Biexponential T_1ρ Relaxation Mapping of Human Knee Menisci

Abstract

Background:

Measuring T_1ρ in the knee menisci can potentially be used as noninvasive biomarkers in detecting early-stage osteoarthritis (OA).

Purpose:

To demonstrate the feasibility of biexponential T_1ρ relaxation mapping of human knee menisci.

Study Type:

Prospective.

Population:

Eight healthy volunteers with no known inflammation, trauma, or pain in the knee and three symptomatic subjects with early knee OA.

Field Strength/Sequence:

Customized Turbo-FLASH sequence to acquire 3D-T_1ρ-weighted images on a 3 T MRI scanner.

Assessment:

T_1ρ relaxation values were assessed in 11 meniscal regions of interest (ROIs) using monoexponential and biexponential models.

Statistical Tests:

Nonparametric rank-sum tests, Kruskal–Wallis test, and coefficient of variation.

Results:

The mean monoexponential T_1ρ relaxation in the lateral menisci were 28.05 ± 4.2 msec and 37.06 ± 10.64 msec for healthy subjects and early knee OA patients, respectively, while the short and long components were 8.07 ± 0.5 msec and 72.35 ± 3.2 msec for healthy subjects and 2.63 ± 2.99 msec and 55.27 ± 24.76 msec for early knee OA patients, respectively. The mean monoexponential T_1ρ relaxation in the medial menisci were 34.30 ± 3.8 msec and 37.26 ± 11.38 msec for healthy and OA patients, respectively, while the short and long components were 7.76 ± 0.7 msec and 72.19± 4.2 msec for healthy subjects and 3.06 ± 3.24 msec and 55.27 ± 24.59 msec for OA patients, respectively. Statistically significant (P≤ 0.05) differences were observed in the monoexponential relaxation between some of the ROIs. The T_1ρ,short was significantly lower (P = 0.02) in the patients than controls. The rmsCV% ranges were 1.51–16.6%, 3.59–14.3%, and 4.91–15.6% for T_1ρ-mono, T_1ρ-short, and T_1ρ-long, respectively.

Data Conclusion:

Our results showed that in all ROIs, T_1ρ relaxation times of outer zones (red zones) were less than inner zones (white zones). Monoexponential T_1ρ was increased in medial, lateral, and body menisci of early OA while the biexponential numbers were decreased in early OA patients.

Level of Evidence:

2

Technical Efficacy Stage:

1

OSTEOARTHRITIS (OA) of the knee is a degenerative joint disease that affects about 10% of adults aged over 60 years.^1,2 Knee menisci are vital for the normal function and long-term health of the knee joint. The menisci help to increase joint stability, shock absorption, load distribution on knee joint, and decreasing friction between articular surfaces; although the relationship between meniscal degeneration and OA is not clear it may be effective in the initiation of knee OA.^3–7 The normal knee meniscus extracellular matrix is composed of 72% water, 22% collagen, and 0.8% proteoglycans and this composition may be altered due to OA.^8,9 Magnetic resonance imaging (MRI) has many advantages, including its ability to achieve high soft-tissue contrast among different tissues in the knee joint; however, clinical MRI protocols for knee imaging such as T₁-weighted spoiled gradient echo (SPGR) sequences and fat-saturated T₂ or proton density-weighted fast spin echo (FSE) are unable to quantitatively assess changes in tissue composition in the early stages of OA. Therefore, T_1ρ relaxation measurement, which is sensitive to changes in the composition of the extracellular matrix (ECM) of the meniscus, might be valuable for identifying individuals in need of early OA treatment and for detecting response to various therapies.^10,11 Estimating T_1ρ and T₂ relaxation times (which are spin-lattice relaxation times in the rotating frame and spin-spin relaxation times) have been studied extensively as potential noninvasive biomarkers in detecting early-stage OA.^12–16 T_1ρ is sensitive to the slow-motional interactions between local macromolecular environments and bulk water¹⁷ and as a result to proteoglycan content in the tissue,¹⁸ while T₂ is found to be sensitive to the collagen fibril orientation and anisotropy.^19,20 Furthermore, a recent meta-analysis²¹ study validated the reliability (compiling 58 studies) and discriminative ability (compiling 26 studies) of T_1ρ and T₂ compositional mapping of knee cartilage, especially in mild knee OA as a motivation. T_1ρ showed better discrimination for mild OA and OA compared to T₂. The increase in water content and the loss of proteoglycans (PGs) in early-stage OA results in an elevation of the T_1ρ relaxation times.^14,22,23 A multicomponent model better represents the relaxation decay behavior than a monoexponential model because the monoexponential relaxation shows the weighted mean of relaxation in different water compartments in cartilage, while a multicomponent model distinguishes the relaxation time of different water compartments (eg, free water, water tightly bound to PG and collagen macromolecules, etc.).²⁴

In this study we sought to demonstrate a method for measuring biexponential T_1ρ relaxation times of knee menisci using MRI of healthy control subjects and early knee OA patients.

Materials and Methods

Study Group

The study was approved by the Institutional Review Board (IRB) and complied with Health Insurance Portability and Accountability Act guidelines. All the participants provided written informed consent prior to participating in this study. Eight healthy volunteers (women: n = 4, men: n = 4) who had no known inflammation, trauma, or prior knee surgeries with the mean age (±SD) of 30 ± 4 years (women: 26 ± 3 years, men: 30 ± 3 years), mean height of 169 ± 12 cm (women: 159 ± 6 cm, men: 177 ± 3 cm), mean weight of 52 ± 6 kg (women: 52 ± 6 kg, men: 74 ± 12 kg), and mean BMI (SD) of 20.9 ± 3.5 for women and 22.9 ± 2.8 for men and three symptomatic subjects (two females and a male) with early knee OA (KL-1, 2) with a mean age (±SD) of 51 ± 8 years, mean height and mean weight of 170 ± 9 and 79 ± 24, respectively, were enrolled in this study.

Ex Vivo Bovine Menisci Study

Fresh bovine meniscus specimens (n = 2, age = ~3 months old) were obtained from a slaughterhouse (Max Insel Cohen, Livingston, NJ) within 36 hours postmortem. The bovine meniscus specimens were equilibrated in phosphate-buffered saline (PBS) for 1 hour before the MRI study. 3D T_1ρ-weighted images were obtained using a Cartesian Turbo-FLASH sequence that was customized to enable T_1ρ imaging with varying spin-lock durations.²⁴ The T_1ρ preparation module consists of a 90° RF excitation pulse along the x-direction, followed by a spin-lock pulse along the y-direction and a −90° along the x-axis. To compensate for B₀ inhomogeneity, an 180° refocusing pulse was applied in the middle of the module. Moreover, the phase of the spin-lock module was altered to compensate the B₁ inhomogeneity effect.²⁴ Series of T_1ρ-weighted images were acquired at 15 different spin-lock lengths (TSLs), including 0.5, 1, 2, 4, 6, 8, 10, 12, 15, 25, 35, 40, 45, 55, and 65 msec. MRI acquisition parameters were as follows: repetition time (TR) = 1500 msec, echo time (TE) = 4 msec, bandwidth = 515 Hz/pixel, matrix size = 256 × 128 × 64; slice thickness = 2 mm, flip angle (FA) = 8°, field of view (FOV) = 120 × 120 mm, frequency of spin-lock = 500 Hz. The frequency of spin-lock was set to 500 Hz as a good trade-off between observing the biexponential behavior and reducing the magic angle effect.²⁴ The experiment was repeated with different acceleration factors (AFs) = 1–4 to examine the effect of reduced signal-to-noise ratio (SNR) due to parallel imaging on the estimated relaxation times. The maps were also estimated for 6 (5, 15, 25, 35, 45, 55) and 10 TSLs (2, 4, 6, 8, 10, 15, 25, 35, 45, 55) to compare the effect of number of time points on estimation error.

In Vivo Knee Menisci Study

All subjects arrived at the MRI center 30 minutes prior to data acquisition and the knee joint of the leg was scanned with a 3 T clinical MRI scanner (Magnetom Prisma, Siemens Healthcare, Erlangen, Germany) in the feet-first supine position with a 15-channel T_X/R_X knee coil (Quality Electrodynamics, Cleveland, OH). Sagittal 3D T_1ρ-weighted scans of the knee were obtained at 10 different spin-lock durations (TSLs) including TSL: 2, 4, 6, 8, 10, 15, 25, 35, 45, and 55 msec. Sagittal 3D T_1ρ-weighted scans of the knee were obtained at 10 different spin-lock durations (TSLs) including TSL: 2, 4, 6, 8, 10, 15, 25, 35, 45, and 55 msec, with the same imaging parameters as the ex vivo study. A generalized auto-calibrating partially parallel acquisition (GRAPPA) acceleration factor of 3 was used to achieve a total scan time of 15 minutes for in vivo study.

Multicomponent T_1ρ Relaxation Analysis

The mono- and biexponential T_1ρ relaxations were reconstructed pixel-by-pixel across the regions of interest (ROIs) in MatLab (v. R2016a; MathWorks, Natick, MA). The monoexponential relaxation time, T_1ρmono, was calculated by fitting the monoexponential model to the data:

$$y\left(t_{sl}\right)=y_m{e}^{-\frac{t_{sl}}{T_1\rho mono}}+y_0$$ (1)

Where y(t_sl) is the pixel signal intensity in the scan acquired at the spin-lock duration, TSL. y₀ is the constant term that was set to the noise standard deviation.

Similarly, the biexponential relaxation components were estimated by fitting the biexponential model to the same data set using a nonlinear approach:

$$y\left(t_{sl}\right)=y_l{e}^{-\frac{t_{sl}}{T_1{\rho }_{long}}}+y_s{e}^{-\frac{t_{sl}}{T_1{\rho }_{short}}}+y_0$$ (2)

Where T_1ρlong and T_1ρshort are the long and short relaxation time components, respectively. The relative fractions of the long (y_l) and the short (y_s) components are reported as $f_{long}=\frac{y_l}{y_l+y_s}$ and $f_{short}=\frac{y_s}{y_l+y_s}$, respectively.

Afterward, the pixels satisfying the below condition were included in the mean estimation²⁵:

$$T1{\rho }_{long}>4\times T1{\rho }_{short}$$ (3)

Meniscus Segmentation

T_1ρ of the menisci of all 11 subjects (eight healthy and three knee OA patients) in the study group were estimated in nine different ROIs including meniscus body (MB), posterior-medial menisci (PM), anterior-medial menisci (AM), posterior-lateral menisci (PL), anterior-lateral menisci (AL), and the segmented menisci regions were then divided into two subcompartments that are red zone (RZ) and white zone (WZ) (Fig. 1). Meniscus segmentation was performed using the shortest TSL = 2 (therefore, with highest SNR) of T_1ρ-weighted images. First, the ROI masks were drawn manually using ITK-SNAP (v. 3.6.0)²⁶ then the segmentation masks transferred to MatLab for T_1ρ measurements.

FIGURE 1:

Representative sagittal T_1ρ-weighted (TSL = 2 msec) illustrating the ROIs of meniscus subregions in lateral, medial, and body menisci.

Mont Carlo Simulation

Monte Carlo simulations²⁷ were performed to determine the effects of different parameters such as number of TSLs and SNR on estimating four parameters in a biexponential model (two relaxation times and the corresponding fractions). A biexponential decay curve with known T_1ρ component fractions and time constants were generated using the array of TSL times. A random Rician noise was then added to the signal and the relaxation components were estimated by fitting the biexponential model Eq. [2] to the noisy signals. The process was repeated for N = 1000 independent noise trails, and the mean absolute percentage error (MAPE) was calculated as:

$$MAPE=\frac{100}{N}\sum \frac{\left|T_a-T_e\right|}{T_a}$$ (4)

Where T_a and T_e are the actual and estimated relaxation values, respectively.

Statistical Analysis

We performed a Kruskal–Wallis test to compare the relaxation components between different ROIs and zones and differences were deemed to be significant at P < 0.05. The histogram of calculated pixel T_1ρ mono, T_1ρ long, and T_1ρ short was characterized by calculating the mean, standard deviation, median, mode, skewness, and kurtosis of the distribution. The mean absolute percentage error (MAPE) used to find differences between the estimation using 15 TSLs (as reference) with the estimation using 10 and 6 TSLs. All statistical testing was performed with Minitab (Minitab 18 Statistical Software, Minitab) and MatLab.

As the measure of repeatability of T_1ρ relaxation components (mono-short-long) in different ROIs, the experiment was repeated on two volunteers on two different days with a 2-week gap. For each subject, the mean and SD were calculated across the two sessions, and the coefficient of variation (CV), representing the relative magnitude (%) was calculated as CV = SD/mean. To depict the repeatability, root-mean-square coefficients of variation (rmsCV) for each ROI was calculated as²⁸:

$$\text{rmsCV}=\sqrt{\frac{{\sum }_{i=1}^{N}C{V}_i^2}{N}}$$ (5)

where i is the subject number and N is the total number of subjects (N = 2) in the repeatability study.

Results

T_1ρ-relaxation maps (mono-short-long) of a bovine meniscus from representative slices are shown in Fig. 2a–c, respectively. Considering the estimated T_1ρ values from fully sampled data as a reference, Fig. 2d shows the estimation error for different GRAPPA acceleration factors (AF = 2–4). Using an AF of 3 will decrease the total acquisition time from 31 to 16 minutes, while the estimated relaxations have less than 6% difference from the estimated values derived from fully sampled data. Therefore, AF = 3 was selected for performing the in vivo experiments. The differences between the T_1ρ values using 15 TSLs as a reference with the estimation using 10 and 6 TSLs are shown in Fig. 2e. The results show less than a 5% bias between the relaxation values estimated using 10 TSL points, which is in agreement with our Monte Carlo simulation (Fig. 3b). Therefore, 10 TSLs points were used in the in vivo studies.

FIGURE 2:

Ex vivo study on the bovine menisci. (a) Mono, (b) short, and (c) long component spatial T_1ρ maps. (d) The estimated parameters from scans acquired with an AF = 3 show less than a 6% difference from the fully sampled scans. (e) The estimated relaxations using 10 TSLs showed less than 6% difference from the estimated relaxations using 15 TSLs points.

FIGURE 3:

Monte Carlo simulations. Effect of different parameters on biexponential model. (a) The estimation mean-absolute percentage error (MAPE) decreases by increasing the number of TSLs. (b) Higher SNR leads to less estimation error. (c) The estimation is more accurate for shorter short component (T_1ρs). (d) The estimation is more accurate for larger long component (T_1ρl). (e,f) The component with larger fraction (short component, A_s, in (e) and long component, A_l in (f)) has better estimation than the component with a shorter fraction.

The representative examples of the T_1ρ relaxation maps in the sagittal plane for a healthy subject and an early knee OA patient are shown in Figs. 4 and 5, respectively. The binary maps (Figs. 3–4, a1–c1) show the pixels that were included in the biexponential map by satisfying Eq. 3. The summary of descriptive statistics in different ROIs for healthy subjects and early OA patients are shown in Table 1. The T_1ρ short and T_1ρ long ranges were 5.49–10.60 msec and 58.85–78.44 msec in healthy subjects and 2.00–8.08 msec and 46.29–73.04 msec in OA patients, respectively. The long component fractions were 44.66% and 40.10% for healthy subjects and early knee OA patients, respectively, while the short component fractions were 53.06% and 59.90% for healthy subjects and early knee OA patients, respectively. The average SNR of the experiments was 32.1 ± 9.32. Based on Monte Carlo simulation results (Fig. 3a), there is less than 15% error in estimating the biexponential components for SNR between 25 and 35.

FIGURE 4:

A representative example of T_1ρ relaxation maps of knee menisci in healthy subject in (a1–a6) lateral, (b1–b6) medial, (c1–c6) body menisci. (a1–c1) The binary maps show the pixels that were included in the biexponential map (yellow) by satisfying the Eq. 3 condition.

FIGURE 5:

A representative example of T_1ρ relaxation maps of knee menisci in early OA patient in (a1–a6) lateral, (b1–b6) medial, (c1–c6) body menisci. (a1–c1) The binary maps show the pixels that were included in the biexponential map (yellow) by satisfying the Eq. 3 condition

TABLE 1.

Descriptive Statistics of Mono- and Biexponential Relaxation Times (Mean ± SD and 95% Confidence Intervals) in Nine Regions of Interest for Eight Healthy Controls and Three Early OA Patients

ROI	T1ρmono (msec)	T1ρshort (msec)	T1ρlong (msec)	Yshort (%)	Ylong (%)
	Mean ± SD	Mean ± SD	Mean ± SD	Mean ± SD	Mean ± SD
Control
PLWZ	39.29 ± 5.1	10.60 ± 0.7	78.44 ± 3.8	44.47 ± 8.3	51.37 ± 8.3
PLRZ	28.42 ± 6.19	9.52 ± 1.1	77.86 ± 8.7	57.76 ± 5.4	42.24 ± 15.3
ALWZ	22.20 ± 4.3	6.69 ± 0.6	75.16 ± 4.2	66.97 ± 7.0	33.03 ± 7.0
ALRZ	20.80 ± 5.8	5.49 ± 0.4	63.02 ± 1.8	62.24 ± 6.0	29.43 ± 6.0
PMWZ	42.19 ± 6.8	9.65 ± 1.4	73.64 ± 8.5	39.05 ± 6.2	56.78 ± 6.2
PMRZ	33.96 ± 5.3	8.08 ± 0.7	68.90 ± 4.2	46.92 ± 11.7	44.75 ± 11.7
AMWZ	31.00 ± 6.0	6.99 ± 0.7	75.87 ± 3.5	56.78 ± 7.6	43.22 ± 7.6
AMRZ	30.19 ± 6.1	7.10 ± 0.8	77.80 ± 4.9	60.84 ± 11.1	39.16 ± 11.0
MB	21.34 ± 2.52	6.39 ± 0.6	58.85 ± 1.82	36.0 ± 7.6	64.0 ± 7.6
Patients
PLWZ	37.25 ± 9.50	2.00 ± 2.72	47.29. ± 19.57	55.13 ± 23.53	44.87 ± 23.53
PLRZ	33.25 ± 9.20	2.26 ± 2.73	46.29 ± 21.48	58.88 ± 23.05	41.11 ± 23.05
ALWZ	37.21 ± 8.42	2.54 ± 3.14	52.69 ± 24.84	57.01 ± 21.95	42.99 ± 21.95
ALRZ	42.57 ± 11.61	3.33 ± 3.13	73.04 ± 22.93	58.89 ± 20.50	41.11 ± 20.50
PMWZ	37.91 ± 11.34	3.06 ± 3.11	61.82 ± 23.84	61.86 ± 23.80	38.14 ± 23.80
PMRZ	35.63 ± 9.72	8.08 ± 3.23	68.90 ± 26.56	46.92 ± 24.39	44.75 ± 24.39
AMWZ	38.58 ± 11.01	3.77 ± 3.79	64.83 ± 23.29	61.87 ± 21.76	38.13 ± 21.76
AMRZ	37.02 ± 8.46	2.62 ± 3.29	50.23 ± 20.98	57.10 ± 25.30	42.90 ± 25.30
MB	34.70 ± 7.19	3.23 ± 3.38	59.05 ± 24.92	62.14 ± 23.06	37.86 ± 23.06

SD: Standard Deviation; PLWZ: Posterior-Lateral White Zone; PLRZ: Posterior-Lateral Red Zone; ALWZ: Anterior- Lateral White Zone; ALRZ: Anterior-Lateral Red Zone;; PMWZ: Posterior-Medial White Zone; PMRZ: Posterior-Medial Red Zone; AMWZ: Anterior-Medial White Zone; AMRZ: Anterior-Medial Red Zone;; MB: Meniscus Body.

Figure 6 shows the comparison of the T_1ρ relaxation components in different ROIs. The pairwise statistically significant ROIs are shown with the black dashed line and the rank sum test P-values.

FIGURE 6:

The boxplot shows the T_1ρ relaxation times component between nine different ROIs. The line in the middle of each box is the median and the bars extending from the boxes are the minimum and maximum values. The dashed line with numbers show the pairwise rank sum test between two ROIs (PLWZ: Posterior-Lateral White Zone; PLRZ: Posterior-Lateral Red Zone; ALWZ: Anterior-Lateral White Zone; ALRZ: Anterior-Lateral Red Zone; TLM: Total Lateral Menisci; PMWZ: Posterior-Medial White Zone; PMRZ: Posterior-Medial Red Zone; AMWZ: Anterior-Medial White Zone; AMRZ: Anterior-Medial Red Zone; TMM: Total Medial Menisci; MB: Meniscus Body).

The monoexponential T_1ρ relaxation in PLWZ was significantly different from that of ALWZ and ALRZ (P = 0.012 and P = 0.002, respectively). Moreover, PMWZ demonstrated significantly higher monoexponential T_1ρ relaxation values than that of ALWZ (P = 0.016) and ALRZ (P = 0.003).

Figure 7 shows the comparison of the T_1ρ relaxation components in different ROIs between females and males. The Mann–Whitney U-test showed no significant difference (P > 0.05) between females and males for each component per ROI.

FIGURE 7:

The boxplot shows the T_1ρ relaxation times component (mono, short and long) between different genders.

The comparison between control and patient groups showed a significantly higher mono T_1ρ relaxation in anterior lateral red (P = 0.04) and white (P = 0.02) zones as well as the meniscal body (P = 0.02) in the patients. However, the global difference was not significant (P = 0.3). No difference was observed in the long relaxation time components (P = 0.7). The short component was significantly lower for patients in posterior lateral red (P = 0.02) and white (P = 0.02), posterior medial red (P = 0.04) and white (P = 0.02), anterior medial white (P = 0.02) and body (P = 0.02) zones. The global short relaxation component was also significantly lower in patients (P = 0.02) than in the control group.

Fig. 8a–c shows the mono- and biexponential fit and their corresponding residuals in representative slices of the lateral, medial, and body of the meniscus, respectively.

FIGURE 8:

Monoexponential vs. biexponential fitting models. T_1ρ decay on a logarithmic scale and the fitted residuals of representative pixels in lateral (a), medial (b), and body (c) of knee meniscus. The deviation of the data from a straight line shows the existence of more than one exponential term.

Table 2 shows all repeatability results (CV%, rmsCV%), together with mean T_1ρ values, for each component and for each of the analyzed ROIs. The rmsCV% range was 1.51–16.6% for T_1ρ-mono and 3.59–14.3% and 4.91–15.6% for T_1ρ-short and T_1ρ long, respectively.

TABLE 2.

Intersubject Repeatability of T_1ρ Mono, Long, and Short Components

ROI	T1ρ,mono			T1ρ,long			T1ρ,short
	Mean (msec)	CV	rmsCV%	Mean (msec)	CV%	rmsCV%	Mean (msec)	CV%	rmsCV%
PLWZ	36.42	0.035	4.85	60.67	0.073	9.63	8.58	0.048	4.91
PLRZ	24.22	0.068	4.39	69.62	0.105	11.0	8.08	0.066	6.65
ALWZ	25.09	0.001	13.8	62.47	0.026	3.59	8.18	0.149	14.9
ALRZ	31.75	0.089	8.81	67.17	0.059	5.98	9.00	0.050	5.01
PMWZ	36.06	0.014	1.51	68.63	0.112	11.4	11.245	0.109	10.9
PMRZ	22.22	0.042	4.25	60.69	0.060	6.08	8.77	0.111	11.1
AMWZ	20.64	0.154	16.6	70.66	0.136	14.3	8.05	0.061	6.12
AMRZ	22.54	0.079	8.09	74.13	0.090	11.1	6.72	0.156	15.6
MB	23.45	0.091	12.1	52.92	0.068	7.10	6.1	0.123	12.3

SD: Standard Deviation; PLWZ: Posterior-Lateral White Zone; PLRZ: Posterior-Lateral Red Zone; ALWZ: Anterior- Lateral White Zone; ALRZ: Anterior-Lateral Red Zone; PMWZ: Posterior-Medial White Zone; PMRZ: Posterior-Medial Red Zone; AMWZ: Anterior-Medial White Zone; AMRZ: Anterior-Medial Red Zone; MB: Meniscus Body; CV = Coefficient of Variation; rmsCV = Root-Mean-Square Coefficient of Variation.

Discussion

In this article we present a 3D-MRI technique for bicomponent T_1ρ analysis of subregional, compartmental, and whole knee menisci in nine different ROIs in vivo. The most significant finding of the present study is that biexponential fitting better represents T_1ρ relaxation time values measured in each ROI. Although prior studies have shown that monoexponential fitting can obtain T_1ρ and T₂ values that correlate strongly with water content in OA menisci, these values do not strongly correlate with proteoglycan content and may have limited ability to detect compositional variation in degenerated menisci.^29,30

We reported the biexponential T_1ρ analysis of knee menisci in vivo in different vertical zones. The model provides both short (which is sensitive to early biochemical changes) and long relaxation times with their corresponding fractions. The short component is related to water tightly bound to macromolecules, while the long component is related to water molecules loosely bound to macromolecules.³¹ Hence, we expect that the biexponential T_1ρ provides higher specificity to macromolecular loss in cartilage compared with a monoexponential model. Most of the published literature on T_1ρ relaxation times have been performed utilizing monoexponential fitting; however, the T_1ρ relaxation range in our study was comparable to that of other studies,^{12–14,16,22,32,33} but there were some differences. These differences may be due to two main reasons. First, the slice thickness in this study (2 mm) was smaller compared to other studies, which were 3 mm or 4 mm; this leads to a decrease in partial volume effects (PVEs). Second, we acquired 10 spin-lock times (TSLs) compared with other studies that used 4 to 6 TSLs.

There have been few studies^16,34,35 describing T_1ρ relaxation times in different vertical zones of knee menisci. Takao et al³⁵ reported that results of T_1ρ relaxation times of the white zone were significantly higher than those of the red zone in all segments, which is in agreement with our results. T_1ρ relaxation times reported by Subburaj et al³⁴ were lower than those in our current work; this difference is probably caused by the difference in the number of TSLs and slice thickness, as mentioned above. Other studies have shown similar zonal variation. For instance, Calixto et al¹⁶ investigated T_1ρ relaxation times for both healthy subjects and patients with knee OA in unloaded and loading conditions. In healthy subjects at the unloaded situation, inner lateral and medial posterior horn values were higher than in the outer zone, suggesting underlying differences in macromolecular organization in these subregions of the meniscus. Similarly, Nebelung et al³⁶ investigated an ex vivo quantitative multiparametric MRI mapping study of degenerated human menisci that showed T₁, T_1ρ, T₂, and UTE-T₂* relaxation times are functionally related to histological meniscus degeneration.

The mean of T_1ρ relaxation times in nine different ROIs for early knee OA patients in our study (35.20 msec) was higher than the mean of T_1ρ relaxation times in healthy subjects (29.72 msec). Also, the estimated range of T_1ρ relaxation times for patients in our study are in agreement with previous studies.^37,38

The mean of T_1ρ-short in healthy subjects (7.83 msec) was higher than T_1ρ-short in early knee OA patients (3.00 msec) and the mean of T_1ρ-long in healthy subjects (72.19 msec) was higher than the mean of T_1ρ-long in healthy subjects (57.01). In addition, the short component fractions in early knee OA patients (59.90) was higher than short component fractions in healthy subjects (53.06), while the long component fractions in early knee OA patients (40.10) was less than long component fractions in healthy subjects (44.66).

The investigated reproducibility of T_1ρ relaxation times (rmsCV) numbers in prior studies^35,39 are in good agreement with our experiments, where the rmsCV of 4.91–15.6% was measured for monoexponential T_1ρ.

This study had several limitations. First and foremost, there was a relatively small number of asymptomatic healthy and OA patients recruited. Second, there could be measurement bias resulting from the manually drawn ROIs in different subjects. Third, although we confirmed our experimental condition (4T_S < T_l) for biexponential mapping, there might be some bias. We chose this condition based on the suggestion in the Juras et al study.²⁵ Finally, there was a difference between healthy and early OA patients’ mean ages (healthy: 30, patient: 51).

In conclusion, a 3D Turbo-FLASH-based sequence was implemented for simultaneous mono- and biocomponent T_1ρ relaxation mapping analysis of the human knee meniscus. Our preliminary results show that in all ROIs, T_1ρ relaxation times of outer zones (red zones) less than inner zones (white zones) and T_1ρ mapping can potentially be used to increase the specificity of early OA diagnosis by estimating the relaxation time of different water compartments and their fractions.

Although there is no ground truth for validation of a biexponential model in vivo, we recently used synthetic phantoms with known relaxation times and fractions and validated the biexponential model with different noise levels.⁴⁰