secant shape, duration of 6.048ms, and bandwidth of 1.643kHz) was used to lock the spin and generate the T_1ρ contrast (10). The adiabatic T_1ρ preparation used was much less sensitive to B₁ inhomogeneity and magic angle effect than the conventional CW spin lock preparation (12). A Fermi pulse with a duration of 8ms and a bandwidth of 160 Hz was used to saturate the magnetization of the macromolecular proton pool to generate the MT contrast (Figure 1C) (11). All the UTE sequences use a short rectangular pulse for excitation followed by 3D Cones sampling (18).

Figure 1.

The diagram of the quantitative 3D UTE sequences. A multi-echo UTE sequence is used for T₂* measurement (A). A train of adiabatic full passage pulses (hyperbolic secant shape, duration of 6.048 ms, and bandwidth of 1.643 kHz) were used to lock the spin to generate the T_1ρ contrast (B). A Fermi pulse with a duration of 8 ms and a bandwidth of 160 Hz was used to saturate the magnetization of the macromolecular proton pool to generate the MT contrast (C). The AFI sequence with interleaved two TRs (D) was used to measure and correct B₁ inhomogeneity for all quantifications.

The standard FatSat module available in GE scanners (a minimum-phase Shinnar-Le Roux (mip-SLR) RF pulse design, duration of 8 ms, bandwidth of 500Hz, center frequency of −440Hz, and flip angle (FA) of 100°) was used for fat suppression in all the evaluated quantitative UTE sequences. When the FatSat option is activated, the FatSat module is placed before the excitation pulse in the multi-echo UTE sequence, between the AFP pulse train and the first excitation pulse in the UTE-Adiab-T_1ρ sequence, and between the MT pulse and the first excitation pulse in the UTE-MT sequence. T₁ values were estimated using a variable flip angle method (19), which was utilized for both T_1ρ and MMF calculations (20). Figure 1D shows the actual flip angle imaging (AFI) sequence (21), which was used to measure B₁ inhomogeneity maps to correct the excitation flip angles for all the quantitative UTE measurements.

The detailed sequence parameters of the quantitative UTE imaging protocols were: A) 3D multi-echo UTE sequence with an echo train of six TEs=0.032, 4.4, 8.8, 13.2, 22, and 35.2ms; repetition time (TR)=100ms; and FA=10°; B) 3D UTE-Adiab-T_1ρ sequence: spin lock times (TSLs) were 0, 12, 24, 36, 48, 72, and 96ms corresponding to the number of AFP pulses of 0, 2, 4, 6, 8, 12, and 16; TR=500ms; FA=10°; number of spokes (N_sp) in a TR=21; and interspoke interval τ=6ms (22); C) 3D UTE-MT sequence: MT pulse FAs=500°, 1000°, and 1500°; MT pulse frequency offsets=2, 5, 10, 20, and 50kHz; TR=100ms; FA=7°; N_sp=9; and interspoke interval τ=6ms (11); D) 3D UTE-AFI sequence: TR₁/TR₂=20/100ms and FA=45°; and E) 3D UTE-VFA sequence: FAs=5, 10, 20, and 30° and TR=20ms (21). The field of view (FOV), matrix size and scan time for the small sample study were 5×5×3.6cm³, 160×60×60 and 110min respectively. The FOV, matrix size and scan time for the whole knee joint specimen study were 15×15×9cm³, 256×256×30 and 90min, respectively. Each sequence was scanned twice, once with and once without FatSat preparation.

Data analysis

The UTE data processing code was written in MATLAB 2018a (MathWorks, Natick, MA). The Levenberg-Marquardt method was used for nonlinear least-squares curve fitting.

For small specimens, regions of interest (ROIs) were drawn on the regions without fat. For knee joint specimens, two different ROIs were drawn on each of the major tissue components independently, including cartilage (1 patellar and 1 femur), meniscus (anterior and posterior horn), quadriceps tendon (upper and lower half parts), posterior cruciate ligament (PCL) (upper and lower half parts), and muscle (gastrocnemius and vastus medialis). The mean intensity within each ROI was used for fitting. All ROIs were carefully selected to ensure that no fat signal was included. Because it is difficult to draw proper ROIs without fat contamination for the anterior cruciate ligament (ACL), this tissue component was not included in the study to avoid any potential bias in the comparisons between UTE quantifications with and without fat suppression. To investigate the impact of fat signal contamination in quantitative UTE measurement, ROIs were also manually drawn in the muscle regions (i.e., gastrocnemius and vastus medialis) that were contaminated with fat.

MTR is calculated by the following equation:

$$\text{MTR}={(\text{MT}}_{\text{off}}{-\text{MT}}_{\text{on}})/{\text{MT}}_{\text{off}}$$ [1]

MT_on and MT_off represent the UTE scans with MT pulse power on and off, respectively. In this study, MT_on data was chosen from the UTE-MT data with MT pulse FA=500° and frequency offset=2kHz. MT_off data was chosen from the UTE-MT data with MT pulse FA=500° and frequency offset=50kHz, which has almost no MT effect-induced signal reduction due to the large frequency offset.

The MMF value can be obtained by fitting a two-pool MT model. This two-pool model divides the spins within a tissue into two compartments (i.e., a water pool and a macromolecular pool) which can then be expressed with a series of Bloch equations (11). Simplified using a rectangular pulse approximation of the MT pulse, these Bloch equations were solved in a matrix form, the final expression of which can be found in Eq. [A7] from Ref. 11. There are five unknown parameters, including the MMF, in this formula. To achieve accurate estimation of these unknown parameters, 15 UTE-MT data (see the Method section) with different MT contrast were acquired to fit this model. More details about the fitting process can be found in Ref. 11.

To evaluate changes in the quantitative UTE biomarkers that were a result of the FatSat module, a quantification difference ratio (QDR, percentage unit) was used, defined as the division of the absolute difference between non-fat suppressed UTE quantification (Q_nonFS) and fat suppressed quantification (Q_FS) divided by the non-fat suppressed quantification, as seen in Equation 1.

$$\text{QDR}=\frac{{|\mathit{\text{Q}}}_{\mathit{\text{nonFS}}}{-\mathit{\text{Q}}}_{\mathit{\text{FS}}}|}{{\mathit{\text{Q}}}_{\mathit{\text{nonFS}}}}\times 100\%$$ [2]

Statistical analysis

Statistical analyses were performed using SPSS software (version 21; SPSS, Chicago, IL, USA). The two ROI data pieces in each anatomical region for the knee joint specimen study were handled independently for statistical analysis. Numerical data were reported as the mean ± standard deviation (SD) of all three sets of measurement. Intraclass correlation coefficient (ICC) of all the quantitative measurements with and without FatSat was calculated. Bland-Altman analysis was also used to assess the agreement between these measurements. The paired Wilcoxon Rank-Sum test was used to compare each biomarker between the measures with and without FatSat, and P < 0.05 was considered statistically significant.

Results

Representative quantitative UTE images for the small sample study with and without the FatSat module are shown in Supporting Information Figures S1 and the two different sets of representative whole specimen images are shown in Figures 2 and Supporting Information Figure 2 respectively. Fat was well-suppressed in all three UTE sequences when FatSat was used.

Figure 2.

Representative UTE images (i.e., UTE multi-echo sequence with TE = 0.032 ms (first column), UTE-Adiab-T_1ρ with TSL = 0 ms (second column), and UTE-MT with MT flip angle of 500° and frequency offset of 2 kHz (third column)) with (second row) and without FatSat (first row) for a knee joint specimen.

Supporting Figure S3 shows the representative fitting curves of UTE-Adiab-T_1ρ and UTE-MT modeling as well as the corresponding quantitative T_1ρ and MMF values for both cartilage and meniscus from a knee specimen. Small fitting errors were found for all data with and without FatSat (including T₂* measurement, whose fitting curves were not shown for simplicity), demonstrating the reliability of the quantification. Supporting Information Figures S4 shows the corresponding T_1ρ, MTR, MMF and T₂* maps for the non-FatSat and FatSat scans.

Representative ROIs for cartilage, meniscus, quadriceps tendon, PCL, and muscle are drawn on the UTE images from a knee specimen (see Supporting Information Figures S5. Supporting Information Figures S6 and S7 show the quantitative results of the small samples and knee specimens respectively. In general, the T₂*, T_1ρ, and MTR values were slightly decreased while the MMF values were slightly increased when the FatSat module is used. Table 1 summarizes all the measurements together with QDR values. For the small sample study, the QDR values of T₂* and T_1ρ measurements for both cartilage and meniscus were quite small, lower than 6%. The QDR values of MT measurements (including both MTR and MMF) were higher, ranging from 5.2 to 12.6%. The QDR results demonstrated that the FatSat module had more of an effect on MT measurements than it did on T₂* and T_1ρ. Significant differences were also observed for the T_1ρ, MTR, and MMF measurements between scans with and without the FatSat module (P<0.05), whereas no significant differences were found for T₂* measurements (P>0.05).

Table 1.

UTE measurements with and without FatSat and corresponding QDR values for all samples.

			T2*(ms)	T1ρ (ms)	MTR (%)	MMF (%)
Small sample	Cartilage	Non-FatSat	9.95±1.27	47.76±5.02	13.35±1.76	11.58±1.17
		FatSat	9.51±1.12	46.51±4.81	12.66±1.55	12.60±1.09
		QDR (%)	4.4	2.6	5.2	8.8
		P-value	0.071	0.034	0.041	0.002
	Meniscus	Non-FatSat	7.90±0.59	26.24±1.93	17.28±1.62	16.00±1.72
		FatSat	7.85±0.40	24.78±2.07	15.59±1.47	18.01±1.96
		QDR (%)	0.7	5.6	9.8	12.6
		P-value	0.477	0.003	0.004	0.003
Whole knee specimen	Cartilage	Non-FatSat	34.11±3.33	48.15±4.69	14.25±1.33	10.21±1.33
		FatSat	34.03±3.34	45.88±4.63	13.21±1.15	10.74±1.28
		QDR (%)	0.2	4.7	7.3	5.1
		P-value	0.937	0.034	0.012	0.028
	Meniscus	Non-FatSat	13.11±1.35	32.28±3.54	18.18±2.53	14.23±1.69
		FatSat	12.84±1.38	30.31±3.09	16.57±2.28	14.62±1.65
		QDR (%)	2.0	6.1	8.8	2.7
		P-value	0.239	0.010	0.002	0.136
	Quadriceps tendon	Non-FatSat	11.28±1.26	22.36±1.96	16.67±1.77	15.53±1.73
		FatSat	12.16±1.66	20.43±1.97	15.24±1.60	17.39±1.78
		QDR (%)	7.7	8.6	9.6	12.0
		P-value	0.023	0.002	0.012	0.002
	PCL	Non-FatSat	19.39±1.88	38.30±3.67	15.06±1.72	12.35±0.94
		FatSat	18.98±1.90	37.15±3.72	13.75±1.91	12.87±1.05
		QDR (%)	2.1	3.0	8.7	4.2
		P-value	0.117	0.023	0.002	0.034
	Muscle	Non-FatSat	31.24±3.46	63.16±5.18	9.54±0.87	6.04±0.28
		FatSat	32.71±3.38	62.06±5.23	8.97±0.79	6.44±0.31
		QDR (%)	4.7	1.7	5.9	6.6
		P-value	0.099	0.239	0.012	0.012

For the knee specimen study, FatSat had a relatively larger impact on all measurements for the quadriceps tendon (QDR≥7.7%) than it did for other tissues. The longer T₂* value of the quadriceps tendon measured with the FatSat UTE sequence was probably caused by strong saturation of the short T₂ components. The direct saturation induced by the FatSat is significant in quadriceps tendon imaging because approximately 80% of the signal in tendon tissues comes from short T₂/T₂* components (23,24). For other tissues, the T₂* values did not change much, nor were any significant differences found between the non-FatSat and FatSat scans (P>0.05). The T_1ρ and MMF values were also generally similar for all other tissues between the scans with and without FatSat (QDR≤6.6%). Interestingly, the tissues that have shorter T₂*s and T_1ρs relaxations had higher MTR and MMF values. This is probably because of the differences in collagen content and integrity between tissues. Most of the measurements show significant differences for T_1ρ, MTR, MMF (P<0.05) between scans with and without FatSat except for T_1ρ measurements in muscle and MMF in meniscus (P>0.05).

The T₂*, T_1ρ, MTR, and MMF values for the fat-contaminated muscle group obtained from the non-FatSat UTE scans were 21.19ms, 61.78ms, 6.74%, and 6.47% respectively. The corresponding measurement errors related to fat contamination were 32.2%, 2.2%, 29.4%, and 7.1% when the UTE measurements of the fat-free muscle group obtained from the non-FatSat scans were regarded as reference standard. These measurement errors were higher than the errors in the FatSat scans (i.e., 4.7%, 1.7%, 5.9%, and 6.6%), demonstrating that the errors associated with fat signal contamination can be more impactful than those associated with the FatSat scans.

Figure 3 shows the correlation curves and Bland-Altman analysis for all the UTE biomarkers with and without FatSat and Table 2 summarizes the corresponding ICC analysis results between scans with and without FatSat. Strong correlation (ICC>0.95) and agreement were found for all biomarkers.

Figure 3.

Correlation curves (left column) and Bland-Altman (right column) plots of T₂* (A and B), T_1ρ (C and D), MTR (E and F) and MMF (G and H) with and without FatSat.

Table 2.

ICC analysis results for all the measured UTE biomarkers between non-Fat and FatSat scans.

	Biomarkers	ICC	95% CI
Small sample	T2*	0.966	0.921 to 0.985
	T1ρ	0.995	0.989 to 0.998
	MTR	0.949	0.883 to 0.978
	MMF	0.942	0.869 to 0.975
Whole knee specimen	T2 *	0.991	0.986 to 0.995
	T1ρ	0.983	0.973 to 0.990
	MTR	0.976	0.961 to 0.985
	MMF	0.982	0.971 to 0.989

Discussion

Both small sample and whole knee joint specimen studies were performed to investigate the effect of fat suppression on quantitative UTE MR imaging techniques including UTE multi-echo acquisition for T₂* measurement, UTE-Adiab-T_1ρ acquisition for T_1ρ measurement, and UTE-MT acquisition for MTR and MMF measurements. The standard FatSat module was used for fat suppression in all the quantitative UTE techniques. We found that all the derived quantitative UTE biomarkers (i.e., T₂*, T_1ρ, MTR, and MMF) had good correlation and agreement between the scans with and without FatSat. There were slight increases in MTR and MMF measurements and slight decreases in T₂* and T_1ρ measurements for most investigated tissues once FatSat was activated. As a result, we believe that these quantitative UTE MRI techniques with FatSat are still valuable for many applications, particularly in cases where fat suppression is necessary.

In the literature, the average T₂ and T_1ρ values of cartilage were changed 14.1% and 14.3%, respectively, in OA patients compared with controls (25). Compared to our knee specimen study, these changes are greater than those induced by FatSat for T₂* (0.2%) and T_1ρ (4.7%) values in cartilage. For meniscus, the reported average T₂ and T_1ρ values were changed 58.9% and 70.2%, respectively, in advanced OA patients compared with controls (26), which are much higher than the FatSat-induced T₂* (2.0%) and T_1ρ (6.1%) changes in our study. Zhang et al. also reported an average MMF change of 9.6% in meniscus for mild OA patients compared with controls (27), which is much higher than that induced by FatSat (2.7%) in our study. Hodgson et al. reported a 21.9% MMF change in the Achilles tendon for patients with psoriatic arthritis compared with controls (28), and Zhu et al. reported an average 26.8% MMF change in the rotator cuff tendon when comparing between patients with mild and severe tendinopathy (15). The MMF changes reported in both these tendon studies are higher than the FatSat-induced change (12.0%) observed in the quadriceps tendon for our study. Therefore, quantitative metric changes induced by pathology in the musculoskeletal system may be higher than those induced by the FatSat module. The Bland-Altman analysis showed high agreement in all the measured biomarkers between FatSat and non-FatSat scans. High correlations were also found, demonstrating that the quantitative UTE technique with FatSat may be still feasible to detect compositional changes in tissue.

The discrepancy in cartilage T₂* values between the small and whole knee specimens was likely induced by the magic angle effect. During the small cartilage sample scan, the cartilage-subchondral bone interface was nearly perpendicular to the B₀ direction, meaning that the major collagen fibers in the cartilage were nearly parallel to the B₀ direction. Of all possible angle orientations, this placement results in the shortest possible T₂* values of cartilage (13,14). For the whole knee specimen study, two different ROIs were drawn in the patellar and femur cartilage (see Supporting Information Figure S5), respectively. The cartilage collagen fibers in these two ROIs deviated largely from the B₀ direction. In these orientations, much longer T₂* values are observed compared with scans where the angle orientation is close to 0°. In comparison to T₂*, Adiab-T_1ρ has been shown to be much less sensitive to the magic angle effect (12,13). Therefore, it was unsurprising to observe fewer variations in the Adiab-T_1ρ measurements between the small sample and whole knee specimen studies.

Consistent with previous studies, the T₂* values measured from the fat suppressed sequences were generally lower than those without fat suppression (29,30). However, these T₂* value did not change much and no significant differences were found for most of the investigated tissues. Fat suppression is always applied to the T_1ρ quantitation sequences since spin lock preparation may be susceptible to the chemical shift of fat (31). In our study, we found that—similar to what was seen with T₂*—T_1ρ values decreased when FatSat was used. Decreased MTR values were obtained from the FatSat UTE MRI sequence, similar to a human calf muscle study that used the MT saturation technique (32). The MMF values increased when FatSat was applied in our study. A simulation study was performed by Li K et al. to investigate the effect of different fat fractions on MMF calculation using a two-pool MT model (33). They found the MMF values first decreased, then increased when the fat fraction was higher, demonstrating the importance of fat suppression in accurate MMF calculation. In our study, the ROIs were carefully selected to eliminate the fat-induced change in MMF values, but the increased MMF values may have been produced by the decreased MT effect introduced by the FatSat pulse. Overall, the QDR values of MTR and MMF were higher than those of T₂* and T_1ρ, suggesting that these two techniques are more affected by the introduction of the FatSat module.

As can been seen in Figures 2A to 2C, some black/white banding artifacts were observed in the fatty regions. These bandings are likely off-resonance artifacts induced by both the chemical shifts of fat and the significant B₀ inhomogeneity in those regions (34). Moreover, the relatively long spiral trajectory (~1.7ms) used for gradient encoding further exacerbated these off-resonance artifacts

This study had limitations. First, no in vivo study was performed. However, our ex vivo study design benefitted from the advantages of long scan time that allowed us to investigate all three UTE sequences without concerns regarding the duration of scans. Our ex vivo results were also more reliable because there were no motion-related artifacts compared to an in vivo study. Second, the total scan time of the ex vivo protocol is too long (i.e., 90min) for clinical use. The total number of acquisitions can be reduced to shorten the scan time for each technique. Moreover, the trajectory stretching strategy can be applied to these sequences to improve the scan efficiency (35).

Conclusion

The UTE biomarkers showed good correlation and agreement with some slight differences between the scans with and without FatSat. The quantitative UTE techniques with FatSat could potentially be used to access the compositional changes in knee OA.

Supplementary Material

Supplemental Figures

Supporting Information Figure S1. Representative UTE images (i.e., UTE multi-echo sequence with TE = 0.032 ms (first column), UTE-Adiab-T_1ρ with TSL = 0 ms (second column), and UTE-MT with MT flip angle of 500° and frequency offset of 2 kHz (third column)) with (second and fourth rows) and without FatSat (first and third rows) for small osteochondral (first two rows) and meniscus (last two rows) samples.

Supporting Information Figure S2. Representative UTE images (i.e., UTE multi-echo sequence with TE = 8.8 ms (first column), UTE-Adiab-T_1ρ with TSL = 36 ms (second column), and UTE-MT with MT flip angle of 1000° and frequency offset of 10 kHz (third column)) with (second row) and without FatSat (first row) for another knee joint specimen.

Supporting Information Figure S3. The representative fitting curves of UTE-Adiab-T_1ρ and UTE-MT modeling as well as the corresponding quantitative T_1ρ and MMF values for both cartilage and meniscus from a whole knee specimen. The T_1ρ and MMF values were presented as fitted value ± standard error.

Supporting Information Figure S4. The quantified T_1ρ (first column), MTR (second column), MMF (third column) and T₂* (last column) maps for the non-FatSat (first row) and FatSat (second row) scans from a representative knee specimen.

Supporting Information Figure S5. Representative ROIs for cartilage, meniscus, quadriceps tendon, PCL, and muscle drawn on the UTE images from a knee specimen. Two different ROIs were drawn for each tissue.

Supporting Information Figure S6. The quantitative T₂* (A), T_1ρ (B), MTR (C) and MMF (D) values of all the small cartilage and meniscus samples (* represents a statistical difference).

Supporting Information Figure S7. The quantitative T₂* (A), T_1ρ (B), MTR (C) and MMF (D) values of cartilage, meniscus, quadriceps tendon, PCL, and muscle in the four whole knee joint specimens (* represents a statistical difference).