|Home | About | Journals | Submit | Contact Us | Français|
We assessed the relationship between cartilage MR relaxation times and biomechanical response of tibiofemoral articular cartilage to physiological loading in healthy subjects and patients with osteoarthritis (OA). Female subjects above 40 years of age with (N1 = 20) and without (N2 = 10) OA were imaged on a 3T MR scanner using a custom made loading device. MR images were acquired with the knee flexed at 20° with and without a compressive load of 50% of the subject’s bodyweight. The subjects were categorized based on the clinical MRI scoring of medial and lateral cartilage surfaces. Data were stratified twice into two equal groups (low and high) at the median value of T1ρ and T2 relaxation time. The change in contact area and cartilage deformation was measured within these groups. Paired Student’s t-test (α = 0.05) was used to analyze the effect of loading on contact area and deformation. The average area of the contact region in the medial compartment was significantly higher in OA subjects compared with normal subjects in both unloaded (314 ± 112 mm2 vs. 227 ± 106 mm2, p = 0.023) and loaded (425 ± 128 mm2 vs. 316 ± 107 mm2, p = 0.01) conditions. The overall relative change of cartilage thickness in the medial compartment was significantly higher than the lateral compartment (−5.3 ± 9.9% vs. −1.9 ± 9.2%, p = 0.042). When cartilage was divided into deep and superficial layers, superficial layers showed higher changes in relaxation time (T1ρ and T2) than the changes in relaxation time of whole cartilage (Normal: 12.5% vs. 6.9%; OA: 10.9% vs. 4.6%). The average T1ρ and T2 times, change in area of contact region, and change in cartilage thickness in subjects with OA were higher when compared to normal subjects. This study provides support for a relationship between the mechanical response of cartilage to physiological loading (cartilage-on-cartilage contact area and cartilage deformation) and MR relaxation times (T1ρ and T2) in both OA patients and normal subjects.
The biomechanical or biochemical mechanisms responsible for cartilage degeneration in osteoarthritis (OA) are not well understood. Biomechanical factors (such as altered gait or load distribution) and pathological response of cartilage to these factors are believed to contribute to the initiation and progression of degeneration.1 Injured or degraded cartilage shows a loss of proteoglycan (PG) from the extracellular matrix (ECM) and a disruption of the collagen fiber network.2 Studies have been proposed to assess the biochemical changes in cartilage using relaxation times (T2 and T1ρ) with the goal of diagnosing OA at its earliest stage when the disease process may be more amenable to intervention.3,4 Damage to the PG-matrix affects the T2 and T1ρ relaxation times. Changes in the biochemical properties of cartilage affect the mechanical behavior of cartilage.5 Though the relationship between mechanical properties and cartilage composition has been studied in specimens6 or cadavers,7 in vivo human studies have received less attention.8,9 Quantitative magnetic resonance imaging (MRI) has been suggested as a tool for studying quantitative cartilage biomechanics in humans.10,11 Cartilage-on-cartilage contact12,13 and cartilage deformation14 have been investigated in normal human joints. However, no studies have reported in vivo tibiofemoral joint contact mechanics with and without OA under static joint loading using high-field MRI.
The non-invasive assessment of the biomechanical response of cartilage to loading and its relation to biochemical changes will provide valuable information regarding the structural changes of cartilage in early OA. Due to degradation of the ECM affected by OA, the biomechanical properties are expected to change considerably. Understanding contact patterns of the articular surfaces of the tibiofemoral joint under physiological loading in healthy and diseased knees may help us predict the mechanical responses of cartilage to loading. Our objectives were to investigate changes in cartilage-on-cartilage contact patterns and deformation of tibiofemoral joint cartilage under physiological loading in subjects with and without OA and to relate loading patterns to prevalence and degree of cartilage degeneration and biochemical cartilage matrix composition as determined with T1ρ and T2.
Ten healthy volunteers and 20 patients with tibiofemoral OA affecting the medial compartment were included (females, age: 40–70 years, BMI: 20–35 kg/m2). Knee radiographic images were acquired from all subjects in a modified-Lyon-schuss weight-bearing position using a Plexiglas Synaflexer (Synarc, Newark, CA) positioning device. None of the volunteers had a history of frequent knee pain, aching, or stiffness during the past year and were free of radiographic evidence of OA (Kellgren Lawrence (KL) score of 0) on both knees. The inclusion criteria for the patient cohort were knee pain, aching, or stiffness on most days of a month during the past year and radiographic evidence of OA (KL score of 2 or 3) on the study knee with either the same or less severe OA on the contra-lateral knee. For OA subjects the medial joint space width was less than the lateral joint space width indicating medial OA disease. Informed consent was obtained from all subjects after the nature of the study had been fully explained. The study was approved by and performed in accordance with the rules and regulations of the Committee for Human Research at our institution.
MR imaging was performed on a 3T scanner (Signa HDx, General Electric, Milwaukee, WI), using an 8-channel phased array transmit-receive knee coil (Invivo, Gainsville, FL) and an in-house built loading apparatus mounted on the scanner table15 (Fig. 1). Two sets of MR images of one knee (dominant knee for controls and selected study knee for OA patients) were acquired under unloaded and loaded conditions. One half hour prior to imaging, the subjects were kept off their feet to avoid placing load on their knees. The first set of images were acquired while subjects were positioned supine on top of a custom-made MRI-compatible loading apparatus, in 20° of knee flexion and 10° of foot external rotation (placed on a footplate and supported in place) with no load applied. The next set was acquired in the same position while applying a load of 50% of the subject’s weight at the bottom of footplate through a pulley system, intended to simulate static standing. Padding was used to ensure no movement, and there was a consistent and comfortable knee positioning during scanning. The protocol included five sequences: Coronal 3D water excitation high-resolution spoiled gradient-echo (SPGR) images, sagittal fat-saturated T2-weighted fast spin-echo (FSE) images, coronal fat-saturated T2-weighted FSE images, 3D coronal T1ρ-weighted images based on SPGR acquisition that was previously developed in our lab [Magnetization-prepared Angle-modulated Partitioned-k-space Spoiled Gradient-Echo Snapshots (MAPSS)], and 3D T2-weighted images covering the same region as the T1ρ sequence.16 Acquisition parameters are given in Table 1.
MR images were evaluated and scored by two radiologists independently with 20 (TML) and 4 (CS) years of experience in musculoskeletal imaging. In instances of conflicting scoring, consensus readings by both radiologists were performed. Cartilage sub-scores of modified Whole-Organ Magnetic Resonance Imaging Score (WORMS) grading were used to semi-quantitatively assess the cartilage in all subjects as previously described.17 The anatomical regions used for grading were medial and lateral femoral condyle (MFC and LFC), and medial and lateral tibia (MT and LT). Cartilage abnormalities were scored using a WORMS scale: 0 = normal thickness and signal; 1 = normal thickness but abnormal signal on fluid sensitive sequences; 2.0 = partial-thickness focal defect <1 cm in greatest width; 2.5 = full-thickness focal defect <1 cm in greatest width; 3 = multiple areas of partial-thickness (Grade 2.0) defects intermixed with areas of normal thickness, or a Grade 2.0 defect wider than 1 cm but <75% of the region, 4 = diffuse (≥75% of the region) partial-thickness loss, 5 = multiple areas of full-thickness loss (grade 2.5) or a grade 2.5 lesion wider than 1 cm but <75% of the region, 6 = diffuse (≥75% of the region) full-thickness loss. Lesions in menisci and ligaments were quantified to assess joint integrity (Meniscus: 0 = no lesion, 1 = intrasubstance abnormalities, 2 = non-displaced tear, 3 = displaced or complex tear without deformity, 4 = maceration of the meniscus; ligaments: 0 = none, 1 = signal abnormalities around tendon/ligament, 2 = signal abnormalities within the tendon/ligament, 3 = partial tear, 4 = complete tear).
Bone, articular cartilage, and cartilage-on-cartilage contact region were segmented in 3D SPGR images using a software program developed in-house using a spline-based semi-automated (automated edge detection and manual correction) segmentation algorithm in MATLAB (Mathworks Inc, Segundo, CA). Cartilage was automatically divided into two layers: Deep (closest to the bone-cartilage interface) and superficial (closest to the articular surface). Cartilage pixels were classified to only one layer based on minimum Euclidean distances to the transferred splines. To establish the tibial anatomic coordinate system, anatomical landmarks on the plateaus (the most medial and lateral points—MPT and LPT) were automatically localized. The mediolateral (ML) axis was defined by connecting MPT and LPT with a line. The mid-point of this axis served as the origin for the tibial coordinate system. An imaginary line (parallel to the ML axis) was created by localizing the most medial and lateral points in the distal tibial shaft. The shaft axis was defined by connecting midpoints of the ML axis and the imaginary line. The anteroposterior (AP) tibial axis was defined by taking the cross product of the ML axis and the tibia shaft axis. The inferior–superior (IS) tibial axis was defined by taking the cross product of the ML and AP axes, providing mutually perpendicular anatomic axes with the origin at the midpoint of the ML axis. Coordinate systems were established for both unloaded and loaded conditions. The loaded tibia was registered to the unloaded tibia using an iterative closest point (ICP) matching method.18 Cartilage-on-cartilage contact regions of the medial and lateral compartments were defined by creating spline lines on each coronal SPGR image along the line of contact. The lines across all images were connected with triangles and integrated to compute total contact area. The centroid of each compartment contact region relative to the tibial coordinate system were calculated and transformed to the unloaded condition, giving the change in position of the contact region with loading (Fig. 2).
Cartilage above (femoral cartilage) and below (tibia cartilage) the region of contact was segmented in four regions for each subject: MFC, LFC, MT, and LT. After segmentation, cartilage thickness of each region was calculated based on a minimal inscribed sphere method. The minimal distance between the cartilage–bone interface and the articular surface was computed and defined as thickness. The average thickness of each slice was calculated, and the weighted averages were summed to determine overall cartilage thickness.19
T1ρ and T2 maps were reconstructed by fitting the image intensity (pixel-by-pixel) to the equation below using a Levenberg–Marquardt mono-exponential fitting algorithm developed in-house: S(TSL) = S0 exp(−TSL/T1ρ) for T1ρ fit-ting, where S0 is the signal intensity when TSL = 0 ms; S(TE) = S0 exp(−TE/T2) for T2 fitting, where S0 is initial signal intensity. T1ρ- and T2-weighted images with the shortest TSL or TE (therefore with highest SNR) were rigidly registered to high-resolution SPGR images acquired in the same exam using the VTK CISG Registration Toolkit. The transformation matrix was applied to the reconstructed T1ρ and T2 maps. Slices used for T1ρ and T2 quantification were selected based on cartilage-on-cartilage contact during both unloaded and loaded conditions. The cartilage regions of interest were overlaid on the registered T1ρ and T2 maps. Cartilage profiles were corrected manually to avoid synovial fluid or other surrounding tissue. To reduce artifacts caused by partial volume effects with synovial fluid, regions with relaxation time >150 ms in T1ρ or T2 maps were manually removed from the data used for quantification.3
The distributions of all variables were plotted and examined. A two-tailed paired t-test was used to detect significant differences in contact area and centroid between the medial and lateral compartments, in normal and OA subjects, before and after loading. All statistical analyses were performed using JMP (SAS Institute, Cary, NC). To assess the utility of baseline (unloaded) T1ρ and T2 as an indicator for changes in cartilage biomechanical properties and ultimately for measuring progression of OA, changes in contact area and changes in cartilage thickness with loading were evaluated. Data were stratified twice into two equal groups (low T1ρ vs. high T1ρ and low T2 vs. high T2) at the median value of base-line (unloaded) T1ρ and T2 relaxation time. The relative changes in contact area and cartilage thickness (deformation) were measured within these groups.
To measure the reproducibility of measurements, we rescanned five subjects within 2 weeks after the initial image acquisition. The reproducibilities for cartilage-on-cartilage contact region and cartilage thickness were measured using the coefficient of variation.
The “normal group” consisted of a complete absence of lesions (WORMS 0) and was observed in 16 subjects for the medial and 21 subjects for the lateral compartment (age = 42–61 years, median = 56 years, BMI = 23–31 kg/m2, median = 28 kg/m2). The “OA group” had small or large focal lesions (WORMS >0) and were observed in 12 subjects for the medial and seven for the lateral compartment (age = 49–68 years, median = 56.5 years, BMI = 24–31 kg/m2, median = 27 kg/m2). Only five of 30 knees had a score >1 for ACL; other ligaments were scored <1 (considered as normal). Six and three knees had a score >2 for medial and lateral menisci, respectively.
The contact area was significantly larger under the loaded condition for both normal and OA subjects, and for both lateral (p < 0.002) and the medial (p < 0.0001) sides (Table 2). The contact area in the medial compartment was significantly higher (p = 0.01) in OA subjects compared with normal subjects in both unloaded and loaded conditions (Table 2). No significant differences in contact area under either unloaded or loaded condition were found between normal and OA subjects in the lateral compartment (p > 0.05). In the loaded condition, the area of cartilage-on-cartilage contact in the medial compartment was significantly larger (p = 0.008) than in the lateral in both normal and OA subjects (Table 2). The change in the medial compartment was significantly higher than the lateral compartment in normal subjects (p = 0.002, Table 2) but not in patients with OA (Table 2). The average coefficient of variation for contact area was 6.7%.
Translation of the contact centroid was observed after loading in both groups and in both compartments. In the lateral compartment, OA subjects demonstrated significantly more translation in the medial direction when compared to normal subjects (1.18 ± 1.46 mm vs. −0.15 ± 1.49 mm; p = 0.025). No other differences were observed in centroid translation (Table 2).
When we pooled the data, the change in cartilage thickness under loading was significant in all regions except the LFC (Table 3). The overall relative change in the medial compartment was significantly higher than in the lateral. The largest cartilage deformation was observed in the MFC region in subjects with OA (−7.85 ± 10.99, p = 0.014), and in the MT region in normal subjects (−6.88 ± 14.09, p = 0.04). Greater deformation was seen in the MFC cartilage in OA subjects compared to normals, though this failed to reach significance (−7.85 ± 10.99% vs. −3.28 ± 7.06%, p = 0.096). When groups were analyzed separately, significant deformation was observed in the MFC and MT of normal and OA subjects, but not in the LFC or LT (Table 3). No significant differences were observed in the cartilage deformation in medial and LT between normal and OA patients (Table 3). The average coefficient of variation for the cartilage thickness was 2.2%.
The T1ρ and T2 times of subjects with OA were higher when compared to normal subjects (Fig. 3). The differences in T1ρ and T2 between OA and controls were larger in the lateral compared to the medial compartment (16.69% vs. 5.56% for T1ρ) and (13.89% vs. 7.86% for T2). The superficial layer showed higher change in relaxation times than the change in the whole cartilage (Normal: 12.45% vs. 6.98%; OA: 10.89% vs. 4.56%, Fig. 3). Analysis of the T1ρ-stratified data (low T1ρ vs. high T1ρ) revealed a significant increase in contact area in the high-T1ρ group compared to the low-T1ρ group (91 ± 59 mm2 vs. 56 ± 47 mm2; p = 0.013, Fig. 4). However, no differences in cartilage deformation were observed between these groups (−4.49 ± 10.49% vs. −3.82 ± 10.32%, for high and low- T1ρ groups, respectively; p = 0.733). In contrast to T1ρ-based groups, there were no differences in cartilage contact area or deformation between the T2-stratified groups. However, with regard to contact area, the same increasing trend was observed in the high-T2 group (80 ± 55 mm2 vs. 63 ± 54 mm2), but it was not significant (p = 0.065).
When data were further stratified based on compartment within each T1ρ group (low T1ρ and high T1ρ), the same behavior was observed in the change in contact area (Fig. 5). A significant increase in contact area was observed in the medial compartment of the high-T1ρ group (75 ± 56 mm2–115 ± 54 mm2; p = 0.038) but not in the lateral compartment (42 ± 35 mm2–58 ± 52 mm2; p = 0.396). For the T2-stratified data, neither the medial compartment subgroup (−5.04 ± 11.75% to −6.48 ± 10.30%; p = 0.612) nor the lateral compartment sub-group (−3.03 ± 9.37% to −1.42 ± 10.24%; p = 0.571) showed a significant difference in cartilage deformation. In line with the same observation made in normal and OA subjects, the relative increase of change in contact area from low-T2 group to high-T2 group in lateral compartment was larger than the medial compartment (22 mm2, 46% vs. 8 mm2, 10%), but not insignificantly so (p = 0.474).
Changes in cartilage-on-cartilage contact area and cartilage deformation were investigated and related with MR relaxation times during acute loading. Contact area and cartilage deformation in the medial compartment was larger than in the lateral compartment in both normal and OA subjects. The same trend was observed in MR relaxation times of cartilage (T1ρ and T2). After loading, larger increases in cartilage-on-cartilage contact area and cartilage deformation were observed in OA patients compared to normal subjects. In addition, when data were stratified based on relaxation times and compartments, increased contact area was observed in the medial compartment of subjects with high T1ρ values.
Higher contact area in subjects with OA is likely due to changes in mechanical and structural properties of damaged cartilage. Higher contact area in the medial compartment suggests a larger share of the load compared to the lateral compartment. The change to loading differed between groups, whereby normal subjects responded with a significantly larger increase in contact area in the medial compared to the lateral compartment, but the same relationship was not observed in OA subjects. The difference in contact area between OA patients and normal subjects was higher in lateral than the medial compartment, suggesting a change in load sharing pattern due to OA. Despite this difference, the overall magnitude and the absolute change in contact area in response to loading was greater in the medial compartment in both groups. Support for altered loading mechanics in subjects with OA comes from Astephen et al.,20 who reported that knee moment and angle data during self-selected gait were significantly different between healthy and OA subjects. They reported a higher adduction moment in OA subjects, consistent with the larger load sharing in the medial compartment observed in our study.
The increase in contact area in subjects with OA may indicate that the damaged collagen network allows migration of proteoglycans out of the ECM, thus losing an important source of compressive stiffness. Severe changes in mechanical properties were reported in OA cartilage.5,21 While recent studies reported contact mechanics using dual-orthogonal fluoroscopy and MR images, differences in methodology make a direct comparison to our study difficult.13 Earlier studies13,22 used simple interaction of 3D geometric models for measuring cartilage deformation and contact area, considering cartilage as a rigid body. But Neu et al.23 reported non-uniform cartilage deformations in a cadaver knee under unconfined compression testing. However, in our cross-sectional study, it cannot be determined if differences in contact areas are the cause of the disease process or a result of it.
Cartilage deformation in the medial compartment was higher than in the lateral compartment when all subjects were combined. These findings agree with previous studies that showed more medial compartment deformation.14,22,23 Consistent with our observations of contact area, deformation was higher in OA patients. These results taken together with the contact area findings suggest that structural degradation indeed affects the load bearing capacity of cartilage.13,14,22 These findings suggest that fibrillation of the collagen network lowers its ability to restrain PG swelling forces, leading to increased tissue permeability, and decreased compressive stiffness, all reflected in the larger deformation and contact areas in the OA subjects.
The same trend was observed in our low-T1ρ and high-T1ρ groups, consistent with the clinical observation of an increased incidence of medial OA in the general population. Earlier studies suggested that exchange of protons between PG and the tissue water content could be an important relaxation mechanism contributing to T1ρ relaxation time.24 This suggests that T1ρ may be sensitive to changes of PG in the cartilage matrix during early OA. Furthermore, T1ρ values seem to be less affected by the orientation of collagen that can affect T2 relaxation techniques.25 T1ρ relaxation times are inversely correlated with PG content, while T2 relaxation times are related to cartilage collagen and water content.26 Elevated T2 values in OA subjects indicate that cartilage degeneration results in alterations in the structure of the collagen network and an increase in water content. Change in T1ρ and T2relaxation times of the superficial cartilage layer in OA subjects was lower than in normal subjects. This change may be due to alteration of contents in the superficial layer than in the deep layer under loading. Histological studies showed that biochemical cartilage composition is heterogeneous; near the articular surface, the PG concentration is relatively low, and water content is high. Conversely, in the deeper region near the subchondral bone, PG concentration is relatively high, and water content is low. Our subjects with elevated relaxation times displayed greater contact area in the medial, but not in the lateral, compartment. To date, no studies have investigated the effects of acute loading using relaxation time stratification.
Limitations of the study must be considered when interpreting our findings. One limitation is that the alignment of the lower limb was not measured, which is an important element in analyzing biomechanics of the limb. Also, a small number of subjects may have prevented significance from being reached in some measures; large standard deviations were observed. However, consistent trends were observed within each subject. Since the quantified contact area and deformation are in “relative % change,” the results are still valid even if a subject had a big knee (initial larger contact area) or thicker cartilage (larger deformation). The joint integrity and muscle strength to hold the dead weight and joint balancing mechanism may also be responsible for the large standard deviations. For example, a relatively large translation of the contact point on the lateral tibial plateau consistently occurred within each subject. Also, the load application was performed at only one flexion angle. Li et al.13 showed that, beyond 30° of knee flexion, there was a minimal change and movement of the contact region. Todo et al.27 showed that flexion is accompanied by internal tibial rotation. Also, tall and large size subjects may not be able to flex the knee with the circular knee coil around it inside the magnetic bore beyond 20–30°. So, we decided to flex the knee 20° while imaging. Replication of this study using high-field open MRI would enable a larger flexion range and various standing postures under true physiological loading. Integration of high-field open MRI and quantitative MRI methods would give insight into the non-invasive evaluation and monitoring cartilage biomechanics.
In conclusion, we provide evidence of an association between the mechanical response of cartilage to physiological loading (cartilage-on-cartilage contact area and cartilage deformation) and MR relaxation (T1ρ and T2) in both OA patients and normal subjects. In clinical applications, these imaging biomarkers and response of cartilage to physiologic load may be used as metrics to measure progression and development of OA, but further large cohort studies are required with long-term follow up to evaluate their effectiveness.
The source of funding was Pfizer, Inc., Groton, CT. The authors would like to thank Eric Han from GE Healthcare for his help with pulse sequence development and Choongsoo Shin for his assistance with data acquisition.