Articular Cartilage

The ability of MR to noninvasively assess cartilage is one of more important advances in orthopaedic imaging. Many pulse sequences have been developed to detect cartilage (Table 7-1). The “ideal” cartilage pulse sequence provides differential contrast among synovial fluid, fibrocartilage (such as meniscus and labrum), and hyaline cartilage, combined with relatively high in-plane and slice resolution. The ability to detect changes in the adjacent subchondral bone is also clinically useful. It is essential to note, however, that traditional T1- and T2-weighted images are not suitable for the assessment of cartilage. Indeed, the use of traditional pulse sequences with poor in-plane resolution results in ineffective visualization of articular cartilage that has been proven inaccurate compared to arthroscopic inspection and correlates poorly with clinical evaluation.^{11,12,13,14,15} Image characteristics on traditional T1- and T2-weighted sequences include the following:

Three-dimensional T1-weighted gradient-echo imaging was one of the first techniques to provide cartilage-sensitive imaging that was considered to be accurate based on comparison with arthroscopic findings. Disler et al. have shown that fat-suppressed T1-weighted 3D gradient-echo sequences (T1-weighted 3D spoiled gradient echo) were considerably more sensitive than standard MR imaging: 93% versus 53% in one study¹⁶ and 75% to 85% versus 29% to 38% in another.¹⁷ These authors also measured interobserver agreement, crucial for MR imaging to be used as a substitute for diagnostic

arthroscopy in cartilage evaluation, and found that images acquired using the gradient-echo sequences provided excellent interobserver and intraobserver agreement (kappa = 0.72 – 1.0), with the exception of the lateral tibial plateau, which showed moderate to substantial agreement (kappa = 0.56 – 0.72). The more convex architecture of the lateral tibial plateau may make evaluation in this location subject to partial volume errors and reduced accuracy.¹⁸

Fat-suppressed T1-weighted gradient-echo techniques provide high contrast between bone and cartilage but are not fluid-sensitive. Three-dimensional gradient-echo sequences with high contrast boundaries between bone and cartilage are more amenable to semiautomatic segmentation algorithms, thus providing reproducible assessment of cartilage thickness and volume (Fig. 7.4).^19,20

The ability to perform segmentation analysis of gradient-echo images makes these pulse sequences suitable for monitoring disease progression over time and assessment of cartilage volume in osteoarthritis (OA). The chosen pulse sequence parameters should be based on a suitable standard and should have high precision or reproducibility.²¹ Routine, standardized cartilage imaging for clinical assessment is generally performed with a slice resolution between 3 and 4 mm and an in-plane resolution between 300 and 350 microns. However, gradient-echo sequences necessary to quantify cartilage often need higher resolution, requiring a slice thickness on the order of 1 to 2 mm, at a pixel size of less than 300 microns.²¹

Attention to echo time is also important. For fat-suppressed T1-weighted gradient-echo sequences, given the relatively short T2 relaxation times of cartilage, fairly short echo times (preferably 10 msec) are generally recommended. As Eckstein et al. reported, an echo time of 11 msec led to underestimation of tibial cartilage thickness (compared to CT arthrography as the standard).²²

With standardized modern gradient platforms, a cartilage volume technique in a single plane may be acquired in less than 13 minutes. As a result of partial volume effects encountered in tomographic imaging of a curved articular surface, more accurate thickness assessment is obtained if computations are performed in all three planes.²¹ Unfortunately, universal application of such techniques is limited by the extensive time required for optimal scan acquisition and post-processing. In an attempt to find a method to reduce post-image processing time, Zhai et al. compared assessment of a whole slab of 1.5-mm-thick slices with measurement of every second to fourth slice.²³ Using the inclusive slice estimate as the standard, the authors found that decreasing the number of slices by extracting one in two to one in four slices led to little over- or underestimation in cartilage volume over the plateau and patella (femoral cartilage volume was not assessed).²³ Others have used a double echo steady state (DESS) gradient-echo sequence at 3 Tesla to analyze cartilage morphology in the femorotibial joint.²⁴ The inherently higher signal-to-noise ratio in this protocol permitted thinner slices (0.7 mm) and accurate analysis of cartilage morphology.

Assessment of cartilage volume is dependent on thickness and surface area. Therefore, longitudinal changes are affected by alterations in cartilage thickness (elevation or reduction) as well as alterations in cartilage surface area (osteophyte formation). Due to these variables, standardization of cartilage segmentation techniques is needed, with strict attention to pulse sequence parameters and in-plane and slice resolution. To obtain cartilage volume, segmentation must be performed either manually (a lengthy, time-consuming process) or by semiautomatic algorithms based on computerized recognition of tissue contrast edge boundaries. These techniques often delineate the line of the steepest gradient between the hyperintense cartilage and the hypointense (fat-suppressed) bone, as well as between the hyperintense cartilage and the hypointense meniscus or capsule.²⁵ Some correction at low-contrast interfaces, as well as the application of smoothing algorithms after segmentation, may be required. Although some degree of expertise is necessary to maintain reproducibility, in general, automatic segmentation provides superior reproducibility in the assessment of cartilage volume and thickness.²⁶

In a study of cartilage segmentation prior to knee arthroplasty, Graichen et al. compared results from quantitative MR imaging with those obtained by direct image analysis of surface area, cartilage thickness, and volume of the explants. Quantitative MR imaging provided accurate assessment of OA, with correlation coefficients ranging between 0.92 for thickness and 0.98 for volume.²⁵ Segmentation of the central and weight-bearing areas of the femoral condyles have been shown to be more accurate than analysis of boundary and non-weight-bearing regions.²⁷ Others have studied lesions at varying field strengths. Comparing 1.5- to 3-Tesla units and using a gradient peak method to define focal cartilage, no statistically significant differences were found between the two field strengths or in the accuracy of the measurements of depth, diameter, area, and focal thickness compared to manual measurements.²⁸

Quantitative MR imaging is an attractive addition to the longitudinal study of OA, providing noninvasive assessment of cartilage morphology and/or response to pharmaceutical intervention. Comparing radiographic evidence of osteoarthritis, including joint space narrowing and osteophytes, with segmentation techniques, Cicuttini et al. noted a strong negative linear association between tibial cartilage volume and increasing joint space narrowing.^29,30 More recently, evaluation of patients with knee OA using quantitative MR imaging and joint space width on radiographs showed a modest but significant correlation between joint space and medial tibiofemoral cartilage volume; no correlation in longitudinal changes over a 2-year follow-up period, however, was found.³⁰ The authors suggested that 3D assessment of cartilage volume may be superior to 2D assessment using radiographs.³⁰

The use of quantitative MRI in the longitudinal evaluation of osteoarthrosis may be affected by transient deformation in cartilage as a response to load (which affects cartilage volume).³¹ In one study a diurnal variation in cartilage thickness over the patellofemoral and femorotibial contact zones has been noted during a day of standing activity.³² There is also a statistically significant decrease in patella cartilage volume in response to physical exercise, suggesting that a period of rest may be indicated prior to imaging of subjects in a longitudinal assessment.³³

Although the study of mechanical deformation of cartilage in normal and abnormal joint loading conditions is of interest, the relationship between alterations in overall cartilage volume and function remains unclear. It has been suggested that because of variations in the relative size of subchondral bone, cartilage volume alone may not be suitable as a criterion for patient evaluation, and that a ratio of cartilage volume to bone interface may prove a more suitable measure.²¹ In a 2-year study of patients with knee arthritis, Raynauld et al. found that there was no statistical correlation between loss of cartilage volume on MR imaging with radiographic changes, or between changes in cartilage volume and alterations in clinical variables, including components of the Western Ontario and McMaster University Osteoarthritis Questionnaire (pain, stiffness, and function) and the “Short Form 36” (SF36) health quality assessment technique.³⁴

Although suitable for volumetric assessment, one of the pitfalls of imaging with gradient-echo techniques is susceptibility to image degradation in the presence of orthopaedic instrumentation or metallic debris following diagnostic arthroscopy. Fast spin-echo techniques, with the ability to directly visualize articular cartilage despite adjacent hardware, have proven more useful in the setting of instrumentation.³⁵ One of the greatest advantages of fast spin-echo imaging of articular cartilage is the ability to generate images suitable for detection of subchondral bone, ligaments, and the surrounding soft-tissue

envelope, providing more efficient use of imaging time. Fast spin-echo images are cartilage-sensitive due to an inherent magnetization transfer contrast and exchange of off-resonance magnetization between slices, the net effect of which is to saturate the bound pool of hydrogen nuclei, resulting in a decrease in signal intensity from the free pool.³⁶ This exchange results in relatively high signal intensity from fluid compared to the lower signal intensity of articular cartilage, providing an effective differential contrast among menisci, articular cartilage, and fluid in the meniscosynovial recesses (Fig. 7.5). With arthroscopy as a standard, Potter et al. found that articular surface evaluation using spin-echo sequences had a sensitivity of 87%, a specificity of 94%, and an accuracy of 92%, with minimal interobserver variability (kappa = 0.93).¹⁸ These findings of reproducibility support the use of MR imaging as a noninvasive and objective outcome measure of surgically manipulated cartilage and in the longitudinal evaluation of traumatic and degenerative cartilage lesions.

Fast spin-echo sequences use a relatively long repetition time (TR) (on the order of 3,500–5,000 msec) with a moderate to intermediate echo time (TE) (30–34 msec) for central K-space filling. High in-plane resolution (on the order of 250–350 μ in the frequency direction) imparts superior spatial resolution, which can make partial-thickness lesions more conspicuous compared to the gradient-echo techniques (Fig. 7.6). An essential element is the use of a wider receiver bandwidth

to minimize chemical shift misregistration as well as to reduce inter-echo spacing. Decreasing inter-echo spacing will minimize the edge blurring routinely encountered with fast spin-echo techniques performed at correspondingly narrower bandwidths with wider inter-echo spacing. The effect of chemical shift causes a misregistration at the interface between fat (subchondral bone) and water (cartilage), resulting in factitious loss of the subchondral plate, abnormal signal hyperintensity in the adjacent cartilage (due to the frequency shift summation), and inaccurate assessment of cartilage morphology (Fig. 7.7).

The chemical shift effect may be minimized by the use of a wider receiver bandwidth and can be largely eliminated by the use of fat-suppression techniques. In a study comparing arthroscopy findings with coronal and axial moderately T2-weighted fast spin-echo images with frequency-selective fat suppression, Bredella et al. reported that MR imaging has a sensitivity of 94%, a specificity of 99%, and accuracy of 98%.³⁷

Fast spin-echo techniques are well suited to discernment of both intrasubstance changes in the cartilage based on the gray-scale contrast and morphologic changes that may be correlated with pathologic and arthroscopic inspection using standardized grading scales (Table 7-2).³⁸

Additional pulse sequences have also been recommended for evaluation of cartilage, most of which are based on modified gradient-echo techniques that yield good fluid-to-cartilage contrast. Examples include driven equilibrium Fourier transform (DEFT) imaging, double echo steady state (DESS) imaging, refocused steady state free precession (SSFP), water selective balanced steady state free precession pulse sequences (WS-bSSFP), and projection reconstruction imaging:

Like ¹H, ²³Na is a suitable nucleus for MR imaging. The presence of negatively charged glycosaminoglycan in cartilage generates an attraction toward the positively charged ²³Na, allowing for relative measurement of fixed charge density in cartilage.^59,60 Although sodium MR imaging is not in widespread clinical use, the ability to detect early OA changes as a result of proteoglycan loss may allow earlier and more accurate diagnosis of disease progression.

Reddy et al., evaluating knee cartilage in human volunteers as well as in enzyme-degraded cartilage explants,⁶¹ found decreased signal intensity on the ²³Na images of the enzyme-degraded specimens and suggested that this reduced signal reflected loss of proteoglycan from the tissue. The clinical feasibility of ²³Na imaging was also demonstrated in the volunteer group.⁶¹ In another study of enzyme-degraded cartilage explants, Borthakur et al. compared the results of ²³Na MR imaging to spectrophotometric assay-determined proteoglycan concentrations and noted that the signal change on ²³Na imaging correlated well with the observed proteoglycan loss, suggesting that ²³Na MR imaging was both sensitive and specific in the detection of proteoglycan concentration.⁶² Interestingly, there was no correlation between proteoglycan loss and changes in proton T1 and T2 relaxation times.⁶² Wheaton et al. studied fixed charge density maps calculated from ²³Na MR imaging in both healthy volunteers and patients with clinical symptoms of early-stage OA. Focal areas of decreased fixed charge density could be found in those with symptomatic OA, suggesting the loss of proteoglycan from the matrix.⁶³

The lower signal-to-noise noted in ²³Na imaging compared to proton MR imaging reflects the natural lower abundance of ²³Na.⁶³ Requirements for special transmit and receive coils, as well as relatively long scan times, may limit its application to clinical cartilage imaging.^64,65

There has been increased interest in the recent literature in the use of T1 rho (T1ρ) imaging to detect an alteration in fixed charge density and regional variation of proteoglycan content. The spin-locking pulse sequence uses a cluster of RF pulses:

T1ρ or spin-lattice relaxation in the rotating frame has been used to investigate low-frequency interactions between hydrogen nuclei in macromolecules and free water.⁶⁶ Although incorporation of a spin-lock pulse into clinical MR imaging protocols has been suggested, it has not become common because of concern over the increased tissue specific absorption rate (SAR) associated with these special pulses.⁶⁷ SAR is increased because T1ρ imaging requires more RF power than conventional pulse sequences. Unlike spin-spin relaxation (T2), however, which is inherently susceptible to local field inhomogeneities and susceptibility variations, the effect of the spin-locking pulse minimizes spin dephasing and thus reduces the effect of regional inhomogeneities and susceptibility artifact.⁶⁶ The spin-locking pulse promotes phase coherence along the locked spins during contrast evolution time, thus making these images potentially less susceptible to artifact.⁶⁷

Correlation of T1ρ with fixed charge density has been performed in both enzyme-degraded bovine and osteoarthritic human explants. Plots of normalized T1ρ rates were found to be strongly correlated to fixed charge density and to corresponding loss of proteoglycan, as confirmed by histologic tissue inspection.⁶⁸ Some investigators, however, have questioned the specificity of T1ρ in the detection of proteoglycan alterations. In molecular suspensions without a matrix, both T2 and T1ρ demonstrate an exponential decrease with increasing collagen and glycosaminoglycan concentration, with the effects of collagen dominating.⁶⁸

T1ρ has been studied in an animal model in which OA was induced by intra-articular injection of recombinant interleukin-1β (IL-1β), causing degradation of proteoglycan via increasing expression and activation of matrix metalloproteinases (MMPs).⁶⁹ In this animal model, the T1ρ relaxation rate (R ₁β_ρ = 1/T1ρ) of the IL-1β–treated patellae was, on average, 25% lower than that found in control (saline-injected) patellae; fixed charge density (as measured via correlative ²³Na MR imaging) was also found to be reduced by an average of 49%.⁶⁹ These studies demonstrate the potential application of noninvasive MR imaging techniques in the assessment of matrix alteration and provide further support for the clinical use of MRI in the study of degraded cartilage and the eventual development of tissue-engineered cartilage.

In a clinical setting, studies have shown the feasibility of performing T1ρ measurements in cartilage without exceeding the FDA limits of SAR. In one study, images obtained with fat suppression at 1.5 Tesla were compared to standardized T2-weighted images.⁷⁰ Although findings suggested that T1ρ is more effective than T2 in highlighting chondral abnormalities,⁷⁰ the T2-weighted sequence used was not an optimized cartilage technique. Additional studies have shown that T1ρ imaging at 1.5 Tesla is feasible and that T1ρ relaxation times are significantly elevated in symptomatic compared to asymptomatic patients.⁷¹ Additional studies will be necessary to gain greater clinical acceptance of these techniques. In addition, ready availability of the spin-locking RF pulses for clinical scanners is needed before T1ρ imaging can be widely used.

Articular cartilage structure and fixed charge density can also be assessed using gadolinium contrast. In this technique, the negatively charged gadolinium salts diffuse from the synovial fluid into the cartilage as a function of regional fixed charge density, with increasing concentration of negatively charged gadolinium salts in areas of relatively depleted negatively charged proteoglycans. MR imaging at 8.4 Tesla and sodium spectroscopy in intact, enzyme-degraded, and IL1-induced cartilage degeneration showed good correlation with histologic evidence of proteoglycan depletion,⁷² and there are differences in T1-weighted and T1-calculated images in the presence of negatively charged gadolinium agents, but not in the presence of a neutral (nonionic) contrast agent.⁷³

Clinical protocols for cartilage evaluation involve the intravenous injection of 0.2 mmol/kg gadopentetate dimeglumine, followed by mild exercise for 10 minutes to facilitate transport of contrast into the cartilage, followed by an 80-minute wait period. A series of T1-weighted images are then acquired with a fast inversion recovery pulse.⁷⁴ This technique, known as delayed gadolinium-enhanced MRI of cartilage (dGEMRIC), creates an index of relative glycosaminoglycan distribution (the T1_Gd index) (Fig. 7.11). The T1_Gd index can be used to evaluate OA and other disorders involving articular cartilage.^74,75,76 Reproducible findings can be found for standardized regions of interest (ROIs), when a large ROI is drawn over the weight-bearing segment of the condyles, using the meniscus in the sagittal plane as a guide.⁷⁵ The dGEMRIC index was found to have a higher correlation with clinical symptoms on the Western Ontario and McMaster University Osteoarthritis Questionnaire than the presence of arthrosis on radiographs. This finding suggests that MR imaging techniques are more suitable than secondary bony changes and joint space narrowing on radiographs in assessing the magnitude of OA.⁷⁶

T2 mapping can be used to reflect the collagen component of the extracellular matrix. T2 mapping is a well-described MR imaging technique that reflects the internuclear (spin-spin) dephasing that occurs as a result of transverse relaxation of the excited hydrogen nuclei. To a large extent, the stratification of gray scale seen in fast spin-echo imaging of articular cartilage reflects the orientation of the collagen in the extracellular matrix. The spatial variation of cartilage T2 follows the architectural arrangement of the collagen fibril network as assessed by polarized light microscopy,⁷⁷ much as MR microscopy zones based on T2 profiles are based on collagen fiber orientation.⁷⁸ In addition, there is no observed change in regional T2 in areas of mechanically undamaged cartilage with decreased proteoglycan content.⁷⁹

Quantitative assessment of T2 relaxation time is also influenced by the structural anisotropy of collagen within the matrix. In highly ordered structures such as normal tendons and ligaments, the diminished signal intensity is largely due to enhanced dipole–dipole interactions, with short T2 relaxation. The magic angle effect in articular cartilage describes the mathematical relationship between the spinning hydrogen nuclei within the collagen component of cartilage and the long axis of the magnetic field, which in a closed unit runs

parallel to the long axis of the patient.⁸⁰ When the angle between the external field and the hydrogen nuclei in collagen reaches approximately 55°, there is a corresponding prolongation of T2 relaxation time according to the relationship 1/T2 = k(3cos²&thetas; – 1). In the radial zone, where the collagen is highly ordered, expected prolongation of T2 has been noted in both high field strength microscopy systems at 7 Tesla as well as in images acquired on 1.5-Tesla units.⁸⁰ Recognition of the magic angle effect in cartilage is important since it may produce apparent increased signal intensity in the radial zone at 55° (Fig. 7.12). Since the magic angle effect may hamper the measurement of cartilage regions of interest, caution should be used to avoid sampling T2 values at the magic angle.

Regional differences in cartilage T2 have also been noted. In one study of cartilage T2 as a function of orientation to the magnetic field, the largest variation in T2 was reported to have occurred in the transitional rather than the radial zone.⁸¹ These regional variations in T2 may be due to differences in the degree of cartilage compression, with resulting alteration in collagen fiber orientation, causing an attenuation of the magic angle effect.^81,82 Others have noted increased T2 in both the radial and transitional zones at 55°, suggesting that the collagen fiber orientation in the transitional zone was not entirely random.⁸³ Goodwin et al. noted a complex, 3D joint geometry in which there was a continuous macrostructure radiating from the subchondral bone into the plane of the joint surface, similar to that noted on fracture sectioning, and suggested that this structure could account for the regional differences in T2 (Fig. 7.13).⁸⁴ The bulk curvature of cartilage also appears to create a depth-dependent variation in fiber orientation, and T2 differences may be related to the finding that the transitional layer appears to be much thicker in the submeniscal zone of tibial plateau compared to the central portion of plateau, where there is an apparent increase in the thickness of the radial zone.⁸⁴ In evaluating T2 mapping obtained from different areas of a joint, it is important not to mistake longer T2 values that reflect regional variation in collagen macrostructure for prolonged T2 in the setting of cartilage damage.^36,84

The method by which quantitative data is attained can also affect measurements.^85,86 Although multiecho pulse sequences are more efficient in obtaining individual echo information, variations in the 180° refocusing pulses may generate both T1 and T2 contrast. Care should be taken, therefore, when using a multiecho, multislice sequence in which there are both stimulated echo contributions in the slice direction (increasing measured T2 values) and magnetization transfer between slices (decreasing measured T2 values).⁸⁵ Chemical shift misregistration through the interface between fat (bone) and water (cartilage) can also generate errors in quantitative measurements.

In the evaluation of OA, T2 mapping can provide information about cartilage permeability and breakdown in collagen structure.^87,88 When collagen is destroyed, hydrogen dipoles become more mobile and T2 values increase, resulting in corresponding loss of stratification (Fig. 7.14). Using MR imaging with a spin-echo sequence at 1.5 Tesla, David-Vaudey et al. confirmed that increased T2 is correlated with histologic degeneration of cartilage.⁸⁹ In studies of OA using

a 7-Tesla magnet, several changes, including a decrease in superficial zone thickness, an increase in total cartilage thickness over the submeniscal regions, and a shift in the maximum T2 of the articular surface were found. These findings indicate that the collagen of the superficial zone is disrupted in early stages of OA.⁹⁰ Further, the increasing T2 of the superficial zone was detected in OA specimens obtained from portions of the tibia covered by meniscus, suggesting that these regions may show earlier signs of cartilage degeneration, before it is evident in other (more central) areas of the joint. These regional differences are most likely due to alteration in structural properties related to weight-bearing areas of the joint.^84,90

Changes in Young—s modulus (mechanical properties) as well as alterations in polarized light microscopy correlate well with T2 mapping.⁹⁰ When the relationship of T2 relaxation time and dGEMRIC is compared to static and dynamic compressive moduli, there is a statistically significant correlation

between MRI and mechanical parameters, especially between T2 and Young—s modulus.⁹¹ Similarly, it has been shown that prolongation of T2 occurs early following meniscectomy, correlating with observable biomechanical changes using indentation probe testing as well as polarized light microscopy.⁹²

The use of T2 mapping techniques may prove helpful in detecting changes in the structure of cartilage prior to significant loss of cartilage thickness or gross signal alterations in the cartilage gray scale, potentially optimizing the timing of surgical procedures, such as meniscal transplantation and patellofemoral realignment, which are aimed at delaying the progression of OA (Fig. 7.15). In a study comparing subjects who were considered healthy, as having mild OA, or as having severe OA, T2 mapping was combined with anatomic gradient-echo models of the femorotibial joint, allowing for comparison of cartilage T2, volume, and thickness.⁹³ Although no significant difference was found between patients with mild and severe OA, all cartilage compartments except the lateral tibial plateau showed a significant increase in T2 values between healthy subjects and OA patients.⁹³

T2 mapping may also be useful in the evaluation of patients with chondromalacia patellae. Spatial variation in T2 has been found as a function of both age and symptoms, with age being associated with an asymptomatic increase in T2 within the transitional zone of cartilage.⁹⁴ An age-dependent increase in T2 was not observed in the radial zone, which the authors attributed to the higher tissue anisotropy of this region.⁹⁴

One of the most popular resurfacing techniques is microfracture, which entails drilling very small holes (“microfractures”) into the subchondral bone beneath the denuded cartilage. The technique prompts release of multipotential stem cells from the marrow, which create a covering of largely reparative fibrocartilage with smaller elements of hyaline-like cartilage over the defect (Fig. 7.23).¹¹⁰ The best results are correlated with good MRI fill grade, low body mass index (BMI), and short duration of preoperative symptoms.¹¹¹

The signal characteristics of microfracture largely reflect the time interval between repair and MR imaging. In the initial postoperative period, the signal characteristics are largely hyperintense to native cartilage, consistent with increased mobility of water due to a less organized matrix.^111,112,113 In addition to the early appearance of hyperintense repair

cartilage, Mithoefer et al. noted an underlying bone marrow edema pattern, followed by progressive filling and a decreasing bone marrow edema pattern.¹¹¹ Overgrowth of subchondral bone has also been described following microfracture,^111,112 resulting in corresponding thinning of the overlying repair cartilage as well as the potential for impaction of the opposing surface of the joint during active loading. Although osseous overgrowth is a not uncommon finding, it does not correlate with a statistically significant worsening in clinical outcome.¹¹¹ The significance of subchondral bone overgrowth is not yet certain, but it might reflect excessive removal of the subchondral bone during debridement and removal of the calcified cartilage layer, thus providing a stimulus for enchondral ossification (Fig. 7.24).¹¹⁴ Postoperative peripheral integration may be incomplete, and discernible fissures can be seen in the majority of patients on postoperative MR imaging.¹¹¹

T2 mapping techniques can be used to assess cartilage repair. In our experience, T2 mapping obtained following microfracture has markedly prolonged T2 relaxation times compared to the adjacent and opposite hyaline cartilage (Fig. 7.25).

The term OATS, originally used as an acronym for osteochondral autograft transplantation or transfer system, is used to refer not only to autografts (autograft OATS) but also to osteochondral allografts (allograft OATS) and to transplantation with tissue scaffolds (scaffold OATS). In autograft OATS, one or more plugs are harvested from a “less important” portion of the joint (typically the intercondylar notch or far anterior margin of the condyles) and transferred into a defect over the weight-bearing margin of the joint. Autografts require careful operative planning to duplicate the geometry of the joint, and they also create some donor-site morbidity. In osteochondral allograft transfers (allograft OATS or OCA), cadaveric bone–cartilage plugs are implanted into a defect. They are most suitable for very large defects, such as those created by areas of osteonecrosis with collapse or by osteochondritis dissecans.

Although assessment of patient outcome, as judged by pain and function scores, is clinically relevant, objective evaluation of repair provides insight into the natural history of cartilage resurfacing techniques. Traditionally, objective assessment was made by histologic evaluation of biopsy specimens.¹¹⁵ MR imaging, however, is extremely well suited to noninvasive evaluation of cartilage repair, since it not only displays the 3D geometry of joints with high in-plane as well as slice resolution but also demonstrates superior soft-tissue contrast. The information gained from MR imaging studies is thus complementary to more subjective clinical data.¹¹² Brown et al. have proposed an MR imaging assessment system for cartilage transplantation and microfracture that includes several variables, including:¹¹²

As it is with microfracture, peripheral integration is a challenge with both autologous osteochondral transfer and osteochondral allograft transplantation. When peripheral integration was studied in a canine model using cartilage-sensitive MR imaging and T2 mapping, trabecular incorporation (defined as apparent osseous integration with no residual bone marrow edema pattern) was found in 89% of specimens at 6 months. On histologic inspection, however, both allograft and autograft specimens demonstrated a cleft at the host–graft cartilage interface, indicating incomplete peripheral integration at the articular surface. These findings support the evidence that articular cartilage cannot regenerate across a physical gap.¹¹⁶ In addition to peripheral integration and osseous incorporation of the plug, the degree of restoration of the geometry of subchondral bone is another challenge encountered with osteochondral transfer.

MR imaging of the site of cartilage repair provides a tomographic assessment of the degree of offset of the subchondral plate and also an evaluation of the articular surface, providing more detailed information than can be obtained at second-look arthroscopy (Fig. 7.26).¹¹⁷ T2 mapping techniques can be very helpful in assessing not only the orientation of the collagen over the site of the cartilage repair but also the response of the adjacent cartilage to the surgically manipulated cartilage (Fig. 7.27). Unlike T2 mapping obtained following microfracture, T2 relaxation times observed

with autologous osteochondral transplantation are closer to that of the remaining portion of the joint (Fig. 7.28).

Another important element in osteochondral transfer is restoration of the radius of curvature, which is important in maintaining the biomechanical integrity of the plugs. Koh et al. noted that peak contact pressures were significantly elevated (approximately 20%) after creation of a defect, were reduced to normal when the plugs were flush, and were significantly increased when the plugs were elevated.¹¹⁸ Of note, contact pressures with depressed plugs were significantly higher than with intact cartilage, but were lower than those associated with an empty defect (Fig. 7.29).¹¹⁸ Others have noted that proud plugs may undergo “repositioning” due to the effects of weight bearing, but subchondral cavitations were often present, suggesting excessive motion between the graft and recipient site.¹¹⁹

Assessment of trabecular incorporation is based on visualization of bony continuity from the plug to the host bone,

restoration of the normal fatty signal intensity of the plug, and lack of displacement. The use of both cartilage-sensitive and fat-suppressed imaging is therefore essential in assessing not only the integrity of the articular surface but also the degree of host subchondral bone turnover at the site of plug incorporation. The presence of low signal intensity on all pulse sequences strongly suggests loss of bone viability, which may lead to eventual implant failure (Fig. 7.30). Often multiple plugs are used, sometimes requiring instrumentation, and appropriate modification of pulse sequence parameters is necessary to reduce the associated susceptibility artifact in the case of metallic fixation. Alternatively, a “press fit” fixation

may be used, and in this situation care must be taken to avoid mistaking the low-signal-intensity compression of trabeculae at the side wall of the implant for failure of the plug to incorporate (Fig. 7.31).

Delayed incorporation may be related to allograft rejection. Sirlin et al. studied 36 shell osteochondral allografts, comparing MR findings to immunohistochemistry, and noted that antibody-positive patients demonstrated a greater degree of bone marrow edema pattern than antibody-negative patients.¹²⁰ Antibody-positive patients also demonstrated a higher proportion of surface collapse.¹²⁰ These findings suggest that a biologically predictable response of recipient to allograft bone may account for delayed integration on MR imaging, and this information may prove predictive of eventual collapse and implant failure.

Concerns about donor-site morbidity associated with autologous osteochondral transfers and disease transmission associated with allografts have prompted the increasing use of newer bone graft substitute implants, which are often composed of a co-polymer with calcium sulfate. The signal characteristics of these agents are distinctly different from those of autologous tissue (Fig. 7.32), and since they are not representative of native or cadaveric bone, they may be mistaken for delayed biologic incorporation on MR studies. In the absence of back-fill using synthetic plugs, the donor site typically fills in with autologous bone from the periphery and the surface material is typically hyperintense to native cartilage (see Fig. 7.27).

Tissue engineered cartilage involves three key elements:

Although the composition of the matrix in tissue-engineered cartilage is variable, requirements include:

Matrices of varying chemical composition have been used, including carbohydrate-based polymers such as polylactic acid and protein-based polymers, including fibrin and collagen. Additional components of tissue-engineered cartilage may include chondroprogenitor cell pools, such as those found in the cambial layer of periosteum and perichondrium, mesenchymal stem cells from the bone marrow or synovial membrane, or chondrocytes themselves.

An example of a tissue engineering type of technique is autologous chondrocyte implantation (ACI),¹²¹ a two-stage technique involving the arthroscopic harvest of autologous chondrocytes, cloning in tissue culture, and reimplantation at arthrotomy with a covering of autologous periosteum. Compared with microfracture, postoperative evaluation of ACI demonstrates consistently better fill. However, graft hypertrophy is not uncommon, accounting for some post-procedure morbidity.¹¹² Hypertrophy of ACI grafts occurs most commonly in the first 6 months following transplantation (Fig. 7.33).^112,122 In the initial few months following ACI, the repair cartilage is hyperintense relative to native cartilage, most likely due to its relatively immature matrix and increased mobility of water. This matrix can be discerned by the low-intensity periosteum (Fig. 7.34).^112,117 Verstraete et al. described an evolution of signal intensity changes after ACI:

The presence of persistent fluid signal intensity between the repair tissue and subchondral bone suggests impending delamination,¹²⁴ which is most commonly encountered in the early postoperative period of 6 months following repair.^112,124 Complete delamination of the ACI graft has also been described as a potential complication.^112,124

dGEMRIC techniques have also been used to evaluate autologous cartilage implantation. Some researchers have noted that in grafts less than 12 months old, the relative glycosaminoglycan levels appeared to be lower than the comparative hyaline cartilage. In grafts imaged after 12 months, relative

glycosaminoglycan levels were more comparable to the adjacent and remote hyaline cartilage.¹²⁵

Peripheral integration in ACI can also be assessed and is potentially one of the most important factors in evaluating cartilage repair. The use of a fluid-sensitive pulse sequence is essential to differentiate the hypertrophic synovium from fluid intensity in fissures, and the use of high in-plane resolution is necessary to evaluate the integrity of the interface.¹¹⁷ Verstraete et al. noted that edge integration of transplanted cartilage can take up to 2 years and, when complete, is seen as a lack of fluid intensity between the native and repair cartilage.¹²³