T1ρ-mapping for assessing knee joint cartilage in children with juvenile idiopathic arthritis — feasibility and repeatability
Ongoing arthritis in children with juvenile idiopathic arthritis (JIA) can result in cartilage damage.
To study the feasibility and repeatability of T1ρ for assessing knee cartilage in JIA and also to describe T1ρ values and study correlation between T1ρ and conventional MRI scores for disease activity.
Materials and methods
Thirteen children with JIA or suspected JIA underwent 3-tesla (T) knee MRI that included conventional sequences and a T1ρ sequence. Segmentation of knee cartilage was carried out on T1ρ images. We used intraclass correlation coefficient to study the repeatability of segmentation in a subset of five children. We used the juvenile arthritis MRI scoring system to discriminate inflamed from non-inflamed knees. The Mann-Whitney U and Spearman correlation compared T1ρ between children with and without arthritis on MRI and correlated T1ρ with the juvenile arthritis MRI score.
All children successfully completed the MRI examination. No images were excluded because of poor quality. Repeatability of T1ρ measurement had an intraclass correlation coefficient (ICC) of 0.99 (P<0.001). We observed no structural cartilage damage and found no differences in T1ρ between children with (n=7) and without (n=6) inflamed knees (37.8 ms vs. 31.7 ms, P=0.20). However, we observed a moderate correlation between T1ρ values and the juvenile arthritis MRI synovitis score (r=0.59, P=0.04).
This pilot study suggests that T1ρ is a feasible and repeatable quantitative imaging technique in children. T1ρ values were associated with the juvenile arthritis MRI synovitis score.
KeywordsAdolescents Arthritis Cartilage Children Gadolinium Juvenile idiopathic arthritis Magnetic resonance imaging Quantification T1ρ
Juvenile idiopathic arthritis is the most common rheumatic disease in childhood . Cartilage can be damaged by the autoimmune-mediated inflammation that originates in the synovial membrane. Ongoing inflammation can subsequently extend to the cartilage and result in degradation of cartilage and bone matrix. Osteochondral damage is presumably irreversible and associated with disability and decreased quality of life [2, 3]. Previous research also indicates that damaged cartilage matrix facilitates binding of synovial inflammatory cells . Thus, even when inflammation subsides, a damaged matrix leaves the child at increased risk for more cartilage degradation whenever a flare occurs. Damage to the cartilage matrix is characterized by loss of proteoglycan and collagen, as demonstrated in studies focusing on rheumatoid arthritis [5, 6, 7, 8, 9]. This microstructural damage is important to recognize because intensification of anti-rheumatic treatment might prevent irreversible cartilage damage. Current imaging techniques, such as radiography, US imaging and conventional MRI, can detect structural bone damage and synovial inflammation but cannot detect microstructural damage to cartilage .
T1ρ is an MRI parameter that is hypothesized to quantify proteoglycan loss of the cartilage matrix in vitro  and in vivo . T1ρ uses a weak radiofrequency pulse to lock protons in phase, which slows transverse relaxation and reduces the effect of dipolar interactions. As a result, the measured relaxation of protons can be attributed to time-constant T1ρ. Earlier studies in osteoarthritis and rheumatoid arthritis [12, 13, 14, 15, 16] showed that T1ρ values increase as a consequence of decreased proteoglycan content. Cartilage integrity was studied once in pediatric patients using T1ρ: knee cartilage of 10 healthy children was assessed and a mean T1ρ value of 76.6 ms was reported . Cartilage in children is distinct from articular cartilage in adults. Postnatally, chondrocytes are small and there is a scant matrix without a distinct zonal orientation . Animal studies demonstrated that proteoglycan content is highest postnatally, with a gradual decrease in proteoglycans during aging, while collagen content increases and collagen fibers show increasing isotropy, resulting in development of a zonal organization of the cartilage [19, 20, 21].
The primary aim of this pilot study was to evaluate feasibility (motion artefacts and patient comfort) of T1ρ knee cartilage imaging in children with juvenile idiopathic arthritis; secondarily we studied repeatability of T1ρ values derived by manual cartilage segmentation. Moreover, we explored T1ρ values in juvenile idiopathic arthritis by comparing T1ρ values with conventional MRI scores using the juvenile arthritis MRI scoring system .
Materials and methods
In this pilot study, we included all consecutive pediatric patients who underwent MRI of the knee between April 2016 and August 2016 in the Academic Medical Center/Univerisity of Amsterdam, Amsterdam, The Netherlands. Patients visited one of the outpatient clinics of three tertiary pediatric rheumatology centers the Academic Medical Center/Univerisity of Amsterdam, OLVG hospital and Reade, all in Amsterdam, the Netherlands. Patients were clinically assessed by one of our pediatric rheumatologists J.M. van den Berg, D. Schonenberg-Meinema, A. Nassar-Sheikh Rashid and K.M. Dolman. Clinical characteristics consisted of age, height, weight, body mass index (adjusted for age and gender using body mass index percentiles and body mass index z-scores , calculated using the pediatric z-score calculator from the Children’s Hospital of Philadelphia ); a global assessment of disease activity measured on a 0–100 visual analogue scale by the physician; and an 84-joint count evaluating the number of actively inflamed joints. Laboratory tests included the erythrocyte sedimentation rate, which is the most widely used laboratory marker for juvenile idiopathic arthritis-related inflammation , and C-reactive protein.
Inclusion criteria for this pilot study were (1) clinically active arthritis (defined as joint swelling or limitation of motion, with pain or tenderness to palpation) involving at least one knee in children with juvenile idiopathic arthritis or suspected new-onset juvenile idiopathic arthritis; or (2) follow-up of children with juvenile idiopathic arthritis with clinically inactive disease who had a history of clinically evident arthritis in at least one knee.
Exclusion criteria were intra-articular corticosteroid injection within the last 6 months, the need for anesthesia during MRI examination and general contraindications for MRI. Written informed consent was obtained from all parents and, if the patient was 12 years or older, written informed consent was also obtained from the patient, as prescribed by local ethics regulations.
Magnetic resonance imaging protocol
Three-tesla magnetic resonance imaging sequences and parameters of our standard juvenile idiopathic arthritis MRI protocol and of the additional sagittal T1ρ
Repetition time (ms)
Echo time (ms)
Field of view
Voxel size (mm)
Slice thickness (mm)
Number of signal averages
Sagittal turbo spin-echo proton density
Sagittal turbo spin-echo 3-D T2 spectral attenuated inversion recovery
Sagittal 3-D turbo spin-echo T1 spectral attenuated inversion recovery
Sagittal 3-D turbo spin-echo T1 spectral attenuated inversion recovery
The conventional MR image sets were scored by an experienced reader R. Hemke (RH), with 7 years’ experience in musculoskeletal radiology) who was blinded to the clinical history of the patients. The reader scored the presence and extent of disease activity and osteochondral damage using the validated juvenile arthritis MRI scoring system, which has been described in detail .
After image acquisition the T1ρ images acquired at longer spin lock times were registered to the T1ρ images of the corresponding shortest spin lock time (5 ms) using an affine registration pipeline implemented in elastix . This was done to correct for small subject motion occurring between the different scans. The T1ρ relaxation maps were obtained by fitting the signal intensities of the registered images to a three-parameters exponential decay model in a pixel-by-pixel fashion.
The total articular cartilage of the knee was manually segmented into eight regions of interest by an experienced reader (RH, with 7 years’ experience in musculoskeletal radiology) on the T1ρ images acquired with spin lock time of 5 ms using ITK-SNAP . Segmentation was performed using the conventional axial, sagittal and coronal images to ensure accurate segmentation of the cartilage. The regions of interest reflected eight anatomical cartilage regions following the juvenile arthritis MRI scoring system . T1ρ values in milliseconds were extracted per region of interest and then combined, resulting in mean T1ρ values of the total articular cartilage per patient. Moreover, the regions of interest were sub-grouped into weight-bearing cartilage (medial and lateral tibial plateau, medial and lateral weight-bearing femur) or non-weight-bearing cartilage (medial and lateral patella, medial and lateral anterior part of condyle) resulting in mean T1ρ values for the weight-bearing cartilage and non-weight-bearing cartilage.
Reliability of repeated cartilage segmentation
To analyze the repeatability of the delineation of the articular knee cartilage on T1ρ, the same reader [blinded] performed the delineation process a second time in five randomly selected patients, 4 months after the first delineations.
All analyses were carried out using SPSS Statistics version 24.0 (IBM Corp., Armonk, NY). We studied normality of the data and whether the data had a non-Gaussian distribution; medians and interquartile range were reported and non-parametric tests were used for analysis. A P-value <0.05 was considered to indicate a statistically significant difference. First, descriptive characteristics of the patients were reported. Second, we studied the intra-reader reliability in five children who were segmented twice by the same reader using the intraclass correlation coefficient in a two-way mixed model with single measures. To further gain insight into T1ρ differences between the first and second segmentation, we created a Bland–Altman plot and calculated and the coefficient of repeatability (1.96 × standard deviation of the differences) . Next, we studied the difference between total mean T1ρ values and individual region-of-interest T1ρ values in MRI-active disease (defined as children with a juvenile arthritis MRI score ≥1) as compared to MRI-inactive disease (defined as children with a juvenile arthritis MRI score of 0) and we studied the difference in T1ρ values between weight-bearing and non-weight-bearing cartilage, all using Mann–Whitney U test. Subsequently, we studied correlation between the T1ρ value and disease activity parameters (juvenile arthritis MRI score, erythrocyte sedimentation rate, number of actively inflamed joints) with a Spearman ρ test. Last, we studied correlation between T1ρ values and age and body mass index of the patients with a Spearman ρ test.
Patient characteristics and disease activity markers of the study cohort
MRI-inactive patients (juvenile arthritis MRI score 0, n=6)
MRI-active patients (juvenile arthritis MRI score ≥1, n=7)
Body mass index percentile
Body mass index z-score
−0.74 (−1.18 to −0.17)
Number of females (%)
Active joints at physical examination
Joints with limitation of motion at physical examination
Erythrocyte sedimentation rate in mm/h
C-reactive protein in mg/L
Visual analogue scale physician
Total juvenile arthritis MRI score
Repeatability of segmentation-derived T1ρ values
T1ρ values in magnetic resonance imaging (MRI)-active versus MRI-inactive disease
Median (interquartile range) T1ρ values of individual cartilage regions of interest in milliseconds
T1ρ in MRI-inactive disease (ms)
T1ρ in MRI-active disease (ms)
Medial tibial plateau
Lateral tibial plateau
Medial weight-bearing femur
Lateral weight-bearing femur
Medial femoral condyle
Lateral femoral condyle
Weight-bearing cartilage versus non-weight-bearing cartilage
Overall, weight-bearing cartilage had a significantly lower T1ρ value compared to non-weight-bearing cartilage, with T1ρ values for weight-bearing cartilage of 29.5 ms (IQR 27.6–36.0) and 34.8 ms (IQR 33.7–40.4) for non-weight-bearing cartilage, P-value=0.04.
Correlation between T1ρ and disease activity parameters, age and body mass index
In this pilot study, we show the feasibility of T1ρ for assessing knee cartilage integrity in children with juvenile idiopathic arthritis. All 13 children in the study underwent the T1ρ acquisition protocol without discomfort and all images were of sufficient quality. None of the 13 children showed structural cartilage damage on conventional MRI. We found excellent repeatability for derivation of T1ρ values using manual cartilage segmentations on the T1ρ images.
Concerning the assessment of cartilage integrity, we found high correlation between T1ρ values and the juvenile arthritis MRI score, an MRI-based disease activity score, in the seven children who had inflammation in the knee, but we also observed that T1ρ values in children with actively inflamed knees were not different compared to T1ρ values in those with non-inflamed knees. Nevertheless, cartilage of the lateral patellar region of interest demonstrated significantly higher T1ρ values as compared to cartilage in children without knee inflammation on MRI. The patellar cartilage borders the patellofemoral synovium, which is often affected if knee arthritis is present . Correspondingly, in our study, 4 of the 7 children with active arthritis indeed showed inflamed patellofemoral synovium. Thus, we hypothesize that increased T1ρ values might represent pre-erosive microstructural damage to proteoglycans and collagen in the cartilage matrix that is not visualized using conventional MR sequences.
Several studies confirmed that T1ρ values can be used to detect macrostructural and microstructural damage to cartilage in osteoarthritis [34, 35, 36]. In rheumatoid arthritis, the use of T1ρ was first described in a study involving five people with rheumatoid arthritis . In this study of cartilage specimens after total knee arthroplasty, T1ρ values correlated with histological Safranin-O staining and macroscopic grade of severity of cartilage degeneration. In another study, radiocarpal cartilage was evaluated in a 3-month follow-up study of nine people with rheumatoid arthritis who used anti-rheumatic medication . T1ρ values correlated with treatment response, showing the potential of T1ρ to measure changes in cartilage structure following treatment. Our results seem comparable with the findings in both osteoarthritis and rheumatoid arthritis [10, 12, 37] in which people with more severe disease activity were found to have higher T1ρ values. When comparing the absolute T1ρ values, we found (lowest-to-highest) 31 ms to 55 ms. Values in people with rheumatoid arthritis have been found to be 38–62 ms , and values in healthy pediatric patients 66–77 ms . Note our values are somewhat lower. For the first comparison, this is probably attributable to the more severely affected cartilage in these children with rheumatoid arthritis who were scheduled for total knee arthroplasty. Another factor to take into account when comparing results from different studies is the spin lock frequency because T1ρ values are higher at increased spin lock frequency. Our scans were acquired at lower frequency (400 Hz) than the scans of people with rheumatoid arthritis and healthy pediatric subjects (both acquired at 500 Hz) [12, 17].
We found a correlation between the juvenile arthritis MRI score and T1ρ values as well as erythrocyte sedimentation rate and T1ρ values. This supports the hypothesis that inflammation in the knee negatively affects the cartilage. We could not confirm our hypothesis that increasing age, and thus lower proteoglycan content, leads to lower T1ρ values because we observed no correlation between age and T1ρ values in this small cohort. However, this could be influenced by the age dispersion in our cohort because all but one child was older than 10 years.
Concerning body mass index, literature shows contradictory results. A recent study found correlation between body mass index and T1ρ in the knee  while others decline a relation between body mass index and T1ρ values in hip cartilage and intervertebral disc cartilage, respectively [39, 40]. In our study, body mass index was not correlated to T1ρ values. It should be noted that our cohort consisted of mainly non-obese adolescents, thus we cannot exclude that age and body mass index could influence T1ρ values in, for example, a 4-year-old or heavily obese child.
Limitations of this study are that none of the children had structural cartilage damage on MRI, hence it was not possible to examine the T1ρ value in actual erosive cartilage damage. Moreover, histochemical proof of the hypothesized pre-erosive microstructural proteoglycan loss in the cartilage is lacking. However, obtaining cartilage specimens using biopsy is not feasible because this would harm the joints of these children. Furthermore, our T1ρ experiments were performed at a spin lock frequency of 400 Hz because of specific absorption rate limitations and the need to keep the acquisition time short enough. Therefore, the used spin lock preparation did not completely remove all the contributions of dipolar interactions to the relaxation process. As a consequence, our readout is not completely specific to proteoglycan content and likely also reflects changes in the collagen matrix, such as degradation or swelling. Additional studies are needed in order to decouple the two contributions and gain more insight into the biochemical modifications induced by the disease. Another limitation is our segmentation. Although the cartilage segmentations were performed meticulously by an experienced reader, we cannot rule out that the cartilage–bone and cartilage–soft-tissue boundaries were imperfect. This could have influenced our results, especially if possible fluid pixels from joint effusion were wrongly included in the segmentation. To prevent this, we used three imaging planes when drawing the segmentations. Second, we performed our segmentation on full-thickness articular cartilage. Because cartilage has a zonal orientation, it would be interesting to subdivide the cartilage into a superficial and deep layer and study spatial variation in more detail. This could, for example, be performed using a normalization procedure to correct for different cartilage thicknesses in children, as has been done by authors studying cartilage with T2 relaxation time mapping in healthy children and children with juvenile idiopathic arthritis [41, 42, 43, 44]. However, because our primary goal was to evaluate the feasibility of T1ρ, we did not perform such in-depth analyses of the cartilage; nevertheless, we would recommend a zonal analysis of the cartilage in studies that include a bigger sample of patients. Last, the small patient sample itself is considered a limitation and further work should focus on inclusion of more patients to validate the results of our pilot study. When more patients are included, the likelihood of scanning patients with structural cartilage damage would increase, which is important to affirm the assumption that higher T1ρ values are seen in structurally abnormal cartilage as seen on conventional, qualitative MRI.
There was acceptable feasibility and repeatability of T1ρ for assessing knee cartilage in our sample. We found some association between T1ρ values and erythrocyte sedimentation rate and the juvenile arthritis MRI score. In further studies, inclusion of more children — including children with erosive cartilage damage in the knee — is warranted to confirm the preliminary findings of this pilot study.
We acknowledge our colleagues for their valuable assistance in conducting this study: E.C. van Gulik, MD, D. Schonenberg-Meinema, MD, A. Nassar-Sheikh Rashid, MD, K.M. Dolman, MD, PhD. A.M.B. received an MD/PhD scholarship from the Academic Medical Center (AMC)/University of Amsterdam, the Netherlands.
Compliance with ethical standards
Conflicts of interest
- 24.Children’s Hospital of Philadelphia (2019) Pediatric Z-score calculator. https://zscore.research.chop.edu/index.php. Accessed 14 Sept 2019
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.