Introduction

Cartilage affected by progressive osteoarthritis (OA) has proteoglycan loss and collagen fiber changes that act as surrogate biomarkers reflecting disease severity—these can be measured using quantitative magnetic resonance (MR) techniques such as contrast-enhanced T1 and mapping of T1ρ and T2 [1,2,3]. Of these techniques, MR T2 measurements are ideal for clinical applications because they do not require injection of contrast agents and advanced sequence programming. Previous studies have demonstrated that T2 values are reflective of subtle changes in water content and orientation of collagen fibers in knee cartilage; these indicators are signals of cartilage degradation [4] and correlate with OA severity [5].

The T2* relaxation time reveals additional information related to local field inhomogeneity, and these may be more sensitive to changes in tissue composition compared with T2 relaxation time [6]. Recently, T2* measurements have been used to further explore T2 relaxation and the local susceptibility effect on knee cartilage [7]. Indeed, T2* imaging can evaluate hip joint cartilage at 1.5 T with shorter acquisition times than T2 mapping using gradient-recalled echo (GRE) MR imaging [8]. Mamisch et al demonstrated that both T2 and T2* values have similar responses in the assessment of articular cartilage and cartilage repair tissues [9]; however, the relationship between altered T2* values and cartilage degradation remains controversial because previous studies used discrepant imaging parameters [6, 10]. The MR T2* measurements have revealed alterations in cartilage composition—as well as microstructural characteristics—which support its potential utility in T2* imaging for early OA detection.

The composition and orientation of collagen fibers in knee cartilage are location dependent and are subject to various biomechanical stresses. T2 values vary significantly between locations in the cartilage in both unaffected knees and knees affected by OA because of the magic angle effect [11, 12]. These regional discrepancies in cartilage T2* measurements can influence the diagnostic value of these tests. Moreover, there are significant changes in the proteoglycan content as a function of age and sex [13, 14]. Although the reason remains unclear, OA is epidemiologically more prevalent in women than in men [15]. Therefore, it might be useful to derive the regional MR T2* values in knee cartilage relative to age and sex.

There is a high prevalence of OA in the weight-bearing tibiofemoral (TF) joint in Asian populations [16, 17]. However, relatively few studies have investigated knee cartilage using fast T2* measurements [18]. The purpose of this study was to determine the effects of age, sex, and location on MR T2* values in TF joint cartilage in asymptomatic people and to characterize the early changes in T2* values as a function of age. In addition, we compared the T2* values between patients with early OA and age- and sex-matched asymptomatic controls.

Methods

Patient enrollment

Sixty unaffected people were enrolled as approved by the institutional review board of Taipei Medical University in Taipei, Taiwan (approval number 201108018); all participants provided informed consent. Inclusion criteria included the following: (1) body mass index (BMI) less than 30 kg/m2 [19]; (2) asymptomatic with normal Lequesne indices [20] of 0 for both knees; (3) no evidence of meniscal tears or meniscal intra-substance fluid on MR imaging; (4) no evidence of ligamentous abnormalities on MR imaging; and (5) no loss of the meniscus or a discoid meniscus. The Lequesne index is a verified questionnaire given to a patient to evaluate if there is any knee discomfort associated with knee osteoarthritis. It comprises five questions relevant to knee pain or discomfort including maximum distance walked and activities of daily living [20].

Here, a meniscal tear was defined as an abnormal linear or complex high signal intensity communicating with the articular surface [21]. The exclusion criteria for unaffected controls were (1) age of less than 20 years; (2) presence of OA risk factors; (3) history of knee injury, obesity, high-intensity exercise or sports, or loss of knee stability; (4) history of non-OA knee arthropathy, knee surgery, or chronic disease; (5) long-term medication or nutritional supplement use; (6) poor image quality; and (7) abnormal findings on MR images (e.g., cruciate ligament tears, meniscal tears, synovitis). In total, 60 normal participants fulfilled the study criteria (Fig. 1).

Fig. 1
figure 1

Schematic representation of study participant selection

Full size image

The institutional review board also approved the participation of OA patients and waived the need for informed consent due to the retrospective nature of the study (approval number N201704004). Patients with early OA symptoms were referred from a single orthopedic surgeon who performed routine standardized physical examinations and applied a MR T2* knee protocol between January 2012 and December 2015. These patients were diagnosed based on the American College of Rheumatology Classification Criteria of 1986 and the more recent European League Against Rheumatism recommendations of 2010 [22, 23], and then patients were eligible for inclusion if they met all of the following three criteria for definite early OA: pain in the knee, a Kellgren-Lawrence grade not greater than 2, and arthroscopic or MRI findings demonstrating degenerative changes of the knee [24]. Furthermore, to avoid the inclusion of definite OA, patients with the joint space width less than 3 mm were excluded [25].

All participants received three X-ray projections of their knees, including anteroposterior (AP) and lateral standing views and skyline (Merchant) view, as routine radiographic techniques. The Kellgren-Lawrence grading system was used to confirm diagnoses. This system considers several radiographic features of OA including joint space narrowing and osteophyte development [26]. Inclusion criteria included (1) aged more than or equal to 41 years, (2) diagnosis of tibia-femoral knee OA by radiography based on a Kellgren-Lawrence grade of 1 or 2, and (3) at least two episodes of symptomatic knee joint pain rated greater than 3 on a 0–10 visual analog scale for a period of 10 days.

Any patient with a history of systemic autoimmune rheumatoid disease, traumatic knee injury, septic arthritis, intra-articular fracture involving a knee joint, knee arthroscopy, meniscal extrusion and maceration, knee malalignment, BMI greater than 30 kg/m2, or whose MR images showed lateral cartilage degeneration and/or meniscal tears was excluded because medial knee OA is more prevalent than lateral OA [27]. Knee angles were obtained based on the effective measurements using AP knee radiographs [28]. Patients with malalignment of greater than 5° in the valgus or varus direction were excluded. In total, 40 OA patients (20 male) were enrolled (Fig. 1).

All 100 participants were evaluated for body height and mass, and each received a routine MR examination before undergoing the T2* imaging. In addition, each completed the Lequesne index questionnaire and was then evaluated by a trained interviewer using identical validated Lequesne index questionnaires. For the knee, the Lequesne index questionnaire consists of 10 items spanning three scales. The largest consists of five items and evaluates pain or discomfort. The smallest consists of one item and evaluates the maximum distance walked. The third scale consists of four items and evaluates the activities of daily living. Each scale provides a score from 0 (no pain or functional limitation) to 8 (extreme pain and functional limitation) resulting in a total score from 0 to 24—this is a direct aggregate of symptoms and functions as a singular global index.

The MR T2* values were determined for the right knee in the 60 asymptomatic controls who were then divided by age. Group 1 was aged 21 to 40 years and had a mean (± standard deviation [SD]) BMI of 22.0 ± 3.1 kg/m2. Group 2 was aged 41 to 60 years, and BMI was 22.5 ± 2.9 kg/m2. Group 3 was aged more than or equal to 61 years, and BMI was 25.0 ± 3.8 kg/m2. The age- and sex-matched controls (a combination of group 2 and group 3) were compared to the 40 OA patients (aged more than or equal to 41 years, and BMI was 24.6 ± 3.9 kg/m2). Although only controls above 40 years were used for further comparisons with OA patients, the asymptomatic participants aged 21 to 40 years were included for assessing the age effect on the T2* measurements. Detailed information about participant characteristics is shown in Table 1.

Table 1 Participant characteristics
Full size table

Data acquisition

All data were acquired on a 1.5-T clinical MR scanner (Magnetom Avanto; Siemens Healthineers). The right knee of each participant was centered in a single-channel knee coil (Siemens Healthineers). The “magic angle effect” on the cartilage T2* measurement was minimized [29] by straightening the leg so that the long axis was parallel to the main magnetic field (B0). The leg was then immobilized using an MR-compatible plastic pad. Pilot images were obtained in the three orthogonal planes using spin-echo sequences, including coronal proton density with and without fat saturation, sagittal T2-weighted with fat saturation, and axial proton density with fat saturation.

Subsequently, oblique sagittal T2*-weighted images were obtained using a fast, multi-slice, multi-echo, gradient-echo sequence prescribed to cover the medial and lateral menisci. The parameters were TR = 403 ms; TE = 4.38, 11.85, 19.32, 26.79, 33.88, and 40.58 ms; matrix size = 256 × 256 (zero-filled to 512 × 512); in-plane resolution = 0.23 × 0.23 mm; slice thickness = 3 mm; slice gap = 1 mm; NEX = 2; 17 slices; and acquisition time = 15 min 30 s.

Data analysis

Selection of regions of interest

Two raters (one musculoskeletal radiologist and one orthopedic surgeon) interpreted all MR images in selecting OA patients and controls, and a consensus of interpretation was reached. After applying an auto-correlation-based motion correction [30], the femoral and tibial cartilages were manually segmented from the sagittal view of the first-echo image of the knee (Fig. 2). The anterior and posterior margins of the menisci were used as landmarks to draw the ROIs in regions of the femoral and tibial cartilages, respectively. Two experienced operators (PHT and WPC with 8 and 20 years of experience, respectively) selected ROIs for three randomly selected participants separately, discussed their disagreements together, and reached a consensus on the ROI selection procedure stated above to minimize discrepancies. The upper and lower boundaries of the articular cartilage (approximate 2–4 pixels wide) were excluded from ROIs to avoid partial volume effects. The inter-operator disagreement in ROI selection was less than 8%, which limited discrepancies in the T2* estimates to less than 6%.

Fig. 2
figure 2

Regions of interest (ROIs) selection. Manual selection of the ROIs in the femoral and tibial cartilages was performed on the first-echo image of the knee

Full size image

T2* calculation

Cartilage T2* analysis was conducted on a zone-by-zone basis improving data fitting precision in the presence of noise and minimizing the partial volume effect [31]. The mean signal intensity was derived in the femoral and tibial cartilages on each motion-corrected image. The T2* values were subsequently analyzed in MATLAB 2010b (version 7.11; MathWorks) based on the least squares single-exponential curve-fitting method. Values for R2 were assessed to confirm goodness of fit to the curve. R2 > 0.95 was used to define an acceptable fit. Averaged T2* values derived over all slices were used for further comparisons.

Statistical analysis

All data were analyzed using Statistical Package for the Social Science software, Version 20.0 (SPSS). Means and SDs for the femoral and tibial cartilage T2* values in each study group (OA patients and asymptomatic participants) were calculated first. A paired t test was used to compare differences between two compartments within a given group. The root-mean-square average coefficient of variation (CVRMS) and intra-class correlation coefficient (ICC, two-way random effects model, multiple raters/measurements) were calculated to assess method reproducibility. Reproducibility was considered good when the CVRMS is less than 10% and the ICC exceeds 0.75.

Two-way repeated measures analysis of variance (ANOVA) was used to examine differences in T2* values between age groups and to compare the cartilage T2* values between men and women in the asymptomatic participants and OA patients, respectively. Pearson’s correlation test was used to analyze the association between T2* values and Lequesne indices in the OA group. Multiple testing was corrected using the Benjamini-Hochberg method for false discovery rate [32]. The findings were considered statistically significant p < 0.05.

Results

Age-dependent differences in T2* values

The mean (± SD) of the femoral and tibial cartilage T2* values for each study group is demonstrated as follows. T2* imaging of the medial and lateral TF cartilages in the groups of 21–40, 41–60, and > 60-year-olds are compared in Fig. 3. The T2* values of the femoral cartilage in 21–40-year-olds were 27.9 ± 1.9 and 27.1 ± 1.9 ms in the medial and lateral compartments, respectively. These values were significantly greater in 41–60-year-olds (30.4 ± 1.4 and 30.3 ± 2.1 ms, respectively) and > 60-year-olds (33.3 ± 2.5 and 32.7 ± 1.9 ms, respectively; p < 0.001). Additionally, the T2* values of the tibial cartilage in 21–40-year-olds were 24.9 ± 1.3 and 24.8 ± 1.2 ms in the medial and lateral compartments, respectively. These values were significantly greater in 41–60-year-olds (27.6 ± 1.6 and 26.8 ± 1.6 ms, respectively) and > 60-year-olds (28.9 ± 1.7 and 28.2 ± 1.3 ms, respectively; p < 0.001). The T2* values did not differ significantly between the two compartments within each group. The CVRMS and ICC for the selected ROIs in the controls were less than 9% and greater than 0.8, respectively, indicating good reproducibility.

Fig. 3
figure 3

Plots of cartilage T2* values with 95% confidence interval measured at femoral (a) and tibial (b) compartments of asymptomatic participants in three different age groups. Asterisks indicate significant increases of T2* values from group 1 to group 3 (p < 0.001) via an ANOVA. Group 1, 21–40 years old; group 2, 41–60 years old; group 3, > 60 years old; MFC, medial femoral cartilage; LFC, lateral femoral cartilage

Full size image

Comparisons of T2* values between OA patients and unaffected controls

Table 2 shows T2* values of the cartilage in patients with OA (age, 59.1 ± 7.8 years) and in age- and sex-matched controls (age, 56.9 ± 9.3 years). Although a significant difference was not found between T2* values for the lateral tibial cartilage (28.2 ± 2.5 and 27.5 ± 1.6 ms, respectively), this was not the case for any other compartment. Values were significantly greater in the OA group for the medial femoral cartilage (36.0 ± 4.1 and 31.9 ± 2.4 ms, respectively), medial tibial cartilage (31.0 ± 2.9 and 28.2 ± 1.7 ms, respectively), and lateral femoral cartilage (33.9 ± 3.2 and 31.5 ± 2.3 ms, respectively) relative to those of controls (p < 0.01).

Table 2 Comparisons of the cartilage T2* values between OA patients and the age- and sex-matched asymptomatic participants
Full size table

Sex differences in T2* values

Table 3 shows comparisons of the cartilage T2* values between men and women in the asymptomatic participants and OA patients, respectively. The T2* values were significantly greater in the medial femoral cartilage (37.6 ± 4.4 and 34.2 ± 2.6 ms, respectively) and medial tibial cartilage (32.2 ± 2.6 and 29.6 ± 2.5 ms, respectively; p < 0.01), but not in the lateral femoral cartilage (33.7 ± 3.3 and 34.1 ± 3.0 ms, respectively) or lateral tibial cartilage (29.1 ± 2.6 and 27.6 ± 2.2 ms, respectively). No significant differences were seen in the T2* values between the sexes among the asymptomatic controls.

Table 3 Comparisons of the cartilage T2* values between men and women in the asymptomatic participants and OA patients
Full size table

Correlation between T2* values and Lequesne index in OA patients

Figure 4 shows the relationship between the T2* values and the Lequesne index in patients with OA. Although the T2* values for the lateral femoral cartilage and the lateral tibial cartilage did not significantly correlate with the Lequesne index (r = 0.088, p = 0.589 and r = 0.089, p = 0.586, respectively), strong and moderate correlations were found between the Lequesne index and the T2* values for the medial femoral cartilage (r = 0.636; p < 0.001) and the medial tibial cartilage (r = 0.433; p = 0.005).

Fig. 4
figure 4

The distribution and correlation of the Lequesne index and T2* values measured at the medial femoral, medial tibial, lateral femoral, and lateral tibial compartments in OA patients. Significant positive correlations between the T2* values and the Lequesne index were found in the medial femoral and medial tibial compartments

Full size image

Discussion

We investigated the feasibility of early detection of cartilage degeneration and differentiation of OA from normal cartilage using quantitative MR T2* measures in vivo. The findings show good reliability of cartilage T2* values as shown in the previous report [7]. Age-related increases in T2* values were found in both the medial and lateral compartments of the femoral and tibial cartilages suggesting possible associations between MR T2* values and age. A similar tendency has been reported elsewhere using MR cartilage T2 measurements suggesting an altered collagen network and an elevated water content in articular cartilage during advanced age or OA progression [33,34,35]. The cartilage T2* measurements offer further insight into altered cartilage geometry and its composition during OA progression while offering the benefit of a shorter acquisition time.

Two prior in vitro studies using three-dimensional GRE-based sequences with highly reduced TR values (less than 100 ms) showed significant decreases in T2* indicating increasing grades of cartilage degeneration [6, 36]. However, other studies showed that mean T2* values were significantly greater in injured cartilage compared to those of healthy human and animal tissue controls using a multi-echo GRE sequence where TR exceeded 400 ms [10, 37]. This controversy highlights the potential influence of imaging sequences and parameters.

In this study, multi-echo GRE images were acquired at TR = 403 ms. A higher water signal was preserved when TR is greater than 100 ms. This occurs because of the long T1 relaxation time of free water protons. Our findings reveal that the group with early OA has greater T2* values compared to controls. This finding is consistent with previous reports [7] and could have resulted from disrupted cartilage organization and an elevated ratio between free water and bound water during OA progression.

Osteoarthritis affects mostly women [37]—a meta-analysis demonstrated that knee OA is more severe in women than in men [38], and that study emphasized the need to understand sex-based effects on cartilage degradation. Prior work showed significant sex differences in asymptomatic controls consistent with a previous report [39]. In contrast, our results demonstrate that the T2* values in the medial femoral and medial tibial compartments are greater in women than those in men among those with early OA.

Both mechanical and biologic dysfunction in knee joints can trigger OA onset. Although the exact pathological mechanism of OA has not yet been elucidated, several previous studies have demonstrated either greater prevalence or greater severity of OA in the medial compartment of the knee [15, 40]. This might be the result of discrepant mechanical stresses between the medial and lateral cartilages originating from joint instability and/or mobility impairment.

While the T2* values in the medial compartment did not differ significantly from those in the tibial compartment in asymptomatic controls, significant differences were found between those with early OA and controls as well as between sexes in the OA group. Moreover, stronger correlations were found between T2* values and the Lequesne index in the medial compartment compared to the lateral compartment among early OA patients. This hints the potential for monitoring the severity of knee OA using quantitative MR T2* measurements.

This study does have some limitations. For one, a single-component exponential T2* fitting was performed on the cartilage. Although this method has been frequently used to detect early degeneration of knee cartilage, multi-component T2* mapping can provide more extensive information such as a signal from short T2 components of cartilage. This could be an important and novel OA biomarker [41]. Thus, ultra-short echo time imaging could be an alternative for assessing cartilage changes in future studies. Second, contamination by the partial volume effect cannot be completely ignored. Higher-resolution three-dimensional MR cartilage imaging might provide a solution. Finally, assessing the association between histological examinations and cartilage T2* values could facilitate further clinical interpretations.

Conclusion

This study demonstrates the feasibility of using MR T2* measurements at 1.5 T to detect the early changes of cartilage degeneration. T2* values of knee cartilage were correlated with age, sex, and location. Cartilage T2* values tend to increase with age and are useful in assessing cartilage degeneration in early OA.