PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Magn Reson Med. Author manuscript; available in PMC 2010 April 1.
Published in final edited form as:
PMCID: PMC2711212
NIHMSID: NIHMS120262

Multicomponent T2 relaxation analysis in cartilage

Abstract

MR techniques are sensitive to the initial phases of osteoarthritis, characterized by disruption of collagen and loss of proteoglycan (PG), but are of limited specificity. Here, water compartments in normal and trypsin-degraded bovine nasal cartilage were identified using a non-negative least squares multiexponential analysis of T2 relaxation. Three components were detected: T2,1 = 2.3 ms, T2,2 = 25.2 ms, and T2,3 = 96.3 ms, with fractions w1 = 6.2%, w2 = 14.5%, and w3 = 79.3%, respectively. Trypsinization resulted in increased (p<0.01) values of T2,2 = 64.2 ms and T2,3 = 149.4 ms, supporting their assignment to water compartments that are bound and loosely associated with PG, respectively. The T2 of the rapidly-relaxing component was not altered by digestion, supporting assignment to relatively immobile collagen-bound water. Relaxation data were simulated for a range of TE, number of echoes, and SNR to guide selection of acquisition parameters and assess the accuracy and precision of experimental results. Based on this, the expected experimental accuracy of measured T2’s and associated weights was within 2% and 4% respectively, with precision within 1% and 3%. These results demonstrate the potential of multiexponential T2 analysis to increase the specificity of MR characterization of cartilage.

Keywords: spin-spin relaxation, macromolecular compartments, cartilage, magnetic resonance

Introduction

Articular cartilage, the connective tissue responsible for transmitting load across joints, is comprised primarily of water and extracellular matrix. The two main matrix constituents are type II collagen, which is configured as a three-dimensional fibrous network resulting in tissue tensile strength, and proteoglycan (PG), a large hydrophilic macromolecule which provides swelling pressure and compressive resistance. By wet weight, cartilage is 65–80% water, 15–25% collagen, and 3–10% PG (1).

A great amount of effort has been focused on non-invasive early detection of osteoarthritis (OA), a highly prevalent and potentially debilitating disease which results in progressive degradation and loss of articular cartilage. The initial phases of OA are characterized by disruption of collagen and loss of PG. MR is increasingly used as a noninvasive diagnostic modality sensitive to early OA, and establishing correlations between MRI parameters such as T1, T2, T, magnetization transfer and apparent diffusion coefficient and cartilage biochemical properties has been an active area of investigation (24). Although attempts have been made to associate average values of such MR parameters to specific tissue characteristics, results have been inconclusive (3).

T2 has consistently been reported to increase in cartilage with pathomimetic enzymatic degradation. T2 is influenced by the mobility of water molecules associated with various effective tissue compartments, defined by cartilage macromolecular components, which would be expected to exhibit a distinct T2 (5).

Multiexponential T2 analysis has previously been applied to identify and characterize multiple water compartments in normal and pathologic tissue (613). As with previous work in other tissues, we expect this approach to lead to insight into water compartmentation in normal and degraded cartilage. We further conjecture that the ability of multiexponential analysis to detect degradation-induced changes in water compartmentation would support its use as a potential diagnostic approach in early OA.

Therefore, the focus of this work was to investigate water compartmentation in cartilage using multiexponential analysis of T2 relaxation data. Analysis was performed using non-negative least squares (NNLS), making no a priori assumptions about the number of relaxation components present. We hypothesized that decomposition of relaxation data into multiple components would permit observation of water compartment shifts in response to pathomimetic enzymatic degradation.

T2 is known to be influenced by tissue anisotropy, such as that found in articular cartilage. In addition, T2 varies with depth in articular cartilage (14,15). Therefore, we used bovine nasal cartilage (BNC), which is largely isotropic and homogeneous, for our investigations (16). We performed enzymatic degradation with trypsin, which cleaves the core and link proteins releasing PG fragments (1).

A specific objective of our study was to determine the optimum acquisition parameters (TE and number of echoes) and minimum SNR required to resolve multiple T2 components. This analysis was performed through NNLS analysis of simulated noisy relaxation data. Results of these simulations allowed us to define experimental acquisition parameters optimized for accuracy and precision in defining T2 components, as well as to specify the error in component T2’s and weights as a function of acquisition parameters and SNR.

Methods

Cartilage preparation

BNC disks (diameter = 8 mm) were excised from the nasal septum of a skeletally mature cow (Green Village Packing, Green Village, NJ). Samples were moistened with DPBS and stored at 4 °C until testing. Within a week of harvest, each sample was removed individually, blotted dry to remove excess surface water, wrapped in Teflon tape to reduce moisture loss, and placed in a 10 mm NMR tube for MR studies. After the initial MR measurements, samples were removed from the NMR tube, unwrapped and again stored at 4 °C. After scanning, samples were incubated in DPBS with 1mg/ml of trypsin (Sigma-Aldrich, St. Louis, MO) for 24 hours at 37 °C and then washed with DPBS. MR measurements were then repeated after this period of enzymatic digestion.

NMR Methods

Data were acquired with a 9.4T Bruker DMX NMR spectrometer (Bruker Biospin, GmbH, Rheinstetten, Germany) at 4 °C in order to minimize degradation of the sample while in the magnet. Relaxation data were measured using a spectroscopic CPMG pulse sequence. Both pre- and post-enzymatic treatment relaxation decays were acquired with parameters TE/TR = 0.6ms/5s and 1024 echoes. In order to achieve desired levels of accuracy and precision, we implemented signal averaging with NEX = 128 in order to achieve SNR in the range of 6000–10000.

Fitting of T2 Relaxation Data

We briefly describe the NNLS method used for multiexponential T2 analysis (17). Multiple T2 component relaxation decay can be represented in discrete form with the following linear system of equations describing measurement of M-1 relaxation components over N echoes (18):

yn=AnmSm,
[1]

where yn is an array containing the N echo amplitudes, Anm is an N × M matrix containing the kernels exp(−n ·TE/T2,m), and Sm is an array containing the M-1 unknown amplitudes associated with each of the M-1 T2’s. The Mth column in matrix A contains all 1’s to allow for a baseline offset to be fit with amplitude SM. For the current study, 80 possible T2 values were evenly spaced on a logarithmic scale over the interval [1, 3000] ms. NNLS finds the non-negative set of Sm which minimizes:

n=1Nm=1MAnmSmyn2.
[2]

In order to minimize the influence of noise on the fit as well as produce a continuous T2 distribution a minimum energy constraint can be imposed so that the following function is minimized:

n=1Nm=1MAnmSmyn2+μm=1MSm2,
[3]

where μ is referred to as a regularizer term. As μ is increased, the permitted misfit increases while reducing the influence of noise on the fit. Eq. [3] is recast in terms of χ2 by appropriate division by the standard deviation of yn, σn (17,18). In the present application, these σn are all equal. For our application, with relatively small N, we use as an optimal condition on μ that χ2 from the regularized fit is 101% of the non-regularized χ2 (19). We reiterate that the NNLS procedure as summarized above, unlike nonlinear least-squares fits to bi- or triexponentials, makes no a priori assumptions about the number of components present.

The T2 analysis was implemented in MATLAB (MathWorks, Natick, MA). The first moment and associated fractions for each identified component of the T2 distribution were obtained.

Statistical Analysis

Means ± standard deviation (SD) for control and trypsin degraded samples are reported. The differences in the mean T2 values between control and trypsin degraded samples were tested for statistical significance using a paired t test, with p < 0.05 indicating statistical significance.

Simulation of T2 Relaxation Data

As will be discussed, three T2 components were consistently detected by NNLS in control BNC samples. Using the average values for these component T2’s and weights as input parameters, data were simulated using the following expression:

y(n·TE)=B+y0m=1Mwm·e(n·TE)/T2,m+ε(0,σ),
[4]

where y(n · TE) represents the amplitude of the nth echo measured at n · TE, B represents a baseline offset, y0 represents overall signal amplitude, wm represents the fractional weight of the mth T2 component, and ε(0,σ) represents additive Gaussian random noise with mean 0 and standard deviation σ.

Simulations were performed for a range of echo spacings TE (0.4 ms to 1.6 ms), number of echoes N (256, 512, or 1024) and SNR, defined as y0/σ (500 to 16000). Combinations of TE and N were selected to permit full decay of simulated relaxation data. 100 trials each with different noise realizations were performed for each set of TE, N, and SNR. Regularized fits were classified as admissible or inadmissible, with admissibility being defined as i) Reliability: 90% of the runs with independent noise realizations identified three T2 components; ii) Accuracy: identified component T2 values and associated weights were accurate to within 10% of the simulation input values and iii) Precision: identified component T2 values and associated weights were precise to within 10% of the simulation input values. Therefore, this analysis describes the SNR required to perform admissible experiments with a given TE and N. These simulation results were used to guide selection of the experimental acquisition parameters, including the amount of signal averaging, and indicate the expected accuracy and precision of the experimental results.

Results

Simulations

The results in Figure 1 indicate the optimum TE for a given number of echoes for reliable T2 relaxation measurements using NNLS in BNC. For all N shown, the relationship between SNR and TE is roughly convex, indicating an optimum echo time for a given N. For example, if N = 256, the required SNR for admissibility is 2600, achieved using a TE = 1.2 ms. Optimum echo times over the range of N simulated were between 0.8 ms and 1.2 ms. In addition, simulations generally indicated an SNR requirement for admissibility that decreased with increasing N.

Figure 1
Simulation results using for the minimum SNR required for admissible experiments over various combinations of TE and number of echoes (N). SNR values are admissible on and above each curve. In general, the SNR requirement is relaxed with increasing N. ...

Figure 2 shows the accuracy of T2 and associated weight determination for TE = 1.2 ms and N = 512. These acquisition parameters are those used experimentally, since only data from even numbered echoes were analyzed. As seen, accuracy increases with increasing SNR. An SNR of 8000, comparable to experimental values, yielded accuracy in T2’s and associated weights within 2% and 4% respectively. At low SNR, the algorithm underestimated the relaxation time of each component. The fractional weight w2 was underestimated while w1 and w3 were overestimated at low SNR. Since weights must sum to 100%, this result is not surprising. As also seen in Fig. 2, at low SNR the percent errors in T2 and w were substantially greater magnitude for component 2 than for those of the other components.

Figure 2
Simulation results for the percent error in component relaxation times and fractions as a function of SNR for TE = 1.2 ms and N = 512. Percent error is calculated as fitted value minus input value divided by input value. This figure shows how accuracy ...

Figure 3 indicates the precision of T2 and associated weight determinations, again for TE = 1.2 ms and N = 512. Parameter precision was defined by their coefficient of variation calculated by dividing the standard deviation over the trials with different noise realizations by the simulation input value for each parameter. For SNR = 8000, the coefficients of variation for T2’s and their associated weights were within 1% and 3%, respectively. T2,1 and w1 exhibited the largest CV, primarily due to the small denominators in the corresponding calculations. However, T2,2 and w2 exhibited the largest SD’s in absolute terms, consistent with Fig. 2.

Figure 3
Simulation results for the coefficient of variation (CV) in component relaxation times and fractions as a function of SNR for TE = 1.2 ms and N = 512. CV is calculated as the standard deviation over 100 trials divided by input value for each parameter. ...

Experiments

Results for monoexponential and NNLS analysis of the experimental decay curves are presented in Table 1. Monoexponential analysis of data from control samples yielded T2 = 84 ± 3 ms, while three distinct compartments were consistently detected with NNLS. These components were: T2,1 = 2.3 ± 0.6 ms, T2,2 = 25.2 ± 6.2 ms, and T2,3 = 96.3 ± 4.7 ms, with associated fractions: w1 = 6.2 ± 2.5%, w2 = 14.5 ± 3.9%, and w3 = 79.3 ± 6.1%, respectively. As expected, trypsin degradation resulted in a significant increase in the monoexponential T2, to 134.5 ± 5.2 ms. T2,1 was not affected by degradation, while T2,2 and T2,3 increased with degradation, to 64.2 ± 19.3 ms and 149.4 ± 11.8 ms, respectively. The corresponding weight fractions were 3.1± 1.3%, 10.9 ± 10.5%, and 86.0 ± 11.2%. It is important to note that these fractions represent relative changes in water compartment size, as the net increase or decrease of water in the sample was not detected. From this analysis it is therefore impossible to determine whether the absolute size of a particular water compartment changed with degradation, and direct comparison of fractions pre- and post-degradation are not valid unless one assumes that total tissue water content is unchanged. However, assessment of relative shifts of water protons between compartments is independent of this assumption. Figure 4 shows typical T2 distribution results for control and trypsin degraded BNC.

Figure 4
Typical T2 distributions obtained using NNLS analysis of relaxation data acquired from a BNC sample before and after trypsin degradation. Trypsin degradation resulted in a negligible change in T2,1 but a lengthening of T2,2 and T2,3.
Table 1
T2 Analysis of control and degraded BNC

Discussion

Previous work has demonstrated the utility of multiexponential T2 analysis to characterize tissue water compartments in a variety of tissues. In studies of normal muscle, Saab et al. (20) found four T2 components, with relaxation times of <5 ms assigned to macromolecule-bound water, ~20 ms and ~40 ms assigned to intracellular water compartments, and >100 ms assigned to extracellular water. Assignment of T2 components to intra- and extracellular water compartments was performed in the study of a rat model in which edema was presumed to result in an increase in extracellular water fraction and a decrease in the intracellular water fraction (21). MacKay et al. found that brain white matter exhibited three T2 components: a minor component with relaxation time between 10 and 55 ms associated with myelin membranes, an abundant component with T2 between 70 and 95 ms associated with cytoplasmic and extracellular spaces, and a slowly-relaxing compartment with T2 > 1 s associated with cerebral spinal fluid (22). Multicomponent T2 analysis has also demonstrated that the myelin-associated water fraction is reduced or absent in pathologic conditions known to effect myelination, such as multiple sclerosis and possibly schizophrenia (10,11). Our work echoes these previous investigations in the context of cartilage water compartments, where multiexponential analysis indicates changes in relative water pool sizes under pathomimetic intervention.

Multiexponential T2 analysis has previously been used to evaluate anisotropy in a variety of tissues, including bovine articular cartilage (23). In that study, also using NNLS analysis, the change in the measured T2 distribution as a function of cartilage orientation was taken as an indicator of macromolecular organization and anisotropy. The examples reported qualitatively resemble our results for the intermediate and long T2 components in relative size and T2 values. Henkelman et al. attribute the appearance of multiple T2 components in articular cartilage to distinct tissue layers exhibiting different relaxation times. In the case of BNC, as discussed above, layered organization is not present to any significant degree and the identified T2 components are attributable to tissue compartments that are not spatially distinct. Mosher et al. also observed multiexponential T2 relaxation in porcine articular cartilage, using using nonlinear least-squares fits to mono-, bi-, and triexponentials (24). These authors found T2 relaxation distributions compatible with a three compartment model; as was also the case in Henkelman et al., this was primarily attributed to spatial heterogeneity, with different tissue regions contributing different values of T2. It is important to note that in both of these previous studies, as also for our work, T2 determinations were made from non-localized spectroscopic experiments, rendering it impossible to differentiate regional variation in relaxation properties from pools that are co-localized but exhibit different relaxation rates. It was for this reason that we performed our study using BNC, which is known to exhibit much less tissue anisotropy and inhomogeneity as compared to articular cartilage; this permits us to interpret our results in terms of water compartments without the confounding effect of regional differences. We note that the matrix composition of BNC is otherwise very similar to that of articular cartilage, so that our results should correspond at least qualitatively to results obtained in analyses of articular cartilage.

Simulations

Graham et al. (19) used simulated decay data to determine the accuracy of multicomponent T2 analysis in white matter, muscle, and breast. Similarly, our simulations were directed towards evaluating the reliability, accuracy, and precision in fitting multiple T2 components in BNC using NNLS. Consistent with the finding of Graham et al., we find that the utility of multiexponential analysis relies upon the appropriate selection of TE and N, in addition to sufficient SNR.

In general, the minimum required SNR decreased with increasing number of echoes for a given TE. This result relies upon our selection of reasonable combinations of TE and N. That is, sampling the relaxation decay extends throughout the full duration of that decay without extending beyond the point of nearly-complete signal decay. This avoids sampling of noise without underlying signal.

Our simulations, based on experimental component T2’s and weights for BNC, also allowed us to predict the error in component T2’s and weights as a function of SNR for experimental TE and N. As expected, the relationship between error and SNR was influenced by the input T2 values and weights. For example, when using input parameters from measurements of undigested BNC, w2 was underestimated while w1 and w3 were overestimated. However, when using simulation input values corresponding to degraded tissue, for which the values of T2,2 and T2,3 were closer together than in control tissue, w1 and w2 were underestimated while w3 was overestimated (data not shown). We also found that the minimum required SNR for admissibility was generally determined by the accuracy criterion, rather than by the precision (data not shown). When using NNLS analysis of relaxation data to categorize cartilage as healthy or degraded, the required minimum SNR could be relaxed when the change in a particular component of interest is expected to be large. In particular, the 10% criterion on accuracy and precision used here for admissibility could be raised to a larger value.

The optimum acquisition parameters and minimum SNR for admissibility were found to be sensitive to the T2’s and weights of each component. In particular, when T2 values of two water components are within a factor of roughly 2, the minimum SNR necessary for consistently resolving those components significantly increases. This is consistent with previous findings (18). Given that T2 components change with matrix degradation, NNLS evaluation of cartilage must make use of experimental parameters and SNR values that will permit appropriate evaluation at all stages of degradation.

In their analysis of multiple T2 compartments in peripheral muscle (19), Graham et al. also found three relaxation components and demonstrated a convex relationship between SNR and TE using simulation inputs applicable to muscle. These authors used a different criterion for admissibility, targeted to distinguishing between slow and fast twitch muscle based on the accuracy of the first moment of the entire T2 distribution as opposed to the accuracy of the first moment of each individual component T2 and weight as in the present work. Using this criterion, Graham et al. found a minimum required SNR in the range of 100 to 400. We used a more restrictive constraint to select minimum SNR since in our application we were interested in identifying differences in water compartments between normal and degraded BNC for which the T2 values were unknown.

Experiments

We used trypsin as a degradative enzyme to model the loss of PG associated with early stages of OA. Our results are consistent with previous work that reported an increase in the monoexponential T2 of cartilage as a result of trypsinization. For example, Menezes et al. reported an increase in monoexponential T2 from 83 ms to 95 ms with trypsinization in bovine cartilage from the femoral groove (3). These authors also attributed this increase in T2 to an increase in the size of the unbound water compartment.

Ghiassi-Nejad et al. (25) characterized the stoichiometry of articular cartilage through use of dehydration and deuteration. They performed a fit of T2 decay data to a biexponential function under various conditions. In deuterated samples, rapidly-relaxing tightly bound protons were visible (using direct measurements of the free induction decay signal) as well as a component with T2 ranging from 15 to 30 ms, assigned to non-exchangeable protons on mobile PG side-chains. The slowly-relaxing component of non-deuterated samples exhibited a T2 of 60 to 80 ms, indicating greater average mobility (25). As noted, the NNLS analysis used in our study makes no a priori assumption about the number of water compartments present, allowing for the separate identification of water tightly bound to PG and water loosely associated with PG. Our reported T2,2 and T2,3 correspond closely to the slowly-relaxing T2 component of deuterated and hydrated cartilage samples respectively, as reported by Ghiassi-Nejed et al. These similarities support our assignment of T2,2 and T2,3 to water tightly bound to PG and water associated loosely with PG, respectively.

We found that enzymatic degradation with trypsin resulted in a significant lengthening of T2,2 and T2,3 in addition to a small corresponding shift from w2 to w3. This shift is consistent with a loss of water tightly bound to PG accompanied by a relative increase in water loosely associated with PG. The lengthening of both relaxation components is consistent with the fact that while trypsin cleaves the core and link protein of PG, it does not ensure the diffusion of PG fragments from the matrix. Water associated with GAG fragments would be expected to demonstrate an increased T2 as compared to that of intact PG. The lengthening of both T2,2 to T2,3 with trypsin suggests that both of these compartments are associated with PG, and further supports their assignment to water tightly bound to PG and water associated loosely with PG. Lattanzio et al. also investigated cartilage before and after trypsin degradation using a biexponential fit of T2 relaxation data (5). Consistent with our results, they described a slowly-relaxing T2 component, assigned to PG, whose relaxation time increased with degradation. No significant change in water fractions was observed. In addition, this slowly-relaxing T2 component accounted for 97% of the signal, similar to the sum of w2 and w3 in our work, in which these slowly-relaxing pool fractions accounted for 92% and 98.6% of the signal in control and degraded tissue, respectively.

Due to the markedly lower mobility of collagen-bound water in comparison with PG (2628), we associateT2,1 with a collagen-bound water compartment. This is supported by the lack of observed change in T2,1 upon degradation with trypsin. Knauss et al. observed biexponential T2 relaxation in collagen gel using multiexponential T2 analysis, reporting fast and slow components with T2’s of 6.5 ms and 120 ms, respectively (29). They assigned the rapidly-relaxing component to collagen-bound water and the more slowly-relaxation component to bulk water. The short T2 of the collagen-bound water fraction is comparable to our value of T2,1 = 2.3 ms. Nightingale et al. (30) described four relaxation components in the nucleus pulposus of the intervertebral disc, and used multiple regression to correlate the water fractions of matrix components with magnetization fractions. Collagen correlated strongly with the most rapidly-relaxing component, with T2 = 3.1 ms and fractional weight of 3%. These are comparable to our values of T2,1 = 2.3 ms and w1 = 6.2%, respectively. We note that the nucleus pulposus has a similar matrix composition to BNC (16), so that these results support the assignment of the rapidly-relaxing pool in our studies to being collagen-bound water.

We found a good correspondence between the water components identified in our experiments, where the weights associated with water bound to collagen, water tightly bound to PG, and bulk water loosely associated with PG were ~6%, ~14%, and ~80%, respectively, and previous biochemical analysis of BNC, in which fractions of collagen, PG, and water by wet weight were ~10%, ~13% PG, and ~75% respectively (16). Our finding of a smaller value for collagen-associated water as compared to the actual matrix fraction of collagen is consistent with the reportedly smaller binding capacity of collagen as compared to PG per mg (25). In spite of the correspondence of our results with biochemical analysis, we note that several factors can influence the relationship between MR-detected magnetization fractions and macromolecular composition. One such factor is signal visibility based on T2; non-exchangeable protons in collagen have been reported to have a T2 of ~30 μs (5), and so would not be detectable with the TE used in our study. In addition, the relationship between magnetization fractions and matrix components may be affected by proton exchange between compartments. For example, Lattanzio et al. used a four-spin component exchange model in their study of articular cartilage (5). A magnetization exchange rate between PG and collagen of 120 s−1 was found, which is intermediate to the corresponding nominal relaxation rates in our analysis of 1/T2,1 = 438 s−1 and 1/T2,2 = 39.6 s−1. We would therefore expect that the T2’s and associated fractions measured in our experiment would be influenced by this exchange.

As indicated previously, avoiding the confounding effect of regional variation is an essential element in interpreting multicomponent T2 data in terms of water compartmentation. For studies of articular cartilage, including clinical application to osteoarthritis, this would be accomplished through use of imaging experiments and region-of-interest analysis of T2 decay in multi-echo experiments. Important differences between spectroscopic and imaging multi-echo sequences are the longer minimum TE’s and lower SNR’s generally available in the latter. Both simulated and experimental data demonstrate that, for realistic values of SNR, a rapidly relaxing water compartment cannot be detected when the minimum TE approaches its T2 (data not shown). Therefore, the compartment we have identified as T2,1 would not be detectable using a clinical imaging sequence, although T2,2 and T2,3 would be visible. To assess the adequacy of imaging experiments for accurate description of these more slowly-relaxing components, we have performed simulations with an SNR of 500 and which consider only T2,2 and T2,3. The results indicate reliability (as defined in the Methods above) of 100% with accuracy and precision of at least ~12% and ~5% for the derived T2,2 values, respectively, with even more favorable results being obtained for T2,3. We also found the accuracy and precision of the estimate of w2 to be ~19% and ~6%, respectively, with even more favorable results being obtained for w3. This is clearly adequate for describing the large changes of ~155% and ~55% in T2,2 and T2,3 demonstrated in this work using trypsin degradation, as well as the possibly much smaller changes expected during the early stages of degenerative cartilage disease.

As noted above, relaxation of the admissibility criteria used in the current study would lead to a decrease in the required SNR for performing acceptable studies. Detection of only the two more slowly-relaxing T2 components in a clinical imaging experiment as detailed here is, in effect, an example of this; violations of admissibility in the spectroscopic simulations were primarily attached to the rapidly-relaxing T2,1 component.

In conclusion, multiexponential T2 analysis detected three distinct water compartments in normal and degraded BNC, assignable to collagen bound water, PG bound water, and water loosely associated with PG. The two slowly-relaxing T2 components were significantly affected by trypsinization of cartilage. These results indicate the potential for multiexponential T2 analysis to play a diagnostic role in the early detection of OA through monitoring PG depletion. This is the first study in which simulations were used to determine the optimum acquisition parameters for resolving multiple T2 components in cartilage. Simulations of decay data over a range of TE, N, and SNR allowed for characterization of the reliability, accuracy and precision of T2 NNLS analysis.

Acknowledgments

This work was supported by the Intramural Research Program of the NIH, National Institute on Aging.

References

1. Hall BK, Newman SA. Cartilage: molecular aspects. Boca Raton, FL: CRC Press, Inc; 1991.
2. Laurent D, Wasvary J, Yin JY, Rudin M, Pellas TC, O’ Byrne E. Quantitative and qualitative assessment of articular cartilage in the goat knee with magnetization transfer imaging. Magnetic Resonance Imaging. 2001;19(10):1279–1286. [PubMed]
3. Menezes NM, Gray ML, Hartke JR, Burstein D. T2 and T1rho MRI in articular cartilage systems. Magnetic Resonance In Medicine. 2004;51(3):503–509. [PubMed]
4. Mlynarik V, Sulzbacher I, Bittsansky M, Fuiko R, Trattnig S. Investigation of apparent diffusion constant as an indicator of early degenerative disease in articular cartilage. Journal of Magnetic Resonance Imaging. 2003;17(4):440–444. [PubMed]
5. Lattanzio PJ, Marshall KW, Damyanovich AZ, Peemoeller H. Macromolecule and water magnetization exchange modeling in articular cartilage. Magnetic Resonance In Medicine. 2000;44(6):840–851. [PubMed]
6. Ababneh Z, Beloeil H, Berde CB, Gambarota G, Maier SE, Mulkern RV. Biexponential parameterization of diffusion and T2 relaxation decay curves in a rat muscle edema model: Decay curve components and water compartments. Magnetic Resonance in Medicine. 2005;54(3):524–531. [PubMed]
7. English AE, Whittall KP, Joy MLG, Henkelman RM. Quantitative two-dimensional time correlation relaxometry. Magnetic Resonance in Medicine. 1991;22(2):425–434. [PubMed]
8. Graham SJ, Bronskill MJ, Byng JW, Yaffe MJ, Boyd NF. Quantitative correlation of breast tissue parameters using magnetic resonance and X-ray mammography. British Journal of Cancer. 1996;73(2):162–168. [PMC free article] [PubMed]
9. Graham SJ, Ness S, Hamilton BS, Bronskill MJ. Magnetic resonance properties of ex vivo breast tissue at 1.5 T. Magnetic Resonance in Medicine. 1997;38(4):669–677. [PubMed]
10. Kolind SH, Laule C, Vavasour IM, Li DK, Traboulsee AL, Madler B, Moore GR, Mackay AL. Complementary information from multi-exponential T2 relaxation and diffusion tensor imaging reveals differences between multiple sclerosis lesions. Neuroimage. 2008;40(1):77–85. [PubMed]
11. MacKay A, Laule C, Vavasour I, Bjarnason T, Kolind S, Madler B. Insights into brain microstructure from the T2 distribution. Magnetic Resonance Imaging. 2006;24(4):515–525. [PubMed]
12. Peled S, Cory DG, Raymond SA, Kirschner DA, Jolesz FA. Water diffusion, T2, and compartmentation in frog sciatic nerve. Magnetic Resonance in Medicine. 1999;42(5):911–918. [PMC free article] [PubMed]
13. Stanisz GJ, Henkelman RM. Diffusional anisotropy of T2 components in bovine optic nerve. Magnetic Resonance in Medicine. 1998;40(3):405–410. [PubMed]
14. Mosher TJ, Dardzinski BJ. Cartilage MRI T2 relaxation time mapping: overview and applications. Seminars in Musculoskeletal Radiology. 2004;8(4):355–368. [PubMed]
15. Xia Y, Moody JB, Alhadlaq H. Orientational dependence of T2 relaxation in articular cartilage: a microscopic MRI (micro MRI) study. Magnetic Resonance in Medicine. 2002;48(3):460–469. [PubMed]
16. Torchia DA, Hasson MA, Hascall VC. Investigation of molecular motion of proteoglycans in cartilage by 13C magnetic resonance. Journal of Biological Chemistry. 1977;252(11):3617–3625. [PubMed]
17. Lawson CL, Hanson RJ. Solving least squares problems. Englewood Cliffs, NJ: Prentice-Hall; 1974.
18. Whittall KP, Mackay AL. Quantitative interpretation of NMR relaxation data. Journal of Magnetic Resonance. 1989;84(1):134–152.
19. Graham SJ, Stanchev PL, Bronskill MJ. Criteria for analysis of multicomponent tissue T2 relaxation data. Magnetic Resonance in Medicine. 1996;35(3):370–378. [PubMed]
20. Saab G, Thompson RT, Marsh GD. Multicomponent T2 relaxation of in vivo skeletal muscle. Magnetic Resonance in Medicine. 1999;42(1):150–157. [PubMed]
21. English AE, Joy MLG, Henkelman RM. Pulsed NMR relaxometry of striated-muscle fibers. Magnetic Resonance in Medicine. 1991;21(2):264–281. [PubMed]
22. MacKay A, Whittall K, Adler J, Li D, Paty D, Graeb D. In-vivo visualization of myelin water in brain by magnetic-resonance. Magnetic Resonance in Medicine. 1994;31(6):673–677. [PubMed]
23. Henkelman RM, Stanisz GJ, Kim JK, Bronskill MJ. Anisotropy of NMR properties of tissues. Magnetic Resonance In Medicine. 1994;32(5):592–601. [PubMed]
24. Mosher TJ, Dardzinski BJ, SM B. Characterization of multiple T2 components in articular cartilage; Proceedings of the International Society for Magnetic Resonance in Medicine 5th Annual Meeting; 1997. p. 1331.
25. Ghiassi-Nejad M, Torzilli PA, Peemoeller H, Pintar MM. Proton spin-spin relaxation study of molecular dynamics and proteoglycan hydration in articular cartilage. Biomaterials. 2000;21(20):2089–2095. [PubMed]
26. Zernia G, Huster D. Collagen dynamics in articular cartilage under osmotic pressure. NMR in Biomedicine. 2006;19(8):1010–1019. [PubMed]
27. Huster D, Schiller J, Arnold K. Comparison of collagen dynamics in articular cartilage and isolated fibrils by solid-state NMR spectroscopy. Magnetic Resonance in Medicine. 2002;48(4):624–632. [PubMed]
28. Schaefer J, Stejskal EO, Brewer CF, Keiser HD, Sternlicht H. Cross-polarization 13C nuclear magnetic resonance spectroscopy of collagen. Archives of Biochemistry and Biophysics. 1978;190(2):657–661. [PubMed]
29. Knauss R, Fleischer G, Grunder W, Karger J, Werner A. Pulsed field gradient NMR and nuclear magnetic relaxation studies of water mobility in hydrated collagen II. Magnetic Resonance in Medicine. 1996;36(2):241–248. [PubMed]
30. Nightingale T, MacKay A, Pearce RH, Whittall KP, Flak B. A model of unloaded human intervertebral disk based on NMR relaxation. Magnetic Resonance in Medicine. 2000;43(1):34–44. [PubMed]