CME JOURNAL OF MAGNETIC RESONANCE IMAGING 33:180–188 (2011)
Original Research
Advanced Morphological 3D Magnetic Resonance Observation of Cartilage Repair Tissue (MOCART) Scoring Using a New Isotropic 3D Proton-Density, Turbo Spin Echo Sequence With Variable Flip Angle Distribution (PD-SPACE) Compared to an Isotropic 3D Steady-State Free Precession Sequence (True-FISP) and Standard 2D Sequences Goetz H. Welsch, MD,1,2* Lukas Zak, MD,3 Tallal C. Mamisch, MD,4 Dominik Paul, PhD,5 Lars Lauer, PhD,5 Andreas Mauerer, MD,2 Stefan Marlovits, MD,3 and Siegfried Trattnig, MD1 Purpose: To evaluate a new isotropic 3D proton-density, turbo-spin-echo sequence with variable flip-angle distribution (PD-SPACE) sequence compared to an isotropic 3D true-fast-imaging with steady-state-precession (TrueFISP) sequence and 2D standard MR sequences with regard to the new 3D magnetic resonance observation of cartilage repair tissue (MOCART) score.
bone interface, surface, subchondral lamina, chondral osteophytes, and effusion (Pearson coefficients 0.514– 0.865). Especially between the standard sequences and the 3D True-FISP sequence, the variables structure, signal intensity, subchondral bone, and bone marrow edema revealed lower, not significant, correlation values (0.242– 0.383). Subjective quality was good for all sequences (P 0.05). Artifacts were most often visible on the 3D TrueFISP sequence (P < 0.05).
Materials and Methods: Sixty consecutive MR scans on 37 patients (age: 32.8 6 7.9 years) after matrix-associated autologous chondrocyte transplantation (MACT) of the knee were prospectively included. The 3D MOCART score was assessed using the standard 2D sequences and the multiplanar-reconstruction (MPR) of both isotropic sequences. Statistical, Bonferroni-corrected correlation as well as subjective quality analysis were performed.
Conclusion: Different isotropic sequences can be used for the 3D evaluation of cartilage repair with the benefits of isotropic 3D MRI, MPR, and a significantly reduced scan time, where the 3D PD-SPACE sequence reveals the best results. Key Words: PD-SPACE; True-FISP; MOCART; cartilage repair J. Magn. Reson. Imaging 2011;33:180–188. C 2010 Wiley-Liss, Inc. V
Results: The correlation of the different sequences was significant for the variables defect fill, cartilage interface,
1 MR Center, Department of Radiology, Medical University of Vienna, Vienna, Austria. 2 Department of Trauma Surgery, University Hospital Erlangen, Erlangen, Germany. 3 Center for Joint and Cartilage, Department of Trauma Surgery, Medical University of Vienna, Vienna, Austria. 4 Department of Orthopedic Surgery, University of Berne, Berne, Switzerland. 5 Siemens Healthcare, Erlangen, Germany. Contract grant sponsor: project ‘‘Vienna Advanced Clinical Imaging Center’’ (VIACLIC), within the ‘‘Vienna Spots of Excellence’’ program; a collaboration of the Medical University of Vienna and Siemens Austria. *Address reprint requests to: G.H.W., MR Center, High field MRA-1090, Department of Radiology, Medical University of Vienna, Lazarettgasse 14, Vienna, Austria. E-mail:
[email protected] Received May 10, 2010; Accepted September 14, 2010. DOI 10.1002/jmri.22399 View this article online at wileyonlinelibrary.com. C 2010 Wiley-Liss, Inc. V
MAGNETIC RESONANCE IMAGING (MRI) of the knee has been reported as the method of choice to depict cartilage injuries or the postoperative constitution of cartilage repair tissue (1–7). A widespread array of surgical cartilage repair techniques require objective, noninvasive, and reliable follow-up examination, for which the MR observation of cartilage repair tissue (MOCART) score is claimed to be a reliable, reproducible, and accurate tool for assessing cartilage repair tissue (8,9). In several recent studies the MOCART score was used and helped to depict the morphological constitution of the repair tissue and the adjacent structures in the follow-up after different cartilage repair procedures (3,6–8,10–14). This basic MOCART score comprises standard MR sequences and is
180
Isotropic MRI of Cartilage Repair
performed, depending on the location of the cartilage repair site, in the sagittal, axial, or coronal 2D planes using high spatial in-plane resolution together with a slice thickness of 2–4 mm. These sequences offer excellent image quality with high spatial resolution, but there are still some limitations in the exact depiction of the repair tissue, its borders, and the adjacent cartilage due to thick slices, the curvature of cartilage layers, and the limited oblique reconstruction. To use the capabilities of new isovoxel 3D sequences and their 3D multiplanar-reconstruction (MPR) with no loss of spatial resolution, the 3D MOCART scoring system was recently introduced, and was based on an isotropic 3D true fast imaging with steady-state precession (True-FISP) sequence (15). The benefits of this new 3D MOCART score and the 3D evaluation of repair tissue could be clearly demonstrated; however, an obvious limitation of the 3D True-FISP sequence in the postoperative evaluation of cartilage repair tissue was the high number of susceptibility artifacts in the repair tissue and the subchondral bone with this gradient echo-based technique. A recently developed isotropic 3D proton-density, turbo spin-echo sequence, called PD-SPACE (sampling perfection with application-optimized contrasts using different flip angle evolutions), might, because of its spin-echo signal behavior, overcome these limitations with potentially fewer artifacts and better suitability for postoperative imaging of a joint (16). The purpose of this study was to evaluate a new isotropic 3D PD-SPACE sequence compared to an isotropic 3D True-FISP sequence and to a set of 2D standard MR sequences with regard to the new 3D MOCART score. The results of the 3D MOCART scoring using these three sequences were correlated and subjective image quality and sensitivity to artifacts was also graded.
MATERIALS AND METHODS Patient Selection The medical university ethics commission approved the study protocol and written informed consent was obtained from all patients prior to enrolment in the study. Sixty MR scans were prospectively included in this study between October 2008 and October 2009. MRI was performed during the clinical routine at standard follow-up intervals of 1, 3, 6, 12, 24, 48, 60, and 72 months after matrix-associated autologous chondrocyte transplantation (MACT) of the knee joint, with a mean follow-up of 24.3 6 24.1 months. The frequency of the follow-up intervals in this cross-sectional evaluation was: 1 month (n ¼ 8); 3 months (n ¼ 8); 6 months (n ¼ 10); 12 months (n ¼ 7); 24 months (n ¼ 10); 48 months (n ¼ 2); 60 months (n ¼ 13); and 72 months (n ¼ 2). The 60 MRI scans were performed in 37 patients with a mean age of 32.8 6 7.9 years. The 60 MR measurements comprised the right knee joint in 33 and the left knee joint in 27 measurements, with 47 knees of male and 13 knees of female patients. Cartilage repair surgery was performed on
181
the medial femoral condyle (n ¼ 28), the lateral femoral condyle (n ¼ 12), the patella (n ¼ 10), and the trochlea (n ¼ 10). In the 60 consecutive MRIs, considering the standard follow-up examination, 23 patients were included once, nine patients were included twice, three patients were included three times, and two patients were included four times. MACT was performed as a two-step surgical procedure. In a first arthroscopic step a biopsy was obtained from a nonweight-bearing area of the knee. After cell extraction and cultivation, the chondrocytes were transferred onto a biomaterial. In a second step, a mini-arthrotomy was performed to debride the cartilage defect to the subchondral bone. The cell matrix transplants were cut to a suitable size and implanted. The edges were fixed with fibrin glue (17). Image Acquisition The clinical routine MRI was performed on a 3T MR system (Magnetom Tim Trio, Siemens Healthcare, Erlangen, Germany) using a dedicated eight-channel knee coil (InVivo, Gainesville, FL). All patients were positioned consistently with the joint space in the middle of the coil and the knee extended in the coil. The MR protocol was identical for all 60 MRI examinations and consisted of a set of localizers in all three planes, as follows. 1) A standard MR protocol with a sagittal or axial (sagittal for both femoral condyles; axial for the patella and the trochlea) high-resolution, nonfat-saturated proton-density turbo spin-echo (PDTSE) sequence, a sagittal (or axial) PD, respectively, T2-weighted dual fast spin-echo (dual-FSE) sequence, and a coronal fat-saturated proton-density turbo spin-echo (FS-PD-TSE) sequence. Furthermore 2) a coronal isotropic 3D True-FISP sequence, as well as 3) a sagittal isotropic 3D PD-SPACE sequence was acquired. The FS-PD-TSE sequence, as well as both isotropic sequences, was performed in all patients in the same direction. The 3D True-FISP sequence and the 3D PD-SPACE sequence were subsequently reconstructed in every direction using MPR. Sequence parameters are provided in Table 1. Two exemplary patients after MACT of the femoral condyle are provided in Figures 1–4. Whereas Fig. 1 and 3 represent the standard 2D seqeunces, Fig. 2 and 4 represent both 3D isotropic approaches. Data Analysis The new 3D MOCART score, published by Welsch et al (15), with the variables 1) defect fill, 2) cartilage interface, 3) bone interface, 4) surface, 5) structure, 6) signal intensity, 7) subchondral lamina, 8) chondral osteophyte, 9) bone marrow edema, 10) subchondral bone, and 11) effusion, was assessed using i) the standard clinical sequences (with the limitation that the evaluation of the transplant could not be performed in 3D using MPR); ii) the MPR of the isotropic3D True-FISP dataset; and iii) the MPR of the isotropic 3D PD-SPACE dataset. This evaluation is detailed in Table 2. The evaluation was performed on a Leonardo Workstation (Siemens) by a senior
182
Welsch et al.
Table 1 i) Standard MR Seqeunces Sequence protocols
PD-TSE
Dual FSE
Repitition time (TR) Echo time (TE) Flip angle (FA) Field of view (FoV) Pixel matrix Slice thickness Number of slides Voxel size Fat supression PAT Bandwidth Aquisition time
2130ms 36ms 180 120 120mm 448 448 2mm 32 0.27 0.27 2mm None Off 180Hz/Px 5:27min.
5090ms 12ms; 85ms 180 160 160mm 448 448 3mm 30 0.36 0.36 3mm None Off 180Hz/Px 6:03min.
FS-PD-TSE 4250ms 27ms 180 150 150mm 384 384 3mm 36 0.39 0.39 3mm Fat supression 2 150Hz/Px 2:47min.
ii) 3D-True-FISP
iii) 3D-PD-SPACE
7.8ms 3.8ms 30 192 192mm 384 384 0.5mm 240 0.5 0.5 0.5mm Water excitation 2 352Hz/Px 6:21min.
1200ms 30ms 180 160 160mm 324 324 0.6mm 192 0.6 0.6 0.6mm Fat saturation 2 540Hz/Px 7:15min.
Parallel Aquisition technique (PAT) was applied with the given acceleration factor using a generalized auto calibrating partially parallel acquisition (GRAPPA)
musculoskeletal radiologist (25 years of experience) and an orthopedic surgeon with a special interest in MRI (10 years of experience), in consensus. During their analysis both were blinded to patient name and postoperative follow-up interval and were advised only about the localization of the cartilage transplants. The evaluation of the 60 knee MRIs was performed in random order with regard to the respective sequences i), ii), or iii). Further evaluation was performed to classify image quality for the 3D MOCART scoring using i) the standard MR sequences (PD-TSE, dual FSE and FS-PDTSE), ii) the 3D True-FISP sequence, as well as iii) the 3D PD-SPACE sequence to guarantee a sufficient analysis of the cartilage repair tissue and of the adjacent cartilage. Image quality was also graded by the two observers, in consensus, after they performed the MOCART scoring. A four-level scale was used in which a score of 4 indicated excellent image quality; a score of 3, good image quality; a score of 2, acceptable image quality; and a score of 1, poor image quality (18), which indicated that the transplant could not be evaluated sufficiently. In addition, since the diagnostic performance at high-field MRI also suffers from artifacts (19), a scaling of possible artifacts and their impact on the evaluation of the cartilage repair tissue and the sur-
rounding cartilage was performed; again, separately for the three sequences i), ii), and iii). Visible artifacts, including motion, susceptibility, metal/postsurgical, banding, etc., were subjectively graded as absent (4), mild (3), moderate (2), and severe (1) (20).
Statistical Analysis Statistical analysis was performed to compare the different sequences using SPSS v. 16.0 (Chicago, IL) for Mac (Apple, Cupertino, CA). First, the 3D MOCART scoring was correlated for the 11 single variables. To address the problem of multiple correlations, a Bonferroni correction was additionally performed. Second, the subjective quality and the visible artifacts were compared. For the whole evaluation, the i) standard sequences were compared/correlated to the ii) 3D True-FISP sequence, and to the iii) 3D PD-SPACE sequence. In addition, the ii) 3D True-FISP sequence and the iii) 3D PD-SPACE sequence were compared (respectively correlated). Correlation was achieved using a bivariate correlation with the Pearson correlation coefficient. Comparison for quality and artifacts was prepared using Student’s t-test. For all evaluations a P-value < 0.05 was considered statistically significant.
Figure 1. A 35-year-old male patient 24 months after MACT of the medial femoral condyle. Standard 2D MR sequences with a sagittal PD-TSE sequence (TR ¼ 2130 msec, TE ¼ 36 msec, flip angle ¼ 180 ) (a), a sagittal T2-weighted Dual-FSE sequence (5090/12;85/flip angle 180 ) (first echo (b); second echo (c)) and a coronal FS-PD-TSE sequence (4250/27/flip angle 180 ) (d). The cartilage repair tissue is nicely integrated to its border zones, the defect fill is 100%, there are only slight signal alterations visible, and the subchondral bone seems largely intact.
Isotropic MRI of Cartilage Repair
183
Figure 3. A 22-year-old female patient 48 months after MACT of the medial femoral condyle. Standard 2D MR sequences with a sagittal PD-TSE sequence (TR ¼ 2130 msec / TE ¼ 36 msec / flip angle ¼ 180 ) (a) and a coronal FS-PDTSE sequence (4250/27/flip angle 180 ) (b). The cartilage repair tissue seems to be integrated to its border zones; however, a clear split-like lesion (>50% of the thickness of the cartilage) becomes obvious in the coronal plane. The subchondral bone reveals alterations; a still slightly visible bone plug was used during surgery for a bony defect caused by osteochondritis dissecans.
Figure 2. A 35-year-old male patient 24 months after MACT of the medial femoral condyle (same as in Fig. 1) assessed with the sagittal and coronal reconstruction based on the MPR of the isotropic 3D True-FISP (7.8/3.8/flip angle ¼ 30 ) dataset (a,b) and the isotropic 3D PD-SPACE (1200/30/flip angle 180 ) dataset (c,d).
11) effusion, with and without Bonferroni correction, with Pearson coefficients ranging from 0.514–0.865. The variables 5) structure, 6) signal intensity, 9) bone marrow edema, and 10) subchondral bone demonstrated lower correlations, especially between the i) standard MR sequences and ii) the 3D True-FISP sequence, with Pearson coefficients ranging from
RESULTS 3D MOCART Score The results for the 3D MOCART evaluation, as performed by i) the standard MR sequences, ii) the 3D True-FISP sequence, and iii) the 3D PD-SPACE sequence, are listed in Table 2. With respect to the single variables, comparable results for all three sequences could be achieved for 1) defect fill, 2) cartilage interface, 3) bone interface, 4) surface, 7) subchondral lamina, 8) chondral osteophytes, and 11) effusion. However, when comparing the different sequences i) to iii), the cartilage repair tissue and its surrounding structures were assessed with slightly inferior results by the 3D True-FISP sequence, compared to the 3D PD-SPACE sequence, with respect to the standard MR sequences. Compared to the abovelisted variables, the depiction of the variables 5) structure, 6) signal intensity, 9) bone marrow edema, and 10) subchondral bone showed larger differences between the three sequences i) to iii) (Table 2). Correlations The correlation between the 3D MOCART scoring as performed by i) the standard MR sequences, ii) the 3D True-FISP sequence, and iii) the 3D PD-SPACE sequence was significant for the variables 1) defect fill, 2) cartilage interface, 3) bone interface, 4) surface, 7) subchondral lamina, 8) chondral osteophytes, and
Figure 4. The same 22-year-old female patient 48 months after MACT as in Fig. 3, evaluated with the sagittal and coronal reconstruction based on the MPR of the isotropic 3D True-FISP sequence (7.8/3.8/flip angle ¼ 30 ) (a,b) and the isotropic 3D PD-SPACE sequence (1200/30/flip angle 180 ) (c,d). The changes within the subchondral bone are clearer visualized with the 3D PD-SPACE sequence compared to the 3D True-FISP sequence.
184
Welsch et al.
Table 2 Three-dimensional (3D) magnetic resonance observation of cartilage repair tissue (MOCART) score assessed using a i) set of standard MR sequences, the ii) MPR of the 3D-True-FISP sequence and iii) the MPR of the 3D-PD-SPACE sequence Variables
i) Standard 2D
ii) 3D-True-FISP
iii) 3D-PD-SPACE
1. Defect fill (degree of defect repair and filling of the defect in relation to the adjacent cartilage) * 0% 0 (0) 0 (0) 0 (0) * 0–25% 0 (0) 0 (0) 0 (0) * 25–50% 0 (0) 0 (0) 0 (0) * 50–75% 9 (15) 4 (6.7) 5 (8.3) * 75–100% 11 (18.3) 13 (21,7) 12 (20) * 100% 28 (46.7) 29 (48.3) 28 (46.7) * 100–125% 7 (11.7) 11 (18.3) 11 (18.3) * 125–150% 5 (8.3) 3 (5) 3 (5) * 150–200% 0 (0) 0 (0) 0 (0) * >200% 0 (0) 0 (0) 0 (0) Localization * Whole area of cartilage repair * > 50% * < 50% * Central * Peripheral * Weight-bearing * Non weight-bearing 2. Cartilage Interface (integration with adjacent cartilage to border zone in two planes) Sagittal (Femur, Patella, Trochlea, Tibia) * Complete 41 (68.3) 40 (66.7) 30 (50) * Demarcating border visible (split-like) 12 (20) 13 (21.7) 23 (38.3) * Defect visible 50% 3 (5) 2 (3.3) 1 (1.7) Coronal (Femur, Tibia); Axial (Patella, Trochlea) * Complete 38 (63.3) 34 (56.7) * Demarcating border visible (split-like) 18 (30) 21 (35.0) * Defect visible 50% 0 (0) 0 (0) Localization * Whole area of cartilage repair * > 50% * 50% * < 50% * Central * Peripheral * Weight-bearing * Non weight-bearing 5. Structure (constitution of the repair tissue) * Homogeneous 22 (45) 15 (25) 12 (16.7) * Inhomogeneous or cleft formation 38 (55) 45 (75) 48 (83.3) Localization * Whole area of cartilage repair * > 50% * < 50% * Central * Peripheral * Weight-bearing * Non weight-bearing 6. Signal intensity (Intensity of MR signal in of the repair tissue in comparison to the adjacent cartilage: normal ¼ identical to adjacent cartilage; nearly normal ¼ slight areas of signal alterations; abnormal ¼ large areas of signal alteration) * Normal 29 (48.3) 24 (40) 27 (45) * Nearly normal 25 (41.7) 32 (53.3) 32 (53.3) * Abnormal 6 (10) 3 (6.7) 1 (1.7) Localization * Central * Peripheral * Weight-bearing * Non weight-bearing 7. Subchondral lamina (Constitution of the subchondral lamina) * Intact 33 (55) 36 (60) 29 (48.3) * Not intact 27 (45) 24 (40) 31 (51.7) Localization * Whole area of cartilage repair * > 50% * < 50% * Central * Peripheral * Weight-bearing * Non weight-bearing
Isotropic MRI of Cartilage Repair
185
Table 2 (Continued) Variables
i) Standard 2D
ii) 3D-True-FISP
iii) 3D-PD-SPACE
8. Chondral Osteophytes (Osteophytes within the cartilage repair area) * Absent 37 (61.7) 31 (51.7) 31 (51.7) * Osteophytes < 50% of repair tissue 16 (26.7) 14 (23.3) 17 (28.3) * Osteophytes > 50% of repair tissue 7 (11.7) 15 (25.0) 12 (20) Localization Size: ——mm (plane: ——) —— mm (plane: ——) * Central * Peripheral * Weight-bearing * Non weight-bearing 9. Bone marrow edema (Maximum size and localization in relation to the cartilage repair tissue and other alterations assessed in the 3D MOCART score). * Absent 14 (23.3) 33 (55) 15 (25) * Small (< 1cm) 14 (23.3) 15 (25) 13 (21.7) * Medium (< 2cm) 20 (33.3) 8 (13.3) 20 (32.3) * Large (< 4cm) 9 (15) 4 (6.7) 10 (16.7) * Diffuse 3 (5) 0 (0) 2 (3.3) Localization Size: ——mm (plane: ——) —— mm (plane: ——) * Central * Peripheral * Weight-bearing * Non weight-bearing * Relation to other alterations within this score of variable No. —— 10. Subchondral bone (Constitution of the subchondral bone) * Intact 31 (33.3) 34 (56.7) 39 (65) * Granulation tissue 18 (30) 25 (33.3) 18 (30) * Cyst 4 (6.7) 4 (6.7) 3 (5) Localization * Whole area of cartilage repair * > 50% * < 50% * Central * Peripheral * Weight-bearing * Non weight-bearing 11. Effusion (Approx. size of joint effusion visualized in all planes) * Absent 17 (28.3) 10 (16.7) 11 (18.3) * Small 26 (43.3) 31 (51.7) 29 (48.3) * Medium 14 (23.3) 16 (26.7) 17 (28.3) * Large 3 (5) 3 (5) 3 (5) Variables 1–11 for 3D MOCART score; subcategories ‘ localization’’ optional.
0.242–0.383. The correlations between the i) standard sequences and iii) the 3D PD-SPACE sequence and between ii) the 3D True-FISP sequence and the iii) 3D PD-SPACE sequence revealed higher (Pearson: 0.307– 0.633) correlations. Whereas the correlation between the ii) the 3D True-FISP sequence and the iii) 3D PDSPACE sequence was again significant with and without Bonferroni correction, the correlations between the i) standard MR sequences and ii) the 3D TrueFISP sequence, for these four variables, were not significant based on the Bonferroni correction. Results are provided in detail in Table 3. Image Quality and Artifacts In the assessment of the subjective quality and possible artifacts, i) the standard MR sequences revealed the highest scoring (quality: 4.0 6 0.0; artifacts: 3.93 6 0.3), with no significant difference compared to iii) the 3D PD-SPACE sequence (quality: 3.87 6 0.3; artifacts: 3.77 6 0.5) (P 0.05). For ii) the 3D True-FISP sequence, the image quality showed comparable values (quality: 3.78 6 0.5; P 0.05), whereas artifacts were significantly more often visible (artifacts: 3.30 6 0.7; P < 0.05) when compared to i) standard sequences and the iii) 3D PD-SPACE sequence. DISCUSSION The present study indicates that with the use of i) standard MR sequences, ii) a 3D True-FISP
sequence, or iii) a 3D PD-SPACE sequence the recently introduced 3D MOCART score (15) can be obtained with comparable results. Thus, the new 3D MOCART score can be calculated with the MPR of only one isotropic 3D sequence, indicating advantages in the depiction of knee cartilage and especially the cartilage repair tissue (6,21–23). The evaluation can be performed in more detail with 3D isotropic MRI, possible changes can be detected, and additional pathologies in other structures in the joint can be diagnosed (7). The isotropic data set is used to move/scroll through the knee joint in three planes, localize the repair tissue, and use the right angulations to depict the structure of the repair tissue and the adjacent cartilage. The variables of the 3D MOCART score, where the more detailed information gained by the isotropic sequences and the associated MPR might be beneficial, are the cartilage interface, the surface of the repair tissue, the subchondral osteophytes, and the effusion. These parameters revealed, in the present study, more ‘‘normal’’ cases for the i) standard 2D evaluation compared to the ii) / iii) isotropic 3D evaluation. This higher number of irregularities in the repair tissue on isotropic sequences might be due to the easier assessment of the repair tissue site by the MPR of the isotropic dataset, with visualization in every plane. Another reason might be the much lower slice thickness of the isotropic sequences of 0.5 mm to 2–3 mm of the standard 2D sequences. Although it has to be
186
Welsch et al.
Table 3 Pearson correlation coefficients with given p-value (p) and Bonferroni corrected p-value (pB) for the 3D - MOCART score as performed by a i) set of standard MR sequences, the ii) MPR of the 3D-True-FISP sequence and iii) the MPR of the 3D-PD-SPACE sequence i) Standard 2D vs. ii) 3D-True-FISP
Correlations Variables: 1. Defect fill 2. Cartilage interface 3. Bone interface 4. Surface 5. Structure 6. Signal intensity 7. Subchondral lamina 8. Chondral Osteophytes 9. Bone marrow edema 10. Subchondral bone 11. Effusion
0.790 0.793 0.514 0.613 0.280 0.383 0.672 0.807 0.242 0.378 0.519
(p (p (p (p (p (p (p (p (p (p (p
< < < < ¼ < < < ¼ ¼
