Intrarater and Interrater Reliability of the Quantification of Knee Cartilage MR Relaxation Metrics

doi:10.21203/rs.3.rs-4926999/v1

Background

Magnetic resonance (MR) imaging is often used to study osteoarthritis (OA), as advanced MR imaging methods can provide a quantitative assessment of tissue biochemistry or composition. For example, the magnetic relaxation times T_1ρ (i.e., 1/R_1ρ) and T₂ (i.e., 1/R₂) of water molecules within articular cartilage have been demonstrated to be imaging biomarkers sensitive to the compositional changes associated with early OA. However, the outcome of MR imaging data analysis depends on relaxation data acquisition methods as well as assessor variability if manual segmentation is performed. Therefore, the goal of the current study was to evaluate the intra- and interrater reliability of established imaging protocols for performing quantitative cartilage MR relaxation metrics of the knee joint.

Methods

Right knee MR images were obtained from five healthy individuals (average age, 24.4 years; 3 females) via a 3.0T MRI scanner equipped with a 16-channel knee T/R coil. A double echo steady state (DESS) sequence was used for anatomical imaging, and the established MAPSS sequences were used for R_1ρ and R₂ mapping. One assessor performed manual segmentations of the knee cartilage on two separate occasions, whereas a second assessor performed segmentations once. Both the R_1ρ and R₂ mean values were then calculated for the tibial, patellar, femoral trochlear, central femoral condylar, and posterior femoral condylar cartilages. Intraclass correlation coefficients [ICC (3,1)] and ICCs (2,1) were used to evaluate intra- and interrater reliability, respectively. The standard error of measurement (SEM) was used to assess absolute reliability.

Results

The intrarater knee cartilage relaxation metrics demonstrated good to excellent reliability, ranging between 0.88 and 0.99, with SEMs ranging between 0.16 and 0.80. The interrater reliability similarly ranged from 0.79–0.97, with SEMs ranging between 0.27 and 1.10.

Conclusions

Manual segmentation of specific MR slices and known subregions is highly reliable and repeatable for the quantification of cartilage MR relaxation metrics. This validation paves the way for the large-scale application of this method in prospective trials that longitudinally monitor OA development and progression in the knee joint.

knee cartilage

reliability

quantitative MRI

Osteoarthritis (OA) is a common joint disorder in middle-aged and elderly people who significantly affects joint functionality, causing disability and increased healthcare utilization. The knee joint is the most commonly affected joint, with an estimated prevalence of 365 million people worldwide, followed by the hip and hand [1]. Knee OA impacts both tibiofemoral and patellofemoral joints, with patellofemoral OA being highly prevalent [2–4], but the latter is often underrecognized. A radiographic study revealed that among adults aged 50 years and above with knee pain, 40% had combined tibiofemoral and patellofemoral OA, 24% had isolated patellofemoral OA, and 4% had isolated tibiofemoral OA [2].

The current gold standard for diagnosing OA, in addition to routine clinical examination, is plain X-ray [5]. This method is safe, cost-effective, and widely available; however, it is known to be insensitive to early OA changes. Since clinical OA represents a late-stage condition with limited opportunities for disease intervention, identifying and characterizing OA in its earlier stages could enhance our understanding of OA mechanisms and aid in the development and evaluation of disease-modifying treatments.

In the early stages of OA, articular cartilage fails to synthesize extracellular matrix components, leading to a loss of proteoglycans and collagen, reduced elasticity, and decreased water content [6]. Quantitative MR imaging provides noninvasive tools for assessing the biochemical composition of articular cartilage, allowing for the characterization of cartilage integrity before irreversible damage occurs. Over the past two decades, studies have shown that water proton relaxation times T_1ρ and T₂ can provide valuable information on the proteoglycan content and collagen orientation within cartilage [7, 8]. These relaxation metrics have been reported to be effective in distinguishing between individuals with and without knee OA [9–12].

Cartilage segmentation is a critical factor influencing the quantification of MR relaxation metrics [13]. Recent advances in artificial intelligence (AI) have significantly improved cartilage segmentation analysis [14–17]. Despite its growing popularity, AI-based segmentation is not yet commercially or readily accessible to all institutions, making manual segmentation valuable due to its inherent benefits. Manual segmentation offers superior accuracy and precision and remains the gold standard for validating and training deep learning models. However, a major downside of manual cartilage segmentation is the potential for variability between different assessors and even within the same assessor over time.

The reproducibility of compositional MR relaxation metrics could be compromised by various sources, ranging from data acquisition to postprocessing, as previously documented in the literature [18, 19]. With the establishment of standardized compositional MR imaging acquisitions [20], significant attention should be directed toward data postprocessing. Therefore, the aim of this study was to examine the intra- and interrater reliability of knee cartilage MR relaxation metrics via manual segmentation by end users.

Subjects

A sample of five healthy individuals volunteered for the reliability analysis (average age, 24.4 ± 1.3 years; 3 females). The exclusion criteria were (1) any preexisting musculoskeletal diseases of the lower extremity, (2) any past surgeries of the lower extremity, and (3) any contraindications to MR imaging, such as the presence of a cardiac pacemaker, metal implants, intraocular foreign body, aneurysm clips, pregnancy, claustrophobia, etc. The study was approved by the University of Michigan Institutional Review Boards (IRB) (approval number: HUM00200085). All procedures followed were in accordance with the Declaration of Helsinki, ensuring that the rights and well-being of participants were protected. Prior to data collection, all participants signed written informed consent forms, acknowledging their understanding of the study and their voluntary participation.

MR Imaging Protocols

All participants were scanned on the right knee in the supine position via a Philips Ingenia 3.0T wide-bore MRI scanner with a 16-channel knee transmit-receive coil. All the images were acquired in the sagittal plane. The double echo steady state (DESS) sequence was used for anatomical imaging with a field-of-view (FOV) of 140*140*112 mm³ and an acquired voxel resolution of 0.4*0.5*0.9 mm³, which was interpolated to 0.36*0.36*0.63 mm³ when 3D volumetric images were reconstructed. The number of reconstructed imaging slices was 178, and the scan time was 5:52 minutes. R_1ρ and R₂ mappings were obtained from the established MAPSS sequences [19, 21] implemented by the vendor as working-in-progress (WIP) prototypes. The FOV was 140*140*96 mm^3, and the acquired or reconstructed voxel size was 0.44*0.66*4.0 or 0.27*0.27*4.0 mm³. For R_1ρ mapping, the spin‒lock durations (TSLs) were 0, 10, 15, 20, 25, 30, 35, and 40 ms, and the spin‒lock strength was 500 Hz. For R₂ mapping, the echo times (TEs) were 0, 18, 36, and 54 ms. The scan duration for R_1ρ or R₂ mapping was 9:13 minutes.

MR Imaging Processing

Manual cartilage segmentations were performed on the reformatted DESS images via ITK-SNAP 4.0 software [22]. R_1ρ− and R₂-weighted images were coregistered, as documented in the literature [23], with the reformatted DESS images as the reference. The reformatted DESS images had the same slice thickness (4 mm) as the R_1ρ− and R₂-weighted images. The regions of interest (ROIs) included the articular cartilages of the tibia (T), patella (P), femoral trochlea (TrF), central femoral (cF) condyle, and posterior femoral (pF) condyle (Fig. 1). In the sagittal images, the tibial, patellar, and femoral trochlear cartilages were identified by tracing the boundaries on the basis of their anatomical landmarks, slice by slice. All slices that displayed cartilage underwent segmentation, except those with partial cartilage volume averaging at the margins.

As the cartilaginous surface of the femoral cartilage is continuous, artificial boundaries had to be defined for separating the ROIs (Fig. 1). Only the weight-bearing portion of the femoral cartilage was considered and further classified into the cF and pF regions. The former was separated from the trochlear plane, and the latter was separated in the coronal plane. The anterior boundary was defined as the first image where the continuous trochlear cartilage layer separated into distinct cartilage layers on the femoral condyles (Fig. 2a). The posterior boundary was defined as 60% of the distance between the anterior boundary and the last image containing the posterior aspects of the femoral condyles, known as “double bulls-eye”, containing both cartilage and bone (Fig. 2b) [24, 25]. These landmarks were used because they could be clearly identified and reproduced in the subjects. While the segmentation was performed in the sagittal images and the landmarks were identified in the coronal images, the boundaries could be defined by using the same landmarks and 3D viewing during segmentation.

The signal intensities (\(\:S\)) in R_1ρ− or R₂₋weighted images were fitted via home-grown IDL 8.9 (Interactive Data Language, Harris Geospatial Solutions, Inc., Broomfield, CO, USA) programming scripts [23], voxel by voxel, to a simple exponential decay relaxation model, i.e., \(\:S=C*{e}^{-\left({R}_{1\rho\:}\:*\:TSL\right)}\:\)or \(\:\:S=C*{e}^{-\left({R}_{2}\:*\:TE\right)}\), where \(\:C\) is a fitted constant. The average R_1ρ or R₂ relaxation rate (in units of 1/s) was subsequently calculated from the fits for cartilage in different ROIs of the knee joint.

Statistical Analysis

Intra- and interrater reliability statistics were calculated for the R_1ρ and R₂ values of knee cartilage. Intrarater reliability was obtained for one rater (ML) who segmented each scan twice on different occasions. Interrater reliability was obtained for two raters (ML and JH), who individually segmented the knee cartilages for each scan. Intraclass correlation coefficients [ICC (3,1)] and ICC (2,1) were used to evaluate intra- and interrater reliability for each ROI for R_1ρ and R₂, respectively [26]. The standard error of measurement (SEM) was used to assess absolute reliability. ICCs less than 0.5 indicated poor reliability; 0.5–0.75, moderate reliability; 0.75–0.9, good reliability; and greater than 0.9, excellent reliability [27]. Bland‒Altman limits of agreement were calculated with an exact 95% confidence interval (CI). The data collected from this study were analyzed via SPSS 28.0.

Figures 3 and 4 show representative images of R₁_ρ-and R_2-weighted images from one subject at various TSL and TE values, respectively. Tables 1 and 2 provide the R₁_ρand R₂relaxation rates in each ROI as estimated from each rater and session. The tibia cartilage presented the greatest R₁_ρand R₂relaxation rates, followed by the femoral cartilage (if all subregions were considered together), whereas the patella cartilage presented the lowest rates.

The intrarater consistency in the segmented ROI areas of the knee R₁_ρ relaxation rates demonstrated that the ICC values ranged between 0.84 and 0.99, and the results were similar for R₂ values ranging between 0.91 and 0.99. The SEM ranged between 0.22 and 0.57 and between 0.21 and 0.74 for R₁_ρ and R₂, respectively (Table 1). For interrater consistency, the ICCs for the R₁_ρ values ranged between 0.80 and 0.99, and the R₂ values ranged between 0.75 and 0.99. Finally, the SEM for interrater reliability ranged from 0.34--1.10 and 0.27--0.84 for R₁_ρ and R₂, respectively (Table 2). For the intrarater analysis, Bland‒Altman agreement revealed a mean difference in relaxation rates of 0.27 s^-1, the limits of agreement were -1.23 and 1.78, and the outer 95% CIs were -0.49 and 0.055 (Figure 5a). For the interrater analysis, the mean difference was 0.81 s^-1, the limits of agreement were -1.50 and 3.13, and the 95% CIs were -1.15 and 0.47 (Figure 5b).

While automatic cartilage segmentation through AI has been reported, manual segmentation remains the gold standard, offering superior accuracy and precision. The goal of this study was to examine the intra- and interrater reliability of knee cartilage MR relaxation metrics via manual cartilage segmentation. Our findings suggested good to excellent intra- and interrater reliability for the quantification of knee MR relaxation metrics, with slightly lower interrater reliability than intrarater reliability. Consistent with the literature [14, 28], tibia cartilage presented the greatest R_1ρ and R₂ relaxation rates, followed by femoral cartilage, whereas patella cartilage presented the lowest rates.

In our study, reliability was greater in the patellar, trochlear, and central femoral condylar cartilages than in the tibia and posterior femoral condylar cartilages. Since our segmentations were based on anatomical identification, the reproducibility of these landmarks became crucial in determining the boundaries for segmentation. Patellar and femoral trochlear cartilages, which comprise the patellofemoral joint, can be easily visualized on MR images with distinct cartilage-to-bone boundaries (Fig. 1). Additionally, the patellar cartilage, followed by the trochlear cartilage, is thicker than the other knee cartilages. Previous studies have suggested that the thicker the cartilage is, the lower the observed variability [29, 30], which is consistent with the results of the present study. However, posterior femoral condylar cartilage exhibited the greatest variability, which can be attributed to observer-dependent differences in the determination of the artificial boundaries, both anteriorly and posteriorly. Nevertheless, the overall high intra- and interrater reliability (ICCs ranging between 0.75 and 0.99) suggests that subtle differences in cartilage selection in specific regions do not meaningfully alter the mean R_1ρ and R₂ relaxation rates.

Since our cartilage segmentation was reproducible, the mean R_1ρ and R₂ relaxation rates were expected to demonstrate good to excellent intra- and interrater reliability, which is consistent with the literature. In the study by William et al. [30], the authors reported intrarater ICCs ranging between 0.80 and 0.98 for T₂ mapping in the knee joint, whereas the current study reported intrarater ICCs ranging between 0.91 and 0.99 among the tibiofemoral and patellofemoral joints. Another study by Welsch et al. [31] reported an average interrater ICC of 0.90 for T₂ mapping in the knee joint, whereas our interrater agreement, though slightly lower, was still considered excellent, with ICCs ranging between 0.75 and 0.99. Collectively, the evidence suggests that manual segmentation of specific MR slices and known subregions is highly reliable.

While AI-based segmentation has the advantage of eliminating the manual segmentation burden, its overall reliability and accuracy remain unclear. Norman et al. [32] utilized a deep learning model to perform automatic segmentation via DESS sequences and reported Dice coefficients ranging from 0.77 to 0.88 for cartilage. Similarly, Gatti and Maly [16] used knee MR images and reported Dice coefficients ranging from 0.88 to 0.91 for different cartilage compartments, with even better results for healthy cartilage. The Dice coefficient, a statistical tool that measures the agreement of segmentations, with 1 being a perfect match and 0.8 to 0.9 generally considered good and useful [13]. While the reliability of AI-based segmentation appears to be comparable with that of manual segmentation, it is important to note that AI-based segmentation was validated against manual analysis, which resulted in a doubling of the errors. Importantly, reliability does not necessarily imply accuracy. Since the true cartilage boundaries for the samples in this study are currently unknown, it is unclear how accurately these MR relaxation metrics are reflected. Future studies will need to address this critical gap for both manual and AI-based segmentations to enhance the assessment of degenerative joint diseases.

Even though this study is limited by having only two raters and five image sets, the results likely reflect the true reliability of generating knee MR cartilage relaxation metrics from manual segmentations of the respective ROIs. However, it is important to note that the study included only healthy subjects. Individuals with OA may have bone spurs and cartilage defects that could complicate the accurate outlining of the cartilage surface. Additionally, this study did not examine the zonal behavior of cartilage relaxation metrics, which have been shown to differ within cartilage layers due to varying loads [31, 33]. Future research should include subjects with OA and assess the zonal behavior of cartilage relaxation metrics to improve our understanding of cartilage health and pathology.

In conclusion, our analysis results suggest that manually segmenting specific MR slices and known subregions is highly reliable and repeatable for the quantification of MR cartilage relaxation metrics, albeit at the cost of human manpower. Our findings contribute to the existing body of literature and indicate that manual segmentation of knee cartilage demonstrates excellent intra- and interrater reliability, with patellofemoral joint assessment being superior to tibiofemoral joint assessment. Additionally, this validation paves the way for the large-scale application of this method in prospective trials that longitudinally monitor the development and progression of knee osteoarthritis in individuals at risk of further developing the disease.

AI: artificial intelligence; MR: magnetic resonance; OA: osteoarthritis

Ethics Approval and Consent to Participate

The study was approved by the University of Michigan Institutional Review Boards, and prior to data collection, all the subjects signed a written informed consent form (HUM00200085).

Consent for Publication: Not applicable.

Availability of Data and Materials

The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Competing Interests:

The authors declare that they have no competing interests.

Funding

Funding was obtained from the College of Health Sciences, University of Michigan-Flint.

Authors' Contributions

TCL and YP secured funding for the study. All the authors were actively involved in subject recruitment and data collection. YP, ML, and JH were instrumental in processing the MR data. TCL conducted the statistical analysis and contributed significantly to the interpretation of the data. Additionally, all authors participated in reviewing the manuscript, providing critical feedback and approval of the final version.

Acknowledgments: Not applicable.

Long H, Liu Q, Yin H, Wang K, Diao N, Zhang Y, Lin J, Guo A: Prevalence trends of site-specific osteoarthritis from 1990 to 2019: findings from the Global Burden of Disease Study 2019. Arthritis Rheumatol 2022, 74(7):1172-1183.
Duncan RC, Hay EM, Saklatvala J, Croft PR: Prevalence of radiographic osteoarthritis--it all depends on your point of view. Rheumatology (Oxford, England) 2006, 45(6):757-760.
McAlindon TE, Snow S, Cooper C, Dieppe PA: Radiographic patterns of osteoarthritis of the knee joint in the community: the importance of the patellofemoral joint. Annals of the rheumatic diseases 1992, 51(7):844-849.
Szebenyi B, Hollander AP, Dieppe P, Quilty B, Duddy J, Clarke S, Kirwan JR: Associations between pain, function, and radiographic features in osteoarthritis of the knee. Arthritis and rheumatism 2006, 54(1):230-235.
Kellgren JH, Lawrence JS: Radiological assessment of osteo-arthrosis. Ann Rheum Dis 1957, 16(4):494-502.
Cucchiarini M, de Girolamo L, Filardo G, Oliveira JM, Orth P, Pape D, Reboul P: Basic science of osteoarthritis. J Exp Orthop 2016, 3(1):22.
Link TM, Neumann J, Li X: Prestructural cartilage assessment using MRI. J Magn Reson Imaging 2017, 45(4):949-965.
Regatte RR, Akella SV, Lonner JH, Kneeland JB, Reddy R: T1rho relaxation mapping in human osteoarthritis (OA) cartilage: comparison of T1rho with T2. Journal of magnetic resonance imaging : JMRI 2006, 23(4):547-553.
Baum T, Joseph GB, Arulanandan A, Nardo L, Virayavanich W, Carballido-Gamio J, Nevitt MC, Lynch J, McCulloch CE, Link TM: Association of magnetic resonance imaging-based knee cartilage T2 measurements and focal knee lesions with knee pain: data from the Osteoarthritis Initiative. Arthritis Care Res (Hoboken) 2012, 64(2):248-255.
Joseph GB, Baum T, Alizai H, Carballido-Gamio J, Nardo L, Virayavanich W, Lynch JA, Nevitt MC, McCulloch CE, Majumdar S et al: Baseline mean and heterogeneity of MR cartilage T2 are associated with morphologic degeneration of cartilage, meniscus, and bone marrow over 3 years--data from the Osteoarthritis Initiative. Osteoarthritis Cartilage 2012, 20(7):727-735.
Li X, Benjamin Ma C, Link TM, Castillo DD, Blumenkrantz G, Lozano J, Carballido-Gamio J, Ries M, Majumdar S: In vivo T(1rho) and T(2) mapping of articular cartilage in osteoarthritis of the knee using 3 T MRI. Osteoarthritis and cartilage 2007, 15(7):789-797.
Atkinson HF, Birmingham TB, Moyer RF, Yacoub D, Kanko LE, Bryant DM, Thiessen JD, Thompson RT: MRI T2 and T1ρ relaxation in patients at risk for knee osteoarthritis: a systematic review and meta-analysis. BMC Musculoskelet Disord 2019, 20(1):182.
Link TM, Joseph GB, Li X: MRI-based T(1rho) and T(2) cartilage compositional imaging in osteoarthritis: what have we learned and what is needed to apply it clinically and in a trial setting? Skeletal Radiol 2023, 52(11):2137-2147.
Razmjoo A, Caliva F, Lee J, Liu F, Joseph GB, Link TM, Majumdar S, Pedoia V: T(2) analysis of the entire osteoarthritis initiative dataset. J Orthop Res 2021, 39(1):74-85.
Gaj S, Yang M, Nakamura K, Li X: Automated cartilage and meniscus segmentation of knee MRI with conditional generative adversarial networks. Magn Reson Med 2020, 84(1):437-449.
Gatti AA, Maly MR: Automatic knee cartilage and bone segmentation using multistage convolutional neural networks: data from the osteoarthritis initiative. Magma 2021, 34(6):859-875.
Liu F, Zhou Z, Jang H, Samsonov A, Zhao G, Kijowski R: Deep convolutional neural network and 3D deformable approach for tissue segmentation in musculoskeletal magnetic resonance imaging. Magnetic Resonance in Medicine 2018, 79(4):2379-2391.
Matzat SJ, McWalter EJ, Kogan F, Chen W, Gold GE: T2 Relaxation time quantitation differs between pulse sequences in articular cartilage. Journal of Magnetic Resonance Imaging 2015, 42(1):105-113.
Kim J, Mamoto K, Lartey R, Xu K, Nakamura K, Shin W, Winalski CS, Obuchowski N, Tanaka M, Bahroos E et al: Multivendor multisite T(1ρ) and T(2) quantification of knee cartilage. Osteoarthritis Cartilage 2020, 28(12):1539-1550.
Chalian M, Li X, Guermazi A, Obuchowski NA, Carrino JA, Oei EH, Link TM: The QIBA profile for MRI-based compositional imaging of knee cartilage. Radiology 2021, 301(2):423-432.
Lartey R, Nanavati A, Kim J, Li M, Xu K, Nakamura K, Shin W, Winalski CS, Obuchowski N, Bahroos E et al: Reproducibility of T1ρ and T2 quantification in a multivendor multisite study. Osteoarthritis and Cartilage 2023, 31(2):249-257.
Yushkevich PA, Piven J, Hazlett HC, Smith RG, Ho S, Gee JC, Gerig G: User-guided 3D active contour segmentation of anatomical structures: significantly improved efficiency and reliability. Neuroimage 2006, 31(3):1116-1128.
Pang Y, Palmieri-Smith RM, Maerz T: An efficient R(1ρ) dispersion imaging method for human knee cartilage using constant magnetization prepared turbo-FLASH. NMR Biomed 2021, 34(6):e4500.
Eckstein F, Hudelmaier M, Wirth W, Kiefer B, Jackson R, Yu J, Eaton CB, Schneider E: Double echo steady state magnetic resonance imaging of knee articular cartilage at 3 Tesla: a pilot study for the Osteoarthritis Initiative. Ann Rheum Dis 2006, 65(4):433-441.
Schneider E, Nevitt M, McCulloch C, Cicuttini FM, Duryea J, Eckstein F, Tamez-Pena J: Equivalence and precision of knee cartilage morphometry between different segmentation teams, cartilage regions, and MR acquisitions. Osteoarthritis Cartilage 2012, 20(8):869-879.
Shrout PE, Fleiss JL: Intraclass correlations: Uses in assessing rater reliability. Psychological Bulletin 1979, 86(2):420-428.
Koo TK, Li MY: A guideline of selecting and reporting intraclass correlation coefficients for reliability research. J Chiropr Med 2016, 15(2):155-163.
Pedoia V, Lee J, Norman B, Link TM, Majumdar S: Diagnosing osteoarthritis from T(2) maps using deep learning: an analysis of the entire Osteoarthritis Initiative baseline cohort. Osteoarthritis Cartilage 2019, 27(7):1002-1010.
Newbould RD, Miller SR, Toms LD, Swann P, Tielbeek JA, Gold GE, Strachan RK, Taylor PC, Matthews PM, Brown AP: T2* measurement of the knee articular cartilage in osteoarthritis at 3T. J Magn Reson Imaging 2012, 35(6):1422-1429.
Williams A, Qian Y, Chu CR: UTE-T2∗ mapping of human articular cartilage in vivo: a repeatability assessment. Osteoarthritis Cartilage 2011, 19(1):84-88.
Welsch GH, Apprich S, Zbyn S, Mamisch TC, Mlynarik V, Scheffler K, Bieri O, Trattnig S: Biochemical (T2, T2* and magnetization transfer ratio) MRI of knee cartilage: feasibility at ultrahigh field (7T) compared with high field (3T) strength. Eur Radiol 2011, 21(6):1136-1143.
Norman B, Pedoia V, Majumdar S: Use of 2D U-Net convolutional neural networks for automated cartilage and meniscus segmentation of knee MR imaging data to determine relaxometry and morphometry. Radiology 2018, 288(1):177-185.
Banjar M, Horiuchi S, Gedeon DN, Yoshioka H: Review of quantitative knee articular cartilage MR imaging. Magn Reson Med Sci 2022, 21(1):29-40.

Table 1. Intra-rater reliability for R₁_ρ and R₂(1/s) cartilage relaxation rates.

Regions of Interest	Trial 1 mean (SD)	Trial 2 mean (SD)	Pearson Correlation (r)	ICC (3,1) (95% CI)	SEM
R₁_ρ (1/s)
Central Femoral	28.57 (3.40)	28.71 (3.44)	0.97*	0.98 (0.86 – 0.99)*	0.46
Posterior Femoral	26.29 (1.22)	26.91 (1.18)	0.84	0.91 (0.16 – 0.99)*	0.35
Tibia	27.41 (2.06)	27.29 (1.42)	0.92*	0.92 (0.27 – 0.99)*	0.47
Patella	20.33 (2.27)	20.33 (2.40)	0.99*	0.99 (0.94 – 0.99)*	0.22
Trochlea	26.84 (2.13)	27.09 (3.16)	0.99*	0.95 (0.56 – 0.99)*	0.57
R₂(1/s)
Central Femoral	32.52 (4.94)	32.73 (5.31)	0.99*	0.99 (0.96 – 1.00)*	0.48
Posterior Femoral	29.46 (2.25)	29.83 (2.21)	0.99*	0.99 (0.98 – 1.00)*	0.21
Tibia	34.09 (3.49)	34.69 (3.54)	0.91*	0.95 (0.57 – 0.99)*	0.74
Patella	26.84 (2.42)	27.05 (2.58)	0.99*	0.99 (0.95 – 1.00)*	0.24
Trochlea	32.80 (2.40)	33.26 (2.83)	0.97*	0.97 (0.77 – 0.99)*	0.43

CI: confidence interval; CV: coefficient of variation; ICC: intra-class correlation coefficient; SEM: standard error of measurement. * indicates p<0.05

Table 2. Inter-rater reliability for R₁_ρ and R₂(1/s)cartilage relaxation rates.

Regions of Interest	Rater 1 mean (SD)	Rater 2 mean (SD)	Pearson Correlation (r)	ICC (2,1) (95% CI)	SEM
R₁_ρ (1/s)
Central Femoral	28.57 (3.40)	29.82 (4.37)	0.92*	0.93 (0.45 – 0.99)*	0.98
Posterior Femoral	26.29 (1.22)	27.60 (1.90)	0.95*	0.79 (-0.26 – 0.97)*	0.74
Tibia	27.41 (2.06)	27.81 (1.23)	0.93*	0.90 (0.26 – 0.99)*	0.50
Patella	20.33 (2.27)	21.00 (2.54)	0.99*	0.97 (0.07 – 0.99)*	0.34
Trochlea	26.84 (2.13)	27.96 (3.32)	0.80	0.83 (-0.12 – 0.98)*	1.10
R₂(1/s)
Central Femoral	32.52 (4.94)	33.34 (5.43)	0.95*	0.97 (0.79 – 0.99)*	0.84
Posterior Femoral	29.46 (2.25)	30.29 (1.75)	0.93*	0.91 (0.23 – 0.99)*	0.56
Tibia	34.09 (3.49)	34.68 (3.00)	0.98*	0.98 (0.81 – 0.99)*	0.43
Patella	26.84 (2.42)	27.27 (2.75)	0.99*	0.98 (0.81 – 0.99)*	0.27
Trochlea	32.80 (2.40)	33.52 (1.87)	0.75	0.84 (-0.18 – 0.98)	0.83

CI: confidence interval; CV: coefficient of variation; ICC: intra-class correlation coefficient; SEM: standard error of measurement. * indicates p<0.05

No competing interests reported.

oldTABLE2.docx

Intrarater and Interrater Reliability of the Quantification of Knee Cartilage MR Relaxation Metrics

Status:

Version 1

Abstract

Background

Methods

Results

Conclusions

Figures

Introduction

Methods

Subjects

MR Imaging Protocols

MR Imaging Processing

Statistical Analysis

Results

Discussion

Conclusions

Abbreviations

Declarations

References

Tables

Additional Declarations

Supplementary Files

Status:

Version 1