Ultrashort–Echo Time MR Imaging of the Patella with Bicomponent Analysis: Correlation with Histopathologic and Polarized Light Microscopic Findings

Short T2* water fraction parallels cartilage degeneration, while Carr-Purcell-Meiboom-Gill imaging T2 values show a low correlation with degeneration.

Abstract

Purpose:

To correlate short and long T2* water fractions, derived from ultrashort–echo time (TE) magnetic resonance (MR) imaging, with semiquantitative histopathologic and polarized light microscopic (PLM) assessment of human cadaveric patellae cartilage.

Materials and Methods:

Twenty human cadaveric patellae were evaluated by using ultrashort-TE imaging, spin-echo imaging, histopathologic analysis, and PLM, with institutional review board approval. Short and long T2* water components were evaluated for each patella by using bicomponent fitting of ultrashort-TE signal decay. Four to six regions of interest (ROIs) within each patella were chosen for correlation between ultrashort-TE bicomponent analysis, histopathologic grading (Mankin score), and PLM grading (Vaudey score).

Results:

Ultrashort-TE imaging with bicomponent analysis showed two distinct water components with a short T2* and a longer T2* in all patellae. ROI analysis showed that the short T2* fraction was correlated significantly with the Mankin (ρ = 0.66, P < .001) and Vaudey (ρ = 0.68, P < .001) scores. The Mankin scores were weakly positively correlated with T2 (ρ = 0.28, P = .13) and short T2* (ρ = 0.24, P = .14) but were negatively correlated with long T2* (ρ = −0.55, P < .01). The Vaudey scores were weakly positively correlated with T2 (ρ = 0.18, P = .16) and short T2* (ρ = 0.22, P = .14) but were negatively correlated with long T2* (ρ = −0.55, P < .01).

Conclusion:

Short T2* water fraction derived from ultrashort-TE imaging with bicomponent analysis correlates significantly with both the Mankin and Vaudey scores and may serve as a biomarker of cartilage degeneration.

©RSNA, 2012

Introduction

Articular cartilage is a dense connective tissue that consists of 15%–22% collagen, 4%–7% proteoglycans (PGs), and 60%–80% water by weight (1). PGs create a high swelling pressure, which is contained by the collagen network and helps maintain the cartilage’s biomechanical properties (1,2). In the early stages of osteoarthritis (OA), cartilage extracellular matrix is disrupted, with loss of PGs, degradation of collagen microstructure, and increase in water content (2).

In OA, affected cartilage water content may increase by about 10% (3–5). Investigators in several studies (6–8) proposed measurement of water content as a biomarker of cartilage degeneration. However, the investigators in all these studies used conventional clinical spin-echo or gradient-echo sequences with relatively long echo times (TEs), which may only have the capability for detection of signal from long T2 water components.

The majority of the water in cartilage exists in the form of free water, with a smaller fraction bound to either PGs or collagen fibrils (9,10). Bound water has a much shorter T2 than free water (10). Ultrashort-TE sequences have been developed to image short-T2 tissues or tissue components in vitro and in vivo and have been implemented on clinical MR imaging units (11–15). Recently, we reported the use of ultrashort-TE imaging together with bicomponent analysis to evaluate short T2* (predictive of bound water) and longer T2* (predictive of free water) components in joint tissues (16). In this study, we aimed to correlate short and long T2* water fractions, derived from ultrashort-TE MR imaging, with semiquantitative histopathologic and polarized light microscopic (PLM) assessment of human cadaveric patellar cartilage.

Materials and Methods

Human Patellae Procurement

Twenty fresh human patellae from 11 donors, seven of whom were male (age range, 48–90 years; mean age, 63.4 years ± 16.0 [standard deviation]) and four of whom were female (age range, 50–92 years; mean age, 94.3 years ± 20.3), were obtained from tissue banks approved by our institutional review board (Scripps Research Institute, La Jolla, Calif) and processed within 24–72 hours of death. After harvesting, a transverse slab of 5–8-mm thickness was cut and stored in phosphate-buffered saline–soaked gauze at 4°C prior to MR imaging.

MR Data Acquisition

A two-dimensional ultrashort-TE imaging sequence was implemented with a 3-T MR unit (Signa TwinSpeed; GE Healthcare Technologies, Milwaukee, Wis), with a maximum gradient strength of 40 mT/m and a maximum slew rate of 150 mT/m/msec. With the ultrashort-TE sequence, half-pulse–radiofrequency excitation, together with radial ramp sampling and fast transmit-receive switching, was used to allow a minimal nominal TE of 8 μsec (17). A homebuilt 1-inch (2.54-cm) transmit-receive birdcage coil was used for signal excitation and reception. The patella samples were placed in perfluorooctyl bromide solution to minimize susceptibility effects at tissue-air junctions. A single section at the center of each patella sample was imaged, with the apex normal to the B₀ field. Typical imaging parameters were as follows: repetition time msec/TE msec, 200/0.008, 0.1, 0.2, 0.4, 0.6, 0.8, 3, 5, 10, 20, 40, 60, and 80 (13 TEs); field of view, 6 cm; reconstruction matrix, 512 × 512; section thickness, 1.7 mm; and acquisition time, 2 minutes per image. Carr-Purcell-Meiboom-Gill (CPMG) acquisitions with eight echoes (TE = 12.2, 24.5, 36.7, 48.9, 61.2, 73.4, 85.6, and 97.9 msec) were acquired with the same spatial resolution and a 12-minute image time.

The reproducibility of ultrashort-TE bicomponent analysis was estimated with coefficients of variation calculated as the ratio of the standard deviation to the average value. Average coefficients of variation were calculated for the relative fractions and T2* values of bound and free water components.

Tissue Processing

After MR imaging, the patellae slabs were immediately fixed in Z-Fix (Anatech, Battle Creek, Mich) for 3 days and then decalcified with a reagent (TBD-2; Thermo Scientific, Waltham, Mass). The centers of the samples were marked with a tissue marking dye (Cancer Diagnostics, Morrisville, NC) on the lateral and medial edges to provide orientation. After full decalcification, dehydration with alcohol, clearing with an agent for that purpose (Pro-Par Clearant; Anatech, Battle Creek, Mich), and infiltration with paraffin embedding medium (Paraplast; McCormick Scientific, Richmond, Ill), transverse sections covering the cartilage and subchondral bone were obtained. Each tissue block was trimmed on a microtome by using the orientation marks for reference. Sections of 5 μm were cut at the defined central location to match the MR images. Several sections from each patella were stained with safranin O–fast green for histopathologic analysis and Picrosirius Red for PLM.

Histopathologic Analysis

Each safranin O–fast green–stained slide was scanned with a slide scanner (SCN4000; Leica Microsystems, Buffalo Grove, Ill) and viewed with software (SlidePath; Leica Microsystems). The slab of an entire patella most likely contains several regions with different histopathologic grades. The goal from a histopathologic analysis standpoint was to cover all the different grades within one specimen. That was done with four to six regions of interest (ROIs) per patella chosen for correlation with histopathologic, PLM, and MR imaging findings. Each ROI was assigned a Mankin score ranging from 0 to 14 by one author (C.P., with 6 years of experience in musculoskeletal histopathology with a primary focus on articular cartilage) (18). Each score was converted to a grade as follows: grade 1 for scores 0–1; grade 2 for scores 2–5; grade 3 for scores 6–9; and grade 4 for scores 10–14. Grade 1 represented normal tissue; grade 2, mild degeneration; grade 3, moderate degeneration; and grade 4, severe degeneration. Table 1 describes details of this histologic scoring system.

Table 1.

Original Histologic-Histochemical Grading Scale with Individual Parameters and Scores

Source.—Reference 18.

Note.—The minimum score is zero, and the maximum score is 14.

Polarized Light Microscopy

PLM for the collagen network analysis was performed with the Picrosirius Red–stained sections centered on a rotating stage of a microscope (Olympus BX60; Shinjuku, Tokyo, Japan). The specimens were placed between a polarizer, a 546-nm interference filter, a Senarmont compensator, and an analyzer. The polarizer and analyzer were orthogonally aligned to each other. Polarized light was used to transilluminate the specimen. Once polarized light was scattered by the sample, it traveled through the analyzer and was detected by using a calibrated charge-coupled device camera (Macrofire Optronics, Goleta, Calif). PLM images were recorded and saved. Each ROI that had been given a histopathology Mankin score was qualitatively assessed by using the grading scale (grade 0–4) published by Vaudey to describe the birefringence characteristics of the articular cartilage matrix (19,20). Details of this scoring system are shown in Table 2.

Table 2.

PLM Characteristics Representing Five Grades

Source.— Reference 19.

Postprocessing and Image Analysis

The T2 and bicomponent analysis algorithms were written in a numerical computing environment and programming language (Matlab; MathWorks, Natick, Mass). The mean signal intensity within each of the ROIs was used for subsequent curve fitting. Biexponential signal decay fitting was performed on data from the ultrashort-TE images to assess short and long T2* water components. Single-component fitting was performed on data from the CPMG images for T2 measurement. Goodness-of-fit statistics, including theR² value and standard error or fitting confidence level, were calculated. Fit curves, along with their 95% confidence intervals, and residual-signal curves were created.

Multicomponent fitting is sensitive to image signal-to-noise ratio, the number of fitting components, and the difference between T2 values (21–23). The following approaches were used to improve curve fitting. First, only two components were assumed. Second, the ultrashort–TE T2* signal was normalized before bicomponent fitting. Third, background noise was estimated by using a fully automated comprehensive four-step noise estimation algorithm primarily based on maximum likelihood estimation model (24). Fourth, a noise-corrected model that was based on the Bessel function of the first kind of n^th order was used for bicomponent fitting (25,26). Because only three parameters, namely short T2*, long T2*, and the short or long T2* fraction, were fitted in this bicomponent model, robust fitting could be achieved with clinically achievable signal-to-noise ratios (16).

Multiple regions in each patella were used for analysis. The number of regions was determined by one author (C.P.), depending on the grade of OA severity. The same ROIs were used for ultrashort–TE T2*, CPMG T2, and PLM analyses. For each region, four sub-ROIs that included the deep zone (the third of the cartilage thickness adjacent to subchondral bone), the middle zone (the third in the middle of the cartilage), and the superficial zone (the superficial third), as well as a global ROI comprising the entire region, were assessed.

Statistical Analysis

Short and long T2* values and their fractions, as well as T2, were correlated with the Mankin and Vaudey scores. Spearman rank correlation was used, and its significance was assessed. Because multiple measurements were obtained from the same donor, the nonparametric bootstrap method was used to assess the significance of the Spearman correlation (27). The resampling in the bootstrap replicates was performed per subject to adjust for within-subject dependence. Significance of the correlation was assessed on the basis of the bias-corrected, accelerated bootstrap confidence interval around the correlation coefficient. The P values for the correlations were calculated on the basis of the bootstrap. A P value of less than .01 was considered to indicate a significant difference.

Results

The average coefficient of variation for the ultrashort-TE bicomponent analysis of one patella sample in three repeated acquisitions was 4.64% for bound water fraction, 1.76% for free water fraction, 4.27% for bound water T2*, and 3.89% for free water T2*. These results show that the two-dimensional ultrashort-TE bicomponent analysis technique provides reliable estimation of relative fractions and T2*s of bound and free water components.

Both spin-echo and ultrashort-TE images of a healthy patella showed significant zonal variations (Fig 1). The T2 decay curves showed only a single component, with T2 decreasing from 64 msec for the superficial zone to 36 msec for the deep zone. The ultrashort-TE T2* decay curves showed two distinct components, with short T2* values increasing slightly from 0.76 msec for the superficial zone to 0.99 msec for the deep zone and long T2* values decreasing from 36 msec for the superficial zone to 24 msec for the deep zone. There was a monotonic increase in short T2* fraction from 15.2% for the superficial zone to 17.4% for the middle zone and 21.2% for the deep zone.

Figure 1:

Curves show single-component fitting of, A, CPMG images of a patella specimen with signal from ROIs drawn in the, B, superficial, C, middle, and, D, deep layers, as well as bicomponent fitting of, E, ultrashort-TE images of the same patella specimen with signal from the, F–H, same ROIs. B–D, T2 curves show a single-component decay behavior with a T2 of 63.53 msec for the superficial layer (B), 51.83 msec for the middle layer (C), and 36.02 msec for the deep layer (D). In F–H, there is an obvious short T2* component with a T2* of 0.78 msec for the superficial layer (F), 0.89 msec for the middle layer (G), and 0.99 msec for the deep layer (H), as well as a long T2* component with T2*s of 36.37, 30.47, and 24.13 msec for the respective three layers. Bound water fractions account for 15.2% of the ultrashort-TE signal for the superficial layer, 17.4% for the middle layer, and 21.2% for the deep layer. Zoom-in figures for short TEs are also shown in F, G, and H for bicomponent fitting. CI = confidence interval.

Cartilage degeneration displays considerable heterogeneity, and regional analysis is required for accurate correlation, as demonstrated by the histologic, PLM, and ultrashort-TE images, as well as bicomponent analysis of three ROIs with mild, moderate, and severe degeneration, respectively (Fig 2). Ultrashort-TE bicomponent analysis showed a significant increase in short T2* water fraction from 21.9% for mild OA to 28.5% for moderate OA and 36.0% for severe OA.

Figure 2:

A, Histologic, B, PLM, and, C, very-short–TE images of a patella specimen, as well as curves showing bicomponent analysis of, D, G, ROI 1 with mild degeneration (Mankin score 2, Vaudey score 0), E, H, ROI 2 with moderate degeneration (Mankin score 7, Vaudey score 2), and, F, I, ROI 3 with severe degeneration (Mankin score 12, Vaudey score 3). Arrows on A and B indicate ROIs. Ultrashort-TE bicomponent analysis shows a short T2* of 0.57 msec with a fraction of 21.9% and a long T2* of 60.94 msec with a fraction of 78.1% for ROI 1 (D), a short T2* of 0.65 msec with a fraction of 28.5%, and a long T2* of 53.33 msec with a fraction of 71.5% for ROI 2 (E), and a short T2* of 0.73 msec with a fraction of 36.0% and a long T2* of 43.86 msec with a fraction of 64.0% for ROI 3 (F). The residual signal (G, H, I) is less than 5% with bicomponent fitting, suggesting that the two-component model is well suited for ultrashort-TE T2* analysis of articular cartilage. CI = confidence interval.

The correlation between short and long T2* water fractions and the Mankin and Vaudey scores in 91 regional ROIs is shown in Figure 3. Short T2* water fraction was positively correlated with the Mankin (ρ = 0.66, P < .001) and Vaudey (ρ = 0.68, P < .001) scores. Free water fraction was negatively correlated with the Mankin (ρ = -0.68, P < .001) and Vaudey (ρ = -0.69, P < .001) scores. Short T2* water fraction was similar for patellae with grades 1 and 2 degeneration but was significantly increased for patellae with grades 3 and 4 degeneration. There was an increase in short T2* water fraction with the Vaudey score. Short T2* water fraction was significantly correlated with the Mankin score in all three zones across articular cartilage, with the highest correlation in the superficial zone (ρ = 0.68, P < .001) and the lowest correlation in the deep zone (ρ = 0.59, P < .001). High correlation was also apparent between short T2* water fraction and the Vaudey score, with the highest correlation in the superficial zone (ρ = 0.66, P < .001) and the lowest correlation in the deep zone (ρ = 0.55, P < .001).

Figure 3:

Graphs show correlation between short and long T2* water fractions with the, A, B, Mankin score and, D, E, Vaudey score, as well as the relationship between short T2* water fractions and, C, histologic grading and, F, Vaudey PLM score. Short T2* water fraction is positively correlated with the Mankin (ρ = 0.66, P < .001) and Vaudey (ρ = 0.68, P < .001) scores, while long T2* water fraction is negatively correlated with the Mankin (ρ = −0.68, P < .001) and Vaudey (ρ = −0.69, P < .001) scores. There is a steady increase in short T2* water fraction with histologic grading and Vaudey score.

The regional correlation between T2, short and long T2* values, and the Mankin and Vaudey scores is shown in Figure 4. There was a low correlation between T2 and the Mankin (ρ = 0.28, P = .123) and Vaudey (ρ = 0.18, P = .16) scores and a low correlation between short T2* and the Mankin (ρ = 0.24, P = .14) and Vaudey (ρ = 0.22, P = .14) scores. However, there was a significant negative correlation between long T2* and the Mankin (ρ = -0.55, P < .01) and Vaudey (ρ = −0.55, P < .01) scores.

Figure 4:

Graphs show correlation between, A, T2, B, short T2*, and, C, long T2* and the Mankin score, as well as correlation between, D, T2, E, short T2*, and, F, long T2* and the Vaudey PLM score. The Mankin score is weakly positively correlated with T2 (ρ = 0.28, P = .13) and short T2* (ρ = 0.24, P = .14) and significantly negatively correlated with long T2* (ρ = −0.55, P < .01). The Vaudey score is weakly positively correlated with T2 (ρ = 0.18, P = .16) and short T2* (ρ = 0.22, P = .14) and significantly negatively correlated with long T2* (ρ = −0.55, P < .01).

Cartilage shows significant magic angle effect on the T2 relaxation times (Fig 5). At later echoes, signal from ROI 2 (near the magic angle) was significantly higher than that from ROI 1. T2 was 79.27 msec for ROI 2 and 30.68 msec for ROI 1, corresponding to an increase of 158%, which is much more than the typical T2 increase due to degeneration. The ultrashort-TE images also showed significant apparent magic angle effects on their T2* values. However, short T2* water fraction, 20.7% for ROI 1 and 18.8% for ROI 2, appeared much less sensitive to the magic angle effect.

Figure 5:

CPMG images show magic angle effect in T2 analysis of a patella sample with normal, A, histologic and, B, PLM findings, as well as, C–J, CPMG images with TEs of 12–96 msec, and single-component T2 fitting of two ROIs, K, ROI 1 and, L, ROI 2. The ROIs show a dramatic T2 increase from 30.68 msec for ROI 1 to 79.27 msec for ROI 2, although both ROIs have the same normal Mankin scores of 1 and Vaudey scores of 0. CI = confidence interval.

The correlation between short T2* water fraction and cartilage degeneration may be complicated because of the complex structural changes. Figure 6 shows a patella with a Mankin score of 12 and a Vaudey score of 4, indicating severe degeneration. However, ultrashort-TE bicomponent analysis showed a short T2* fraction of only 18.7%. Histologic analysis showed significant structural degradation, with clefts across the whole thickness.

Figure 6:

A, Histologic, B, PLM, and, C, CPMG images and, D, single-component analysis of the ROI shown, as well as, E, ultrashort-TE image and, F, bicomponent analysis curves of the same ROI. Green outline on B is the same ROI as on A. CPMG single-component analysis shows a T2 of 92.4 msec, indicating a high degree of degeneration. Ultrashort-TE bicomponent analysis showed a short T2* of 0.90 msec with a fraction of 18.7% and a long T2* of 44.72 msec with a fraction of 81.3%. A high long T2* water fraction was observed in this patella sample, with severe degeneration probably caused by water residing in clefts across the whole cartilage thickness. CI = confidence interval.

The Mankin and Vaudey scores were highly correlated (ρ = 0.93, P < .001); however, there was significant overlap between different grades within the two scoring systems (Fig 7).

Figure 7:

Graphs show correlation between the, A, Mankin and Vaudey scores and between, B, histologic grade and Vaudey score. The Mankin score is highly correlated with the Vaudey score (ρ = 0.93, P < .001). However, there is substantial overlap between different grades within the two scoring systems.

Discussion

Our study findings indicated significant differences in short and long T2* water fractions between patellae with healthy or mild degeneration and patellae with moderate or severe degeneration. Short T2* fraction was significantly correlated with histopathologic and PLM assessments and probably reflects microstructural changes in collagen matrix. Long T2* values showed a significant negative correlation with the Mankin and Vaudey scores, and this correlation is consistent with the results reported by Williams et al (15). Both short T2* and T2 values were poorly correlated with the Mankin and Vaudey scores.

The two forms of water in articular cartilage, bound and free water, are the subject of intense discussion in the literature. Mankin and Thrasher (4) reported that cartilage with OA had 12% higher bound water fraction than normal cartilage. Studer et al (28) found a prominent bright halo around fibrils under electron microscopic examination, indicating a distinct water-rich layer surrounding collagen fibrils in articular cartilage. Recent nuclear MR spectroscopy of simple collagen-water systems shows that water molecules exist either as bulk water, which does not interact with collagen, or as interior hydration water molecules (9,29). The existence of bound and free water pools has also been confirmed by findings in recent cross-relaxation imaging experiments (30), as well as results from MR spectroscopy studies and our own study (10,16).

With the clinical CPMG sequence, a minimal TE of 12 msec, which is too long to evaluate water components bound to PG and collagen, was used, and therefore, a single-component exponential T2 fitting was assumed. The principal confounding factor with T2 measurement is likely to be the magic angle effect. Dipolar interactions become zero when collagen fibers are oriented at 55° or 125° to B₀, and this factor may result in a severalfold increase in T2 (31–33). This increase may exceed the change produced by OA and lead to a low correlation between T2 and the Mankin and Vaudey scores. It is likely that higher correlation between T2 and histopathologic findings could have been achieved if the samples were rotated 90°, with the apex parallel to the B₀ field (more fibers aligned to B₀ and less magic angle effect). The poor correlations between T2 and the Mankin and Vaudey scores represent the worst case scenario based on angular orientation of the tissue. Fiber orientation is an important consideration when using T2 to evaluate OA (31–33).

Water concentration has been proposed as a sensitive biomarker of OA (4,6–8). However, most of the MR studies of water concentration have been based on spin-echo sequences, with which TEs are used that are too long to aid evaluation of short T2 water components (6–8). Ultrashort TE bicomponent analysis allows both short and long T2* water components to be evaluated. Short and long T2* water components have more than an order of magnitude difference in relaxation times, and this improves fitting robustness. Articular cartilage has been reported to have 6% water bound to collagen, 14% bound to PGs, and 80% free water (10). The ultrashort-TE bicomponent analysis showed similar results, with the 15%–20% water with short T2* found in healthy patellae, which corresponds to the sum of the components bound to collagen and PGs. It is technically difficult to separate these two bound water components by using clinical MR systems because of signal-to-noise ratio limitations and the small difference in their T2*s.

Our preliminary results show that the short T2* fraction gradually increases from the articular surface to the deep zone in healthy patellae. This finding may be partly explained by the gradual increase in concentration of PGs (34). Furthermore, the collagen fibrils near the articular surface are of much smaller diameter and more closely packed (35), which reduces the surface area available for water binding and, therefore, leads to a reduction in bound water fraction near the articular surface. The short T2* water fraction increased with cartilage degeneration. Articular cartilage contains numerous collagen fibrils, which bind PG into a structural gel that traps water (4,36). The individual collagen fibrils within the bundles are of approximately uniform diameter of approximately 60 nm. Numerous collagen fibrils are bundled together to form bundles in cartilage with diameter of approximately 50 mm. Degeneration results in loss of PGs, with reduction in PG-bound water. More advanced OA is also associated with a loss of collagen content. However, disorganization of the collagen matrix may result in a significant increase in collagen fibrils’ surface areas available for water binding, which may more than offset the loss of water bound to PGs and lost collagens, and lead to a net increase in total bound water fraction (4). Further work is needed to validate this assumption. Our results show very similar bound water fractions between histologic grades 1 and 2, suggesting its reduced sensitivity to early degeneration.

There were several limitations of this study. First, we have no direct validation that the short T2* component corresponds to bound water and that the long T2* component corresponds to free water. Ultrashort-TE bicomponent analysis of tritiated cartilage may provide such information (4). Second, the ultrashort-TE sequence is subject to eddy currents, which could distort the T2* decay curve and lead to errors in bicomponent fitting (37,38). Third, the reference standards we used, Mankin and Vaudey scores, have their own limitations. For the histopathologic assessment, we chose the Mankin score because it is the most widely used system (18,20). It mainly is used to detect changes in morphologic and histochemical characteristics, and cartilage volume and bone changes are not considered (18). In addition, the Mankin grading system includes as a criterion safranin O staining intensity, which is affected by protocol variability, such as tissue fixation and decalcification, and these factors could lead to errors in the measure (39,40). Fourth, there is a discrepancy in the image scales between histologic sections that are a few micrometers in thickness and MR images that are several millimeters thick. We observed a higher discrepancy in the Mankin score compared with the Vaudey score. This discrepancy may be because the Mankin score includes parameters that are not considered in the Vaudey score (18,19), which mainly reflects collagen matrix and surface structure. In the Mankin score, surface structure accounts for only approximately 25% of the total score (reflecting surface structure, cells, safranin O stain intensity, and tidemark). Each parameter has subcategories, and the scores are summed to provide the total score (18). A sample with a high-quality surface structure (ie, low Vaudey score or relatively normal) may receive a high Mankin score (relatively abnormal) because of abnormal cellularity or poor safranin O staining. In addition, histopathologic and PLM analysis are performed on adjacent slides. This technique may lead to further inconsistencies when microscopic changes are present. Last, the ROIs were chosen by one person, who was blinded to the ultrashort-TE MR imaging findings, and were registered between MR imaging and histologic sections through visual assessment by using landmarks. There is potential for inconsistency in data analysis introduced through this method.

In conclusion, we showed that ultrashort-TE bicomponent analysis can be performed with a clinical MR imaging unit to evaluate short T2* (bound water) and long T2* (free water) components in articular cartilage. Short T2* water fraction parallels cartilage degeneration, while CPMG T2 values show a low correlation with degeneration.

Advances in Knowledge.

• • Ultrashort–echo time (TE) MR technique with bicomponent analysis can be used to evaluate short and long T2* water components in patellae by using a clinical 3-T MR system.

• • The short T2* fraction is correlated significantly with the histopathologic Mankin (ρ = 0.66, P < .001) and polarized light microscopic Vaudey (r = 0.68, P < .001) scores.

Implication for Patient Care.

• • Ultrashort-TE bicomponent analysis can be used in clinical systems to evaluate short and long T2* water components in articular cartilage; changes in bound and free water may provide an early surrogate marker of degeneration.

Disclosures of Potential Conflicts of Interest: C.P. No potential conflicts of interest to disclose. W.C.B. No potential conflicts of interest to disclose. M. Lee No potential conflicts of interest to disclose. M. Lotz No potential conflicts of interest to disclose. G.M.B. No potential conflicts of interest to disclose. D.L.D. No potential conflicts of interest to disclose. C.B.C. Financial activities related to the present article: none to disclose. Financial activities not related to the present article: institution received a grant from GE Healthcare. Other relationships: none to disclose. J.D. No potential conflicts of interest to disclose.