Effectively Measuring Exercise‐Related Variations in T1ρ and T2 Relaxation Times of Healthy Articular Cartilage

Determining the compositional response of articular cartilage to dynamic joint‐loading using MRI may be a more sensitive assessment of cartilage status than conventional static imaging. However, distinguishing the effects of joint‐loading vs. inherent measurement variability remains difficult, as the repeatability of these quantitative methods is often not assessed or reported.

O VER THE LAST TWO DECADES, in vivo magnetic resonance imaging (MRI) has increasingly been used to determine the mechanical properties of knee articular cartilage. Previous studies have shown that cartilage loading activities affect the morphology and biochemical composition of articular cartilage and have provided important information on the behavior of cartilage when exposed to different compressive loads. 1-3 T 1ρ and T 2 relaxation time mapping techniques allow the assessment of cartilage compositional alterations in response to joint-loading, as they have been demonstrated to be sensitive to variations in the water and macromolecular content of cartilage. [4][5][6] Normalized changes in T 1ρ and T 2 relaxation times of cartilage following different exercise regimes have been shown to be on the order of -2.6% to -14.3% and +3.7% to -12.5%, respectively. 2,3,[7][8][9] Since the measured changes resulting from joint-loading can be small, determining the intrasessional repeatability of these quantitative measures is essential for reliable assessment of joint-loading-related effects on cartilage structure and composition.
A systematic review showed that studies assessing the repeatability of these quantitative relaxation techniques without any joint-loading activity have reported root-meansquared coefficient of variation (RMS-CV) for large regional analysis of T 1ρ values in the range of 2.3-6.3% and of T 2 values in the range of 2.3-6.5%. 10 When subregional or laminar cartilage analysis was performed, test-retest CVs for T 1ρ were up to 19% and for T 2 as high as 22%. 10 Intrasessional repeatability assesses the repeatability of measurements of 1) consecutive scans without repositioning and 2) consecutive scans with repositioning of the subject. 11 Evaluating the repeatability of consecutive scans without repositioning is important when measuring T 1ρ and T 2 at multiple timepoints after joint-loading for determining longitudinal cartilage recovery, as previous studies have reported. 1,12,13 Healthy cartilage is maintained with regular deformation and compression of the cartilage structure and its extracellular matrix (ECM) through physiological loading, such as experienced during exercise. 14,15 However, both overuse and disuse can have degenerative effects on the cartilage and are important risk factors in the development of osteoarthritis (OA). [15][16][17] When exposing the cartilage repeatedly to excessive loads, such as may occur during high-impact sports or, to minimal or no load following injury, the cartilage structure and microstructure begin to break down. 15,18 Morphological changes in articular cartilage volume, thickness, and joint space narrowing are not necessarily present in the early stages of OA and may change very slowly during disease progression. Therefore, measuring differences in cartilage deformational responses during or after loading may represent a more sensitive biomarker for detecting the early onset of OA. 19,20 The aim of this study was to measure the intrasessional repeatability of both T 1ρ and T 2 of knee articular cartilage and to determine if these quantitative relaxation measurement techniques are sensitive to permit effective measurement of short-term cartilage compositional responses after a jointloading activity.

Materials and Methods
All imaging was performed on a 3T MRI system (MR750, GE Healthcare, Waukesha, WI) using an 8-channel transmit/receive knee coil (Invivo, Gainesville, FL). Participant imaging had local ethical approval, and written informed consent was provided by each participant.

Study Procedures
PHANTOM REPEATABILITY. To assess the test-retest repeatability of the quantitative T 1ρ and T 2 relaxation time measurements for a range of relaxation times, two consecutive T 1ρ and T 2 relaxation mapping datasets were obtained from a phantom. The phantom consisted of five vials having different T 1ρ and T 2 relaxations. Two vials had T 1ρ and T 2 relaxation times similar to cartilage (~40-50 msec) at 3T, while the relaxation times of the remaining three vials were greater. 21,22 To additionally assess the intersessional variability (scanning the same phantom on different days), two further T 1ρ and T 2 relaxation mapping datasets were acquired 2 days later. On each, the same knee coil and setup was used with the phantom centered in the coil.

GROUP 1: IN VIVO REPEATABILITY STUDY.
To assess the intrasessional repeatability of T 1ρ -and T 2 -relaxation mapping of cartilage, the right knee of 10 healthy participants (five men, five women, mean age 28.9 AE 5.5 years) with no current knee pain symptoms, nor known history of joint disorder, was imaged. Imaged knees were unloaded for 15 minutes prior to the imaging session to minimize short-term loading effects on the joint.
The MR session consisted of a sagittal 3D fat-saturated spoiled gradient recalled-echo (3D-FS SPGR) sequence, and sagittal T 1ρand T 2 -mapping sequences. For details on pulse sequence parameters used, see section "Sequence Parameters," below. Following repositioning of the participant and imaged knee, two consecutive acquisitions of T 1ρ -and T 2 -mapping were performed using the same pulse sequences as before repositioning (Fig. 1a). During knee repositioning, the participants removed their knee from the coil and sat up on the side of the MR table. The coil was repositioned, followed by participant positioning. The time required for repositioning and the continuation of the imaging protocol was~5 minutes.
GROUP 2: EXERCISE AND RECOVERY STUDY. A second group was used to assess the magnitude of effect that mild exercise had on T 1ρ -and T 2 -relaxation mapping of cartilage. The right knee of nine healthy participants (five men, four women, mean age 31.6 AE 6.0 years) with no current knee pain symptoms, nor known history of joint disorder was imaged. Imaged knees were unloaded for 15 minutes prior to the imaging session to minimize short-term loading effects on the joint.
The study design consisted of a 3D-FS SPGR sequence, followed by T 1ρ -and T 2 -relaxation imaging before exercise, and at four timepoints after exercise to assess cartilage compositional recovery. The standardized exercise protocol involved 5 minutes of stepping onto a step-stool (height ≈ 24 cm) with one leg and stepping down onto the other side of the step-stool with the leg to be imaged (Fig. 1b). This resulted in~20 stepping cycles per minute in which the knee joint was repeatedly loaded.

Sequence Parameters
where M(TSL) is the signal intensity of the T 1ρ -weighted image at a specific TSL and M 0 is the initial magnetization / signal intensity. T 1ρ relaxation times >130 msec in T 1ρ maps were excluded from analysis to avoid partial volume effects with synovial fluid. 25,26 T 2 MAPPING. T 2 maps were obtained with a sagittal T 2 -prepared PSS 3D-FSE sequence using a composite 90 x -180 y -90 x pulse train for T 2 -preparation. 23,27 Images were acquired using the following parameters: acquisition time = 5:25 min; matrix = 320 × 256 interpolated to 512 × 512; FOV = 160 × 144 mm 2 ; reconstructed voxel size = 0.31 × 0.31 × 3 mm 3 ; flip angle = 90 ; TR = 1580 msec; TEs = 6.5, 13.4, 27.0, 40.7 msec; 72 slices per TE; echo train length = 45; NEX = 0.5; and bandwidth = AE62.5 kHz. The T 2 maps were created using a log-linearized least-squares algorithm to fit a monoexponential decay function to the signal intensities: where M(TE) is the signal intensity of the T 2 -weighted image at a specific TE and M 0 is the initial magnetization / signal intensity. As with T 1ρ , T 2 relaxation times >100 msec in T 2 maps were excluded from analysis to avoid partial volume effects with synovial fluid. 25,26 Imaging Analysis PHANTOM REPEATABILITY. Mean relaxation times from all five vials of the phantoms were determined using rectangular regions-ofinterest (ROIs) placed on two central sequential slices of the sagittal T 1ρ and T 2 maps. After having the participant sit and keep the imaged knee in an unloaded state for~15 minutes prior to imaging, initial T 1ρ and T 2 relaxation mapping was acquired. Following knee repositioning, two successive T 1ρ and T 2 relaxation mapping measurements were acquired. (b) in vivo assessment of the change in cartilage composition following mild exercise. The imaged knee (green) was kept in an unloaded state for~15 minutes before acquiring the initial T 1ρ and T 2 relaxation measurements. Following mild exercise, four repeats of T 1ρ and T 2 relaxation mapping measurements were acquired to evaluate cartilage compositional change and recovery following exercise.
IN VIVO SURFACE ANALYSIS. All T 1ρ -and T 2 -weighted images were rigidly registered to the high-resolution 3D-FS SPGR images using the Elastix toolbox 28 before calculating the respective quantitative maps. Surface-based analysis (3D Cartilage Surface Mapping, 3D-CaSM) of femoral, tibial, and patellar cartilage was performed using the freely available Stradwin software v. 5.4a (University of Cambridge Department of Engineering, Cambridge, UK; now freely available as "StradView" at http://mi.eng.cam.ac.uk/Main/ StradView/). 29 After creating sparse manual cross-sections (on every 2 nd -4 th sagittal slice) of the patella, tibia, and femur including their surrounding cartilage on the 3D-FS SPGR datasets, a triangulated surface mesh object of each segmented bone-cartilage structure was automatically generated using shape-based interpolation and the regularized marching tetrahedra method. 30 Following cartilage thickness calculation and the generation of inner and outer cartilage surfaces, these surfaces were used to analyze the registered quantitative T 1ρ and T 2 maps. At each vertex, the T 1ρ and T 2 values along a perpendicular line between inner and outer surface (surface normal) were sampled and averaged.
Canonical (average) femoral, tibial, and patellar meshes were created from all participants to be able to compare the T 1ρ and T 2 value distributions between participants. Canonical surfaces were calculated from all participants involved in the exercise and recovery imaging. All quantitative surface data from both the repeatability and exercise-recovery cohorts were mapped onto the canonical surface following surface registration. Canonical surface generation and the subsequent registration and mapping of the individual surfaces was performed using the freely available wxRegSurf software v. 18 (University of Cambridge Department of Engineering, Cambridge, UK; freely available at http://mi.eng.cam.ac.uk/~ahg/wxRegSurf/). The full 3D-CaSM analysis pipeline is illustrated in Fig. 2.

Statistical Analysis
PHANTOM REPEATABILITY. CVs were calculated from the two successive repeatability scans on each day (CV Phant,Day1 , CV Phant, Day2 ) for all five vials using: with σ being the within-vial standard deviation and μ the within-vial mean of measurements. The intraphantom variability was evaluated by calculating the CV from the mean and standard deviation of the relaxation values obtained from both days (CV Phant,All ).
The intrasessional repeatability of T 1ρ and T 2 acquisitions was assessed by calculating RMS-CV from the surface-averaged T 1ρ and T 2 measurements of all participants for femoral, medial tibial, lateral tibial, and patellar cartilage surfaces. The RMS-CV between repeatability measurements 1 (before repositioning) and 2 (first measurement following repositioning) were calculated (RMS-CV S1-S2 ) to evaluate the effects of knee repositioning on repeatability. The RMS-CV between measurements 2 and 3 (with no repositioning between either measurement) were determined to assess repeatability without knee repositioning (RMS-CV S2-S3 ).
The smallest detectable difference (SDD) 31 was calculated as the repeatability coefficient from the AE95% confidence intervals (CIs) from a Bland-Altman analysis 32 of all surface vertices of the repeatability data for all four cartilage surfaces and for both T 1ρ and T 2 .
GROUP 2: EXERCISE AND RECOVERY STUDY. To determine the effects of the dynamic joint-loading stepper activity on mean MR relaxation times of entire cartilage surfaces, linear mixed-effects models with timepoint as a fixed effect and participant as a random effect for each surface/parameter combination were created. For all statistical analysis, a level of significance of 0.05 was used.
The upper (+1.96 Á σ) and lower (-1.96 Á σ) limits of agreement as determined from the AE95% CIs of the Bland-Altman plots of the repeatability data were used to establish thresholds.
Exercise-induced changes in vertex-wise T 1ρ and T 2 relaxation times greater than the SDD signify variations that have a 95% probability of representing a true change rather than a variation due to measurement error. 33 Thresholds were determined for all four cartilage surfaces of interest. The determined thresholds were applied to the canonical surface data to only present cartilage regions undergoing a statistically significant exercise-induced compositional change at each surface vertex.
Vertex-wise percentage changes in T 1ρ (%T 1ρ change) and T 2 (%T 2 change) following exercise were calculated as the normalized change in cartilage relaxation time measurements: where T relax,post is the relaxation time measurement at a postexercise timepoint and T relax,pre is the relaxation time measurement prior to exercise. The variability of T 1ρ and T 2 relaxation values during cartilage compositional recovery following exposure to the mild stepping exercise was assessed only in the cartilage regions determined as regions experiencing significant exercise responses.

Phantom Imaging
The phantom test-retest repeatability on both days (CV Phant, Day1 , CV Phant,Day2 ) was ≤2.29% for T 1ρ and ≤0.74% for T 2 relaxation time measurements for all five vials. The CVs for the two phantoms having relaxation times comparable to cartilage were ≤0.64% for T 1ρ and ≤0.21% for T 2 . The intersessional repeatability (CV Phant,All ) calculated from all phantom repeatability scans over both days was ≤2.94% and ≤1.43% for T 1ρ and T 2 relaxation time measurements, respectively. The measured relaxation times and determined CVs are listed in Table S1in the Supplemental Material.

Group 1: in vivo Repeatability Study
The intrasessional repeatability RMS-CV for in vivo relaxation time measurements averaged over the entire femoral, medial tibial, lateral tibial, and patellar cartilage surfaces are listed in Table 1   For RMS-CV calculation, the vertex-wise T 1ρ and T 2 measurements were averaged over whole femoral, lateral tibial, medial tibial, and patellar cartilage surfaces. Between repeatability scans 1 and 2, the knee was repositioned (RMS-CV S1-S2 ). Repeatability scans 2 and 3 were obtained successively and without knee repositioning (RMS-CV S2-S3 ). patellar cartilage (51.2 msec ! 54.8 msec, RMS-CV S1-S2 = 4.8%) and the mean surfaced-averaged T 2 relaxation times of the lateral tibial cartilage (32.0 msec ! 32.9 msec, RMS-CV S1-S2 = 2.0%). The Bland-Altman plots for vertex-wise T 1ρ and T 2 repeatability measurements with knee repositioning of all four cartilage surfaces under investigation are shown in Fig. 3a,b, respectively.
The determined SDD and 95% limits of agreement from the Bland-Altman plots of all four cartilage surfaces and both compositional MRI methods are listed in Table 2.

Group 2: Exercise and Recovery Study
The T 1ρ and T 2 relaxation times averaged over whole femoral, lateral tibial, medial tibial, and patellar cartilage surfaces are illustrated in Fig. 4. The determined mean baseline T 1ρ relaxation times from the exercise-recovery cohort for femoral, lateral tibial, medial tibial, and patellar cartilage surfaces were 50.9 AE 3.6 msec, 44.3 AE 4.5 msec, 44.9 AE 3.7 msec, and 51.2 AE 8.9 msec, respectively. Mean baseline T 2 relaxation times for femoral, lateral tibial, medial tibial, and patellar car-  Table  S2 in the Supplemental Material. Figures 5 and 7 highlight the cartilage regions experiencing statistically significant changes in T 1ρ and T 2 relaxation times following the mild stepping exercise, respectively. Correspondingly, Figs. 6 and 8 respectively illustrate the alteration ("recovery") in participant-averaged femoral T 1ρ and T 2 percentage (%T 1ρ and %T 2 ) changes determined from the four postexercise measurements (scans 2-5) and the one preexercise baseline measurement (scan 1). Plots illustrating the variations in average lateral tibial, medial tibial, and patellar %T 1ρ and %T 2 changes are shown in Figures S1-S3 in the Supplemental Material, respectively. Table 3 shows the total number of vertices of each canonical cartilage surface and the percentage of cartilage surface area covered in regions experiencing changes (increases and decreases) in T 1ρ (T 1ρ -%SC) and T 2 (T 2 -%SC) relaxation time measurements above the determined measurement errors.
An average %T 1ρ change of -7.9 AE 5.5% and %T 2 change of +2.8 AE 8.6% were determined from all canonical patellar cartilage areas experiencing a significant change in relaxation times immediately following exercise. For the canonical femoral cartilage surface, average %T 1ρ and %T 2 changes of -8.0 AE 4.9% and -5.3 AE 2.3% were observed in response to exercise, respectively. Average %T 1ρ and %T 2 changes determined from all canonical lateral tibial cartilage regions displaying significant responses to exercise were -6.9 AE 3.2% and -5.9 AE 2.8%, respectively. Average medial tibial cartilage %T 1ρ change of +5.8 AE 5.2% and %T 2 change of +2.8 AE 9.5% were determined.
The highest negative normalized change of -25.5% was observed in the patellar cartilage T 1ρ followed by -17.3% in femoral cartilage T 1ρ and -15.0% in lateral tibial cartilage T 2 . The largest positive normalized change of +28.4% was displayed in the patellar cartilage T 2 followed by +15.7% in medial tibial cartilage T 2 and +12.1% in medial tibial cartilage T 1ρ .
When looking at cartilage compositional recovery following exercise and comparing the surface %T 1ρ and %T 2 changes calculated from the first postexercise measurements with the %T 1ρ and %T 2 changes determined from last postexercise measurements, patella cartilage %T 1ρ change recovered by 15%, while the T 2 "recovered" by 171%. The overall femoral cartilage %T 1ρ change dropped by 13% and the % T 2 change increased by 2% compared to the initial, first postexercise percentage change. While the lateral tibial cartilage % T 1ρ change decreased by 15% of its initial value, the medial tibial %T 1ρ change increased by 1%. The overall %T 2 change of both lateral and medial tibial cartilage increased by 12% and 50% compared to their initial values, respectively.

Discussion
This work determined the effects of a mild dynamic stepping exercise on the MR relaxation times of cartilage surfaces related to variation in biochemical composition.
The intrasessional repeatability CVs for T 1ρ and T 2 in this study were lower than or comparable to those determined in previous studies. 10 When looking at the surface-averaged T 1ρ and T 2 repeatability measurements without knee repositioning, both T 1ρ and T 2 were very repeatable on all surfaces. Repositioning of the knee had the greatest effect on the T 1ρ relaxation time measurements of patellar cartilage. During repositioning the knee joint experienced bending which could lead to larger changes in cartilage composition at the patellofemoral cartilage contact areas though friction than at the tibiofemoral areas. Averaging of relaxation times over large surfaces could mask these effects on the femoral cartilage surface due to its greater size in comparison to the smaller patellar surface. However, knee repositioning did not show a similarly strong effect on the patellar T 2 relaxation time measurements. This could be a consequence of the time delay (≈10 minutes) required for patient positioning, localization and T 1ρ data acquisition before the T 2 acquisition started, and therefore allowing compositional recovery during this time period.
In this study, 3D surface analysis was performed to help gain a better insight into how different cartilage regions respond to and recover from exercise. When averaging the T 1ρ and T 2 measurements over the entire femoral, lateral tibial, medial tibial, and patellar cartilage surfaces, no statistically significant exercise-related changes were determined when comparing the preexercise scan with the first postexercise scan. As a previous study also reported, determining mean relaxation time changes from individual slices or across large ROIs may mask significant focal changes. 34 When the individual vertex-wise relaxation times measurements in this study were regridded onto a canonical surface, significant exercise-related focal changes in T 1ρ and T 2 were observed. Although individual participants showed different cartilage compositional response to the exercise performed, cartilage regions experiencing compositional responses consistent across all participants became evident. By thresholding the exerciserelated changes in MR relaxation time measurements with the predetermined threshold limits from the repeatability measurements, cartilage regions undergoing significant responses to the mild dynamic joint-loading activity were highlighted. : T 1ρ (top) and T 2 measurements (bottom) averaged over whole femoral, medial tibial, lateral tibial, and patellar cartilage surfaces for all exercise recovery scans. Each color represents an individual participant, with the black curve representing the mean average trend (loess) of all participants with shaded 95% CIs. Between the baseline scan (timepoint 0) and the first postexercise scan (timepoint 1), the participant performed a stepping activity dynamically loading the imaged knee for 5 minutes. The first postexercise T 1ρ -and T 2 -mapping sequences were acquired~5 and 10 minutes after patient positioning, respectively. The last postexercise T 1ρ -and T 2 -mapping sequences (timepoint 4) were acquired~35 and 40 minutes after patient positioning, respectively. The acquisition of the postexercise imaging protocol took~45 minutes.
Since greater overall normalized changes were seen with T 1ρ than with T 2 relaxation time measurements, T 1ρ may be a more sensitive biomarker for detecting compositional cartilage responses to joint-loading activities. The %T 1ρ changes of patellar (-7.9%), femoral (-8.0%), and lateral tibial (-6.9%) cartilage and the %T 2 changes of femoral (-5.3%) and lateral tibial (-5.9%) cartilage observed in this study are comparable to those seen in previous studies. Mosher et al showed a %T 2 change of approximately -2.5% to -3.2% in femoral and -1.3% to -3.6% in lateral tibial cartilage following a 30-minute running activity. 35 Similarly, Subburaj et al demonstrated a %T 1ρ change of -4.1% to -14.3% and a % T 2 change of -3.0% to -9.3% in femoral, tibial, and patellar cartilage following running for 30 minutes. 2 The joint movements during the stepping activity performed in this study are comparable to the movements during the stair activity carried out in the study by Chen et al. 3 Similarly, the 5-minute stepping activity performed in this study showed a greater effect on patellofemoral cartilage T 1ρ relaxation times than on those of femorotibial cartilage, especially in the region of patellofemoral cartilage contact.
We not only observed regions experiencing significant decreases, but also significant increases in relaxation time measurements immediately following exercise, especially in medial tibial T 1ρ and T 2 , and patellar T 2 . Farrokhi et al also demonstrated a slightly increased %T 2 relaxation time change of 0.3% of healthy patellar cartilage following 50 deep knee bends. 7 Gatti et al showed an increased medial femoral %T 2 change after participants bicycled for~45 minutes. 9 Areas of increased normalized change could result from water redistribution rather than expulsion, increasing the water content and decreasing collagen and proteoglycan concentrations in these regions.
Various compositional "recovery" time-courses were determined for the four different cartilage surfaces. While patellar cartilage volume has been shown to recover in an almost linear fashion following 100 knee bends, we did not observe this linear recovery pattern in patellar cartilage  composition. 1 Overall, we only observed a drop in compositional normalized change in four instances (%T 1ρ change of patellar, femoral, and lateral tibial cartilage; %T 2 change of patellar cartilage), while in the other four instances (%T 1ρ change of medial tibial cartilage; %T 2 change of femoral, medial, and lateral tibial cartilage) an increase in normalized change was observed during the recovery period (postexercise scan 2 à scan 5). Cartilage morphology (thickness, volume), independent of cartilage health state, has been shown to recover almost fully in about 45-90 minutes following 30 36 and 100 knee bends 1 and a 30-minute 13 and 20 km run. 12 Based on our results, the focal compositional changes appear to require more time to return to baseline. More cartilage surfaces experienced some degree of compositional recovery in T 1ρ compared to T 2 , suggesting that the proteoglycan concentration is recovering faster due to water uptake than the changes in the collagen network after cessation of dynamic joint-loading.
The stepping exercise performed in this study is mild and of short duration. This exercise type was chosen as it is thought to be feasible and extendable for use in patients with early-stage knee joint disease and minimal accompanying pain. Knowledge of the effects that deformational loads have on cartilage structure and biochemical composition are important when evaluating clinical imaging studies aiming at determining differences in healthy and diseased cartilage. Differences in cartilage compositional MR relaxation time measurements between healthy and osteoarthritic cartilage have been shown to be in the range of 2-13% for T 1ρ and 1-12% for T 2 for large regional analysis. 21,25,37 As the diseaseinduced compositional changes in cartilage reflected in T 1ρ and T 2 measurements can be of the same order, and appear in similar cartilage regions, as exercise-induced changes, it is important to mitigate these effects when conducting clinical OA trials. A 3D surface analysis provides the possibility of spatially localizing the deformational and compositional  effects of joint-loading on articular cartilage and could also assist in determining the regions most prone to exhibit cartilage degeneration. 29 Limitations As the number of participants in the repeatability and exercise-recovery groups was limited, a larger sample size would increase the precision of the study results. A major limitation to in vivo studies assessing cartilage response to different joint-loading activities is that the compositional behavior of cartilage cannot be determined immediately after cessation of the exercise, but only some short time after, as time is required to position the participant back in the MRI system and for acquiring the data. Additionally, the T 1ρ and T 2 relaxation time mapping data were not acquired simultaneously but sequentially. Although both sequences are fast spin-echo-based sequences, the T 2 mapping was always performed about 6 minutes after T 1ρ , during which time further compositional recovery could take place, preventing an exact comparison between T 1ρ and T 2 results. A sequence capable of simultaneous T 1ρ and T 2 acquisition, such as the sequence proposed by Li et al, 38 could help address this issue.

Conclusion
We have shown that exercise-related changes in cartilage T 1ρ and T 2 relaxation times exceed measurement error and can reliably be determined when using the described 3D-CaSM analysis approach. Based on the results presented here, we hypothesize that mapping of cartilage T 1ρ and T 2 relaxation times are measuring dissimilar compositional features, as similar cartilage regions showed different T 1ρ and T 2 responses to exercise. However, while complete morphological recovery has previously been shown, the question of when, whether, and how the different cartilage regions recover completely from compositional variations following joint-loading activities persists.