Whole Knee Joint Mapping Using a Phase Modulated Ultrashort Echo Time Adiabatic T_1ρ (PM-UTE-AdiabT_1ρ) Sequence

Abstract

Purpose:

To develop a 3D phase modulated ultrashort echo time adiabatic T_1ρ (PM-UTE-AdiabT_1ρ) sequence for whole knee joint mapping on a clinical 3T scanner.

Methods:

This new sequence includes six major features: 1) a magnetization reset module, 2) a train of adiabatic full passage pulses for spin locking, 3) a phase modulation scheme (i.e., RF cycling pair), 4) a fat saturation module, 5) a variable flip angle scheme, and 6) a 3D UTE Cones sequence for data acquisition. A simple exponential fitting was used for T_1ρ quantification. Phantom studies were performed to investigate PM-UTE-AdiabT_1ρ’s sensitivity to compositional changes and reproducibility as well as its correlation with continuous-wave T_1ρ (CW-T_1ρ) measurement. The PM-UTE-AdiabT_1ρ technique was then applied to five ex vivo and five in vivo normal knees to measure T_1ρ values of femoral cartilage, meniscus, posterior cruciate ligament (PCL), anterior cruciate ligament (ACL), patellar tendon, and muscle.

Results:

The phantom study demonstrated PM-UTE-AdiabT_1ρ’s high sensitivity to compositional changes, its high reproducibility, and its strong linear correlation with CW-T_1ρ measurement. The ex vivo and in vivo knee studies demonstrated average T_1ρ values of 105.6±8.4 and 77.9±3.9ms for the femoral cartilage, 39.2±5.1 and 30.1±2.2ms for the meniscus, 51.6±5.3 and 29.2±2.4ms for the PCL, 79.0±9.3 and 52.0±3.1ms for the ACL, 19.8±4.5 and 17.0±1.8ms for the patellar tendon, and 91.1±8.8 and 57.6±2.8ms for the muscle, respectively.

Conclusion:

The 3D PM-UTE-AdiabT_1ρ sequence allows volumetric T_1ρ assessment for both short and long T₂ tissues in the knee joint on a clinical 3T scanner.

Introduction

Spin lattice relaxation in the rotating frame (T_1ρ) has been proposed as an MR imaging biomarker of tissue degeneration in the musculoskeletal (MSK) system ¹. Many studies have shown that T_1ρ is sensitive to compositional changes in cartilage, e.g., loss of proteoglycans (PGs), which can be potentially useful for early diagnosis of osteoarthritis (OA) ^2–4.

However, the commonly used continuous wave type T_1ρ (CW-T_1ρ) imaging on clinical MRI is subject to the magic angle effect which may affect the quantification of highly anisotropic tissues such as the articular cartilage, menisci, ligaments, and tendons ^5–10. The CW-T_1ρ values may increase more than 100% when the tissue fibers reorientate from 0º to 55º relative to the B₀ field ^6–8. These magic angle-induced CW-T_1ρ changes may be much greater than those induced by tissue degeneration, making the clinical assessment of tissue pathological changes complicated.

Recently, a new strategy using a train of adiabatic full passage (AFP) pulses for spin locking was proposed to produce T_1ρ contrast (i.e., AdiabT_1ρ) ^10–12. AdiabT_1ρ is able to detect degenerative and biomechanical changes in cartilage and tendons ^13–17. These new AdiabT_1ρ sequences have two major advantages over the conventional CW-T_1ρ sequences. The first advantage is that the measured T_1ρ from the AdiabT_1ρ sequence is much less sensitive to the magic angle effect than that measured from the CW-T_1ρ sequence at 500 Hz on a clinical MRI scanner, as shown in both cartilage and tendon studies ^8,9. The second advantage is that the AdiabT_1ρ imaging is insensitive to B₁ inhomogeneity because of the use of adiabatic pulses for spin locking ¹¹. In contrast, the conventional CW-T_1ρ imaging suffers banding artifacts related to B₁ inhomogeneity ^18–20.

Most recently, we have combined a 3D ultrashort echo time sequence with the AdiabT_1ρ preparation (UTE-AdiabT_1ρ) for volumetric T_1ρ imaging ^17,21. This UTE-AdiabT_1ρ sequence allows a comprehensive assessment of all the major tissue components in the knee joint, including both short (e.g., the deep radial and calcified cartilage, menisci, ligaments, and tendons) and long T₂ (e.g., the middle and superficial cartilage, and muscles) tissues and tissue components ^21,22. However, the tissue T₁ value should be estimated for accurate T_1ρ quantification in this sequence, as can be seen in Eq. [1] in Ref. ²¹. This T₁ measurement increases the total scan time, and errors in T₁ quantification lead to inaccurate T_1ρ quantification.

In this study, inspired by the widely used 3D magnetization‐prepared angle‐modulated partitioned k‐space spoiled gradient echo snapshots (MAPSS) sequence ²³, we developed a new UTE-AdiabT_1ρ framework, i.e., a phase-modulated UTE adiabatic T_1ρ (PM-UTE-AdiabT_1ρ) sequence, for quantitative T_1ρ imaging of all major tissue components in the knee joint. With this new PM-UTE-AdiabT_1ρ sequence, T_1ρ is estimated using a simple exponential fitting model, and T₁ information is no longer needed for accurate T_1ρ quantification. To investigate the accuracy of this new PM-UTE-AdiabT1rho sequence, phantoms with different agarose concentrations were made and scanned on a 3T clinical MRI scanner. Phantoms with different agarose concentrations and MnCl₂ concentrations (i.e., with different T₁ relaxation times) were also scanned to investigate potential T₁ contamination caused by the usage of AFP pulses with relatively long durations for spin locking, a concern raised by Pang Y in a recent letter article ²⁴. Finally, the new sequence was applied to five ex vivo human knee joint specimens and five in vivo knee joints of healthy volunteers for T_1ρ measurements for the femoral cartilage, meniscus, posterior cruciate ligament (PCL), anterior cruciate ligament (ACL), patellar tendon, and muscle.

Methods

Sequence and Signal Model

Figure 1 shows the diagram of the proposed PM-UTE-AdiabT_1ρ sequence. This new sequence includes six major features: 1) a magnetization reset module to produce a consistent magnetization recovery ²³, 2) a train of AFP pulses to produce T_1ρ contrast ²¹, 3) a phase modulation scheme (i.e., RF cycling scheme using a 90° pulse pair) to mitigate the influence of T₁ dependence during the signal evolution within the data acquisition period ²³, 4) a fat saturation module between the T_1ρ preparation and acquisition blocks for fat suppression, 5) a variable flip angle (VFA) scheme to reduce signal variation along the multiple data acquisition spokes ²³, and 6) a 3D UTE sequence for data acquisition with an efficient Cones encoding trajectory and a minimal nominal echo time (TE) of 32 μs (pulse duration is not included in the nominal TE calculation) ²⁵.

Figure 1.

Diagram of the PM-UTE-AdiabT_1ρ sequence. This new sequence includes six major features: 1) a magnetization reset module (a), 2) an AdiabT_1ρ preparation using a train of AFP pulses for spin locking (a), 3) a phase modulation scheme (i.e., RF cycling scheme using a 90° pulse pair) (a), 4) a fat saturation module (a), 5) a VFA excitation scheme (a), and 6) a 3D UTE sequence for data acquisition (b). The adjacent two acquisitions (e.g., for n^th and (n+1)^th TRs) have identical parameters except for the RF cycling module (a). During the second acquisition, the phase of the second 90° pulse in the RF cycling pair alternates 180°. A rectangular pulse (i.e., 270 μs) is used for signal excitation and the minimal nominal TE (defined as the gap between the excitation pulse and data acquisition window) of this sequence is 32 μs (b). An efficient 3D Cones trajectory is utilized for k-space encoding (c).

The RF cycling strategy (i.e., the third feature) is applied to adjacent acquisitions with the same gradient encodings ²³. During the second acquisition, the phase of the second 90° pulse in the RF cycling pair alternates 180°. The two acquisitions are interleaved. The final image is reconstructed from the subtracted data between these adjacent acquisitions with opposite cycling phases. The magnetization reset, phase modulation, and VFA design features have been successfully applied to the 3D MAPSS sequence ²³.

The transverse signal after the i^th (i = 1, 2, 3, …, N_sp, N_sp is the total number of excitations or spokes in each TR) excitation in the n^th TR is expressed as follows:

$$M_{xy}(i)=a(i)M_{prep}+b(i)$$ [1]

where $a(i)=e^{-t_0/T_1}{e}_1^{i-1}[{\prod }_{j=1}^{i-1}\cos ({\alpha }_j)]e_2\text{sin}({\alpha }_i)$

$$b(i)=M_0\left(1-e^{-t_0/T_1}\right)\{\sum _{p=1}^i[e_1^{i-p}\prod _{j=p}^{i-1}\text{cos}({\alpha }_j)]\}{e}_2\text{sin}({\alpha }_i)$$

$$e_1=e^{-\tau /T_1}\text{and}{e}_2=e^{-TE/T_2^{\text{*}}}$$

${\alpha }_i$ is the flip angle (FA) of i^th excitation. $M_{prep}$ represents the longitudinal signal at the end of the RF cycling pair, and $t_0$ is the time gap between the RF cycling pair and the signal acquisition segment. $\tau$ is the time interval between adjacent excitations or spokes. $a\left(i\right)$ and $b(i)$ are terms related to tissue T₁.

In the next TR (i.e., (n+1)^th TR), the phase of the second 90° pulse in the RF cycling pair alternates by 180°. Thus, the longitudinal signal at the end of the RF cycling pair becomes ${-M}_{prep}$. The transverse signal after the i^th excitation in the (n+1)^th TR is expressed as follows:

$$M_{xy}(i)=-a(i)M_{prep}+b(i)$$ [2]

After subtraction between signals acquired from the above two adjacent TRs (i.e., Eq. [1] – Eq. [2]), the signal equation becomes:

$$M_{xy}(i)=2a(i)M_{prep}$$ [3]

The T_1ρ independent term $b(i)$ is canceled. Since $M_{prep}$ is proportional to the term of $e^{-{TLS/T}_{1\rho }}$ and $a\left(i\right)$ is independent of tissue T_1ρ, the subtracted signal $M_{xy}\left(i\right)$ is proportional to $e^{-{TLS/T}_{1\rho }}$. TSL is the spin locking time, determined by the multiplication of the total number of AFP pulses (N_AFP) and the AFP pulse duration.

As seen in Eq. [3], the signal intensities in different spokes are modulated by $a\left(i\right)$, which can produce signal variations and lead to image artifacts. To mitigate this signal variation, the VFA scheme is incorporated into the sequence, and the FA of $i$^th excitation (i.e., ${\alpha }_i$) is determined according to the following equation:

$$a(i)=a(i+1)$$ [4]

which requires

$$\tan ({\alpha }_i)=e_1\sin ({\alpha }_{i+1})\text{or}{\alpha }_i=\text{a}\tan [e_1\sin ({\alpha }_{i+1})]$$ [5]

Once the last FA (i.e., ${\alpha }_{N_{sp}}$) is determined, the FAs of the rest excitations in a TR can be easily calculated by Eq. [5]. Then the signals acquired among different spokes have identical intensities for a tissue.

To estimate T_1ρ, multiple data are acquired by the PM-UTE-AdiabT_1ρ sequence with different TSLs (i.e., different numbers of AFP pulses) and fitted with a single exponential function ²³:

$$M_{xy}=A{e}^{-TSL/T_{1\rho }}$$ [6]

where A is a constant term that is independent of T_1ρ.

Phantom Study

The PM-UTE-AdiabT_1ρ sequence was implemented on a 3T clinical MRI scanner (GE Healthcare Technologies, Milwaukee, WI) and scanned with an 8-channel transmit/receive knee coil. An AFP pulse (hyperbolic secant type I pulse) with a duration of 6.048 ms, a bandwidth of 1.643 kHz, and a maximum B₁ amplitude of 17μT was used to generate T_1ρ contrast in the PM-UTE-AdiabT_1ρ sequence ^26,27. A rectangular pulse with a duration of 270 μs was used for UTE signal excitation.

To validate the accuracy of the developed PM-UTE-AdiabT_1ρ sequence in T_1ρ measurement, four phantoms were prepared with different agarose concentrations (1%, 2%, 3%, and 4%, weight/volume (w/v)) and scanned together with the following parameters: field of view (FOV) = 16×16 mm², matrix = 192×192, slice number = 30, slice thickness = 4 mm, spoke interval ($\tau$)/TE = 4.2/0.032 ms, magnetization recovery time (T_rec) = 330 ms, TSL = 0, 12.1, 24.2, 36.3, 48.4, 72.6, and 96.9 ms, number of TRs = 182, duration of each TR = ~650 ms, total number of spokes = 11758, N_sp = 65, ${\alpha }_{N_{sp}}={60}^{°}$, bandwidth = 125 kHz, acquisition window stretch factor = 1, and total scan time = ~14 min. Reproducibility of the PM-UTE-AdiabT_1ρ sequence was investigated using these agarose phantoms by repeating the protocol two times and the MRI system was reset before each PM-UTE-AdiabT_1ρ scan.

The phase modulation strategy employed in the proposed PM-UTE-AdiabT_1ρ sequence was expected to minimize T₁ dependence during signal evolution in the data acquisition period. To investigate this issue, a PM-UTE-CW-T_1ρ sequence was also implemented with an identical sequence structure as the proposed PM-UTE-AdiabT_1ρ except the utilization of the CW-T_1ρ preparation. The CW-T_1ρ scheme in the PM-UTE-CW-T_1ρ sequence was similar to that used in the MAPSS CW-T_1ρ sequence ²³. Four series of phantoms with different agarose concentrations (1%, 2%, 3% and 4% w/v) were prepared to compare the T_1ρ quantification between PM-UTE-CW-T_1ρ and MAPSS CW-T_1ρ. In each series, four different phantoms were prepared with different MnCl₂ concentrations to change their T₁ relaxation times. Supplemental Information Table S1 shows more detailed information of these phantoms in terms of the agarose and MnCl₂ concentrations as well as their T₁ values as measured by a VFA technique with B₁ correction ²⁸. These phantoms were then scanned with 3D PM-UTE-CW-T_1ρ and MAPSS CW-T_1ρ sequences for comparison with following parameters: 1) PM-UTE-CW-T_1ρ: FOV = 14×14 mm², matrix = 192×192, slice number = 36, slice thickness = 4 mm, $\tau$/TE = 4.2/0.032 ms, T_rec = 800 ms, spin lock frequency = 500 Hz, TSL = 0, 10, 30, and 60 ms, N_sp = 64, ${\alpha }_{N_{sp}}={60}^{°}$, bandwidth = 125 kHz, and total scan time = ~17 min; 2) MAPSS CW-T_1ρ: FOV = 14×14 mm², matrix = 192×192, slice number = 36, slice thickness = 4 mm, $\tau$/TE = 4.9/2.2 ms, T_rec = 800 ms, spin lock frequency = 500 Hz, TSL = 0, 10, 30, and 60 ms, N_sp = 64, ${\alpha }_{N_{sp}}={90}^{°}$, bandwidth = 125 kHz, and total scan time = ~11 min.

Furthermore, to investigate the relationship of the T_1ρ measurements between AdiabT_1ρ and CW-T_1ρ, three series of phantoms with different agarose concentrations (1%, 2%, and 3% w/v) were prepared and scanned with both 3D PM-UTE-AdiabT_1ρ and MAPSS CW-T_1ρ sequences. In each series, four different phantoms were prepared with different MnCl₂ concentrations to change their T₁ relaxation times. Detailed information of these phantoms can be seen in Table 1. The sequence parameters of the PM-UTE-AdiabT_1ρ were identical to those used in the first phantom study above except for the following parameters: FOV = 14×14 mm², slice number = 36 and slice thickness = 4 mm. The MAPSS CW-T_1ρ sequence was scanned with identical sequence parameters used in the second phantom study.

Table 1.

Agarose and MnCl₂ concentrations as well as quantitative T₁ values for twelve phantoms (see Figure 3a).

Phantom ID	#1	#2	#3	#4	#5	#6	#7	#8	#9	#10	#11	#12
Agarose concentration (w/v)	1%	1%	1%	1%	2%	2%	2%	2%	3%	3%	3%	3%
MnCl2 concentration (mM)	0.019	0.039	0.071	0.136	0.017	0.034	0.062	0.118	0.015	0.030	0.056	0.106
T1 (ms)	1503±7.7	1334±7.3	882±3.1	596±1.3	1507±6.2	1196±3.9	890±1.9	595±0.8	1494±6.0	1205±3.9	913±2.6	606±1.3

Ex Vivo Knee Study

Five normal cadaveric knee joints from five donors (aged 38–87 years, mean age 55.6 years; 2 male, 3 females) were scanned by the PM-UTE-AdiabT_1ρ sequence at room temperature using the following parameters: FOV = 15×15 mm², matrix = 256×256, slice number = 40, slice thickness = 2 mm, $\tau$/TE = 6/0.032 ms, T_rec = 330 ms, TSL = 0, 12.1, 24.2, 36.3, 48.4, 72.6, and 96.9 ms, number of TRs = 316, duration of each TR = ~700 ms, total number of spokes = 20530, N_sp = 65, ${\alpha }_{N_{sp}}={60}^{°}$, bandwidth = 125 kHz, acquisition window stretch factor = 1, and total scan time = ~25 min.

In Vivo Knee Study

Five normal knee joints from five volunteers (aged 34-40 years, mean age 37.5 years; 4 males, 1 female) were also scanned. Informed consent was obtained from all subjects in accordance with the guidelines of the institutional review board. The PM-UTE-AdiabT_1ρ sequence was used to scan these knee joints with the following parameters: FOV = 15×15 mm², matrix = 256×256, slice number = 32, slice thickness = 3 mm, $\tau$/TE = 5.5/0.032 ms, T_rec = 330 ms, TSL = 0, 12.1, 24.2, 36.3, 48.4, 72.6 and 96.9 ms, number of TRs = 150, duration of each TR = ~900 ms, total number of spokes = 11186, N_sp = 75, ${\alpha }_{N_{sp}}={60}^{°}$, bandwidth = 125 kHz, acquisition window stretch factor = 1.5, and total scan time = ~16 min.

Data Analysis

Single exponential fitting was performed on the magnitude data, and the image background noise was estimated and incorporated as a constant term in the fitting process. The mean value of the background noise was calculated from a carefully selected 8×8 object-free region (avoiding regions with artifacts) on the PM-UTE-AdiabT_1ρ image with the longest TSL. The data processing code was written in Matlab (The MathWorks Inc., Natick, MA, USA), and the Levenberg-Marquardt algorithm was used to solve the non-linear least-squares optimization problem ^29,30. The Intraclass Correlation Coefficient (ICC) analysis was used to evaluate the reproducibility of T_1ρ measurement for the phantom study. Pearson correlation was used to investigate the correlations between PM-UTE-AdiabT_1ρ, MAPSS CW-T_1ρ and T₁, and between PM-UTE-CW-T_1ρ and MAPSS CW-T_1ρ as well as between PM-UTE-AdiabT_1ρ and MAPSS CW-T_1ρ. A P-value less than 0.05 is considered statistically significant. Regions of interest (ROIs) were manually drawn for various tissues, including the articular cartilage, meniscus, PCL, ACL, patellar tendon, and muscle in all knee joints. Mean T_1ρ and standard deviation (STD) values for all the major tissue components in both ex vivo and in vivo knee joints were calculated.

Results

Figure 2 shows the results from the first phantom study. As can be seen in the T_1ρ map in Figure 2b, the T_1ρ values decrease with the higher agarose concentrations in the phantoms. Excellent linear correlation is achieved between R_1ρ (=1/T_1ρ) values and agarose concentrations with an R² of 0.9999. The T_1ρ map of the second scan (Figure 2d) shows similar T_1ρ distributions to that of the first scan (Figure 2b). Two repeated scans show an excellent ICC (R² = 0.9996), demonstrating great reproducibility of the PM-UTE-AdiabT_1ρ technique.

Figure 2.

Phantom imaging results for the studies of tissue compositional sensitivity and scan reproducibility using the PM-UTE-AdiabT_1ρ sequence. Phantoms with four different agarose concentrations (i.e., 1, 2, 3, and 4%, w/v) (a) are imaged twice and the corresponding T_1ρ maps are seen in (b) and (d). The R_1ρ values show excellent linear correlation with agarose concentration with an R² of 0.9999 (c). Excellent reproducibility of the R_1ρ measurement is achieved between the 1^st and 2^nd scans with an R² of 0.9996 (e).

Supplemental Information Figure S1 shows the correlation curve between PM-UTE-CW-T_1ρ MAPSS CW-T_1ρ measurements for the twelve phantoms. T₁ values of these phantoms ranged from ~600 to ~1700 ms (see Supplemental Information Table S1), which cover the spectrum of T₁ values for all major tissues in the knee joint ²⁸. A very strong correlation was found between PM-UTE-AdiabT_1ρ and MAPSS CW-T_1ρ measures (R² = 0.9971, P < 0.0001). The obtained linear regression equation demonstrated a similar range of T_1ρ measurements (i.e., [20, 100] ms) between these two techniques.

Figure 3 shows the T₁, PM-UTE-AdiabT_1ρ, and MAPSS CW-T_1ρ mapping results from the second phantom study. This T₁ map displays a wide range of T₁ relaxation times from ~600 to ~1500 ms (T₁ values are shown in Table 1). The T_1ρ values in the PM-UTE-AdiabT_1ρ maps are generally higher than the corresponding MAPSS CW-T_1ρ measurements. Notably, the T_1ρ values of these phantoms demonstrate a similar visual trend for the two kinds of T_1ρ measurement techniques. Figure 4 shows the correlation curves between MAPSS CW-T_1ρ, PM-UTE-AdiabT_1ρ and T₁, and between PM-UTE-AdiabT_1ρ and MAPSS CW-T_1ρ for the twelve phantoms. A similar moderate correlation was found between MAPSS CW-T_1ρ measures and T₁ values (R² = 0.3846, P = 0.031) and between PM-UTE-AdiabT_1ρ measures and T₁ values (R² = 0.3633, P = 0.038), while a very strong correlation was found between PM-UTE-AdiabT_1ρ and MAPSS CW-T_1ρ measures (R² = 0.9985, P < 0.0001).

Figure 3.

Phantom imaging results using T₁, PM-UTE-AdiabT_1ρ, and MAPSS CW-T_1ρ mapping techniques. The agarose concentrations are 1%, 2%, and 3% w/v for the twelve phantoms listed in the first (1-4), second (5-8), and third (9-12) rows, respectively (a). The quantitative T₁, PM-UTE-AdiabT_1ρ, and MAPSS CW-T_1ρ maps are shown in panels (b), (c), and (d), respectively.

Figure 4.

Correlation curves for the measurements between MAPSS CW-T_1ρ, PM-UTE-AdiabT_1ρ , and T₁, and between PM-UTE-AdiabT_1ρ and MAPSS CW-T_1ρ for the twelve phantoms. A similar moderate correlation is found between MAPSS CW-T_1ρ measures and T₁ values (R² = 0.3846, P = 0.031) (a) and between PM-UTE-AdiabT_1ρ measures and T₁ values (R² = 0.3633, P = 0.038) (b), while a very strong correlation is found between PM-UTE-AdiabT_1ρ and MAPSS CW-T_1ρ measures (R² = 0.9985, P < 0.0001) (c).

Figure 5 shows representative PM-UTE-AdiabT_1ρ images acquired from a knee joint specimen (45-year-old male donor). Both long T₂ (e.g., the cartilage and muscle) and short T₂ tissue (e.g., the meniscus, quadriceps tendon, patellar tendon, ACL and PCL) signals are detected by the PM-UTE-AdiabT_1ρ sequence. Fat signals are efficiently suppressed in all images at different TSLs. As can be seen in Figure 6, excellent single exponential fitting is achieved for the major knee tissue components including the femoral cartilage, meniscus, PCL, ACL, patellar tendon, and muscle, where the T_1ρ values of these tissues are 90.3±10.1, 37.7±1.6, 46.8±1.1, 76.7±9.5, 13.4±0.6, and 74.8±5.8 ms, respectively.

Figure 5.

Representative PM-UTE-AdiabT_1ρ images of two different slices (a and b) from a normal knee specimen (45-year-old male donor) with TSLs ranging from 0 to 96.87 ms. Both short and long T₂ tissue signals are detected by this new sequence. Short T₂ tissue signals decrease faster with increased TSLs. Fat is efficiently suppressed in all images.

Figure 6.

The PM-UTE-AdiabT_1ρ imaging results of the normal knee specimen from the 45-year-old male donor. The fitting curves and corresponding T_1ρ values of the femoral cartilage (a), meniscus (b), PCL (c), ACL (g), patellar tendon (h), and muscle (i) (tissue ROIs shown inside of the red closed curves) are 90.3±10.1 ms (d), 37.7±1.6 ms (e), 46.8±1.1 ms (f), 76.7±9.5 ms (j), 13.4±0.6 (k), and 74.8±5.8 ms (l), respectively.

Figure 7 shows representative PM-UTE-AdiabT_1ρ images acquired from an in vivo knee joint (36-year-old female volunteer). Similar to the ex vivo knee study, both short and long T₂ tissue signals in the knee joint are well detected by this new sequence, while fat signals are efficiently suppressed. As can be seen in Figure 8, excellent single exponential fitting is achieved for all the major knee tissue components. The T_1ρvalues of the femoral cartilage, meniscus, PCL, ACL, patellar tendon, and muscle are 73.2±2.6, 34.6±1.2, 30.0±0.8, 47.6±2.0, 15.6±0.3, and 61.2±3.2 ms, respectively.

Figure 7.

Representative PM-UTE-AdiabT_1ρ images of two different slices (a and b) from a normal knee joint (36-year-old female volunteer). Both short and long T₂ tissue signals are detected by this new sequence. Signals for short T₂ tissues (e.g., the meniscus and patellar tendon) decrease faster with longer TSLs. Fat is efficiently suppressed in all images.

Figure 8.

The PM-UTE-AdiabT_1ρ imaging results of the normal knee joint from the 36-year-old female volunteer. The fitting curves and corresponding T_1ρ values of the femoral cartilage (a), meniscus (b), PCL (c), ACL (g), patellar tendon (h), and muscle (i) (tissue ROIs shown inside of the red closed curves) are 73.2±2.6 ms (d), 34.6±1.2 ms (e), 30.0±0.8 ms (f), 47.6±2.0 ms (j), 15.6±0.3 ms (k), and 61.2±3.2 ms (l), respectively.

Table 2 summarizes the T_1ρ measurements for both ex vivo and in vivo knee joints. The average T_1ρ values of femoral cartilage, meniscus, PCL, ACL, patellar tendon, and muscle for the five normal knee joint specimens are 105.6±8.4, 39.2±5.1, 51.6±5.3, 79.0±9.3, 19.8±4.5, and 91.1±8.8 ms, respectively. The average T_1ρ values of the femoral cartilage, meniscus, PCL, ACL, patellar tendon, and muscle for the five normal in vivo knee joints are 77.9±3.9, 30.1±2.2, 29.2±2.4, 52.0±3.1, 17.0±1.8, and 57.6±2.8 ms, respectively. The ex vivo T_1ρ measures are generally higher than the in vivo measurements because of the lower tissue temperature (room temperature vs. body temperature) and potential degeneration of knee joint specimens.

Table 2.

T_1ρ and its fitting standard errors of cartilage, meniscus, PCL, ACL, patellar tendon and muscle in five ex vivo and five in vivo normal knee joints as well as the summarized mean and STD values for each tissue.

T1ρ (ms)		Cartilage	Meniscus	PCL	ACL	Patellar tendon	Muscle
Ex vivo	#1	90.3±10.1	37.7±1.6	46.8±1.1	76.7±9.5	13.4±0.6	74.8±5.8
	#2	112.7±11.3	46.1±0.3	59.4±0.8	87.4±5.3	21.5±2.3	95.3±1.1
	#3	105.4±8.4	31.8±0.8	44.5±0.4	92.0±4.6	17.2±0.9	97.3±1.2
	#4	113.9±8.8	36.7±1.5	52.9±1.6	71.8±1.8	26.9±0.5	89.1±4.2
	#5	105.5±11.6	43.5±2.0	54.2±2.4	67.2±7.9	19.9±2.5	99.2±12.7
	mean±STD	105.6±8.4	39.2±5.1	51.6±5.3	79.0±9.3	19.8±4.5	91.1±8.8
In vivo	#1	73.2±2.6	34.6±1.2	30.0±0.8	47.6±2.0	15.6±0.3	61.2±3.2
	#2	76.2±1.9	30.7±0.9	33.1±1.0	56.5±3.9	16.4±0.5	58.7±1.7
	#3	76.3±2.6	29.5±0.4	28.8±0.4	54.1±2.4	17.1±0.5	52.8±0.7
	#4	79.0±3.6	28.3±0.7	25.6±1.7	50.0±3.8	15.7±0.3	57.2±2.9
	#5	84.8±2.9	31.8±1.0	28.3±0.3	52.0±1.4	20.4±0.8	58.3±1.8
	mean±STD	77.9±3.9	30.1±2.2	29.2±2.4	52.0±3.1	17.0±1.8	57.6±2.8

Discussion

We have demonstrated that the PM-UTE-AdiabT_1ρ sequence allows volumetric T_1ρ mapping of both short and long T₂ tissues in the whole knee joint on a clinical 3T scanner. Phantom studies suggest that T_1ρ measures provided by the PM-UTE-AdiabT_1ρ technique are sensitive to compositional changes and these measurements are reproducible. The ex vivo and in vivo knee studies demonstrate the feasibility of quantifying T_1ρ for all the major tissue components in the knee joint including the articular cartilage, meniscus, PCL, ACL, patellar tendon, and muscle.

The 3D MAPSS CW-T_1ρ sequence has been widely used to evaluate tissue degeneration and monitor compositional changes longitudinally in the musculoskeletal system ^31–36. Some of the sequence features in 3D MAPSS, such as magnetization reset, phase modulation, and VFA scheme ²³, have been successfully incorporated in the PM-UTE-AdiabT_1ρ sequence. The successful implementation of these features was validated by the second phantom study, which demonstrated a strong linear correlation (R² = 0.9971) and a similar value range (i.e., [20, 100] ms) between PM-UTE-CW-T_1ρ and MAPSS CW-T_1ρ measurements. With these features implemented in PM-UTE-AdiabT_1ρ, T₁ measurement is no longer needed, and a simple exponential fitting is sufficient for accurate UTE-AdiabT_1ρ quantification. In addition, the first phantom study further demonstrates the excellent reproducibility of this framework used in the PM-UTE-AdiabT_1ρ sequence.

We have demonstrated in this study that the PM-UTE-AdiabT_1ρ sequence is able to quantify all the major tissue components in the knee joint, including both short (e.g., the deep cartilage, menisci, ligaments, and tendons) and long T₂ (e.g., the superficial cartilage, and muscle) tissues. This is because the PM-UTE-AdiabT_1ρ sequence uses a very short TE (e.g., 0.032 ms) for data acquisition ²⁵. Thus the collagen-rich short T₂ tissues can also be imaged and quantified. This allows more comprehensive assessment of musculoskeletal diseases, such as knee OA. Knee OA has been recognized as a whole organ disease ³⁷. Any tissue deterioration or misalignment in the knee joint can accelerate the progression of OA ³⁸. Studies also show that meniscal positioning or collateral ligament damage can lead to cartilage loss ^39,40. As such, the developed PM-UTE-AdiabT_1ρ sequence provides great potential in studying knee OA comprehensively.

A 3T cartilage study has shown that the 3D MAPSS-T_1ρ sequence with a CW type spin-lock scheme is sensitive to the magic angle effect ⁶. The CW-T_1ρ values are more than doubled when the tissue fibers reorientate from 0° to 55° relative to the B₀ field ^6–8. Most recently, both ex vivo enzymatic degradation studies and in vivo OA studies have shown that AdiabT_1ρ is sensitive to biochemical changes in articular cartilage ^13,15,17. More importantly, AdiabT_1ρ is much less sensitive to the magic angle effect than CW-T_1ρ as demonstrated in previous cartilage and tendon studies on a clinical 3T MRI ^8,9.

In addition to the reduced sensitivity to the magic angle effect, the AdiabT_1ρ preparation typically has a lower total RF power deposition than the CW-T_1ρ preparation. This is because the AFP pulse employs both amplitude and frequency modulation rather than amplitude only, which is used in CW type preparation for spin locking ¹¹. The specific absorption rate (SAR) is directly related to the amplitude of the spin locking pulse ⁴¹. Moreover, the PM-UTE-AdiabT_1ρ imaging is insensitive to B₁ inhomogeneity because of the adiabatic pulse character ¹¹. In contrast, the conventional CW-T_1ρ imaging suffers from B₁ inhomogeneity related banding artifacts ^18–20. These advantages of AdiabT_1ρ preparation will be even more favorable in high field MRI (e.g., 7T) where the SAR issue and B₁ inhomogeneity become more significant.

The contrast mechanism of AdiabT_1ρ is still not very well-understood. The spin-lock direction of AdiabT_1ρ varies from the z-axis to the x-y plane and then back to the z-axis. In comparison, the spins are only locked in the x-y plane during CW-T_1ρ preparation. Dipole-dipole interaction may affect apparent T_1ρ more in the x-y plane than in the z-axis, as demonstrated in the orientation sensitivity of T₂ and T₁ relaxation times, respectively ^42,43. This could lead to a shorter T_1ρ measured by CW-T_1ρ in comparison to AdiabT_1ρ when the tissue fiber orientation is parallel to the B₀ field, which means that CW-T_1ρ is more sensitive to the magic angle effect ⁴⁴. If the spin-lock frequency of CW-T_1ρ is sufficiently high (e.g., ≥1 kHz), the effect of dipole-dipole interaction for T_1ρ measurement may be negligible compared with the slow-motion interactions between motion-restricted water molecules and their local macromolecular environment ¹⁰. Due to the relatively higher spin-lock frequency of the AdiabT_1ρ preparation (from 724 to 860 Hz in this study, calculated by the square root of the real and imaginary parts of the AFP pulses), the measured T_1ρ values from the AdiabT_1ρ type sequences are generally higher than those measured from the typical CW-T_1ρ type sequences (≤500 Hz) on a clinical scanner ^10,13,23. Our measured in vivo cartilage T_1ρ value (77.9±3.9 ms) is similar to the reported values (from 69.0 to 79.9 ms) from a recent AdiabT_1ρ study ¹³.

A recent letter article from Pang Y explained how a T_1ρ relaxation time measured using AdiabT_1ρ preparation was relatively longer than when using a CW-T_1ρ preparation ²⁴. The article claimed that the AdiabT_1ρ quantification was subject to T₁ contamination during each relatively long AFP pulse (even for a 2 ms pulse), but there was no validated theory or dedicated experimental data provided in the letter article to support the T₁ contamination assumption. Though our phantom study showed a moderate correlation between PM-UTE-AdiabT_1ρ and T₁ measurements (R² = 0.3633, P = 0.038), this does not necessarily mean there is T₁ contamination in the AdiabT_1ρ measurement given that a similar moderate correlation was also found between MAPSS CW-T_1ρ and T₁ measurements (R² = 0.3846, P = 0.031), and the MAPSS CW-T_1ρ measurement is T₁-independent in theory. More interestingly, the T_1ρ values measured by the proposed PM-UTE-AdiabT_1ρ linearly correlated with the MAPSS CW-T_1ρ measurements (R² = 0.9985, P < 0.0001). This strong correlation demonstrates that the PM-UTE-AdiabT_1ρ measurements may be also unaffected by T₁ contamination, considering that the MAPSS CW-T_1ρ measurement is theoretically T₁-independent. Moreover, the strong linear correlation also demonstrates the comparable sensitivity between PM-UTE-AdiabT_1ρ and MAPSS CW-T_1ρ techniques in detecting compositional changes. We believe that one of the major contributing factors for the higher AdiabT_1ρ values is the higher spin-lock frequencies used in the AdiabT_1ρ preparation compared to the CW-T_1ρ preparation on a clinical scanner, which was not mentioned in Pang Y’s letter article. Indeed, more research is needed to better understand the mechanism of AdiabT_1ρ relaxation.

To further investigate the contrast mechanism of both AdiabT_1ρ and CW-T_1ρ, we plan to perform a comparison study (i.e., PM-UTE-AdiabT_1ρ vs. PM-UTE-CWT_1ρ) on a newly installed small-bore Bruker 3T scanner. In addition to the state-of-art preparation scheme for CW-T_1ρ, a few new CW-T_1ρ developments will also be incorporated with the PM-UTE sequence structure in this comparison study ⁴⁵. Typically, the maximum spin-lock frequency of CW-T_1ρ is limited by the peak power of the RF amplifier and SAR restriction on a clinical 3T scanner. The Bruker scanner offers more flexibility in adjusting the spin-lock frequencies than the clinical 3T scanner used in this study, thus providing more useful information in studying the mechanisms of CW-T_1ρ and AdiabT_1ρ. Magic angle effects will also be investigated on different spin-lock frequencies as well.

This study has several limitations. First, the total number of ex vivo and in vivo knee joints is relatively small, which is justified for this technical evaluation study. Second, we have only demonstrated the technical feasibility of the 3D PM-UTE-AdiabT_1ρ sequence for volumetric T_1ρ mapping of both short and long T₂ tissues in normal knee joints of healthy volunteers. Patients with knee OA are not recruited for this study. The ex vivo knee joints also have relatively normal morphology. Histology was not performed to confirm the degeneration status in each knee joint specimen. Third, although the new sequence eliminates the need for T₁ measurement, it comes at the cost of prolonged scan time since two acquisitions with the same gradient encodings must be scanned. The scan time can be reduced with a smaller number of TSLs, longer spiral interleaves for data acquisition, and more spokes (i.e., a higher N_sp value) in each TR ^46,47. Moreover, fast imaging techniques such as compressed sensing or deep learning can be incorporated to further accelerate data acquisition ^48–51. Fourth, the image signal-to-noise ratio increases when a longer T_rec is used, leading to a longer total scan time. Longer T_recs will be investigated when fast imaging techniques are available to accelerate the PM-UTE-AdiabT_1ρ scan.

Conclusion

The newly proposed PM-UTE-AdiabT_1ρ sequence allows comprehensive quantification of all the major tissue components in the knee joint, demonstrating its potential in future clinical studies of knee OA.

Supplementary Material

Supplementary Material

Supplemental Information Table S1. Agarose and MnCl₂ concentrations as well as quantitative T₁ values for sixteen phantoms (see Supplemental Information Figure S1a).

Supplemental Information Figure S1. Correlation curve of the measurements between PM-UTE-CW-T_1ρ and MAPSS CW-T_1ρ for the sixteen phantoms. The agarose concentrations are 1%, 2%, 3% and 4% w/v for the sixteen phantoms listed in the first (1–4), second (5–8), third (9–12) and fourth (13–16) rows, respectively (a). The T₁ values of all these phantoms can be found in Supplemental Information Table S1. A very strong correlation is found between PM-UTE-CW-T_1ρ and MAPSS CW-T_1ρ measures (R² = 0.9971, P < 0.0001) (b). The obtained linear regression equation also demonstrates a similar range of T_1ρ measurements (i.e., [20, 100] ms) between these two techniques.