A three-dimensional quantitative method to measure meniscus shape, position, and signal intensity using MR images: A pilot study and preliminary results in knee osteoarthritis



This pilot study presents a technique for three-dimensional and quantitative analysis of meniscus shape, position, and signal intensity and compares results in knees with (n = 20) and without (n = 11) radiographic osteoarthritis. 3-T MR images with 2mm section thickness were acquired using a proton density–weighted, fat-suppressed, coronal, fast spin-echo sequence. Segmentation of the tibial, femoral, and external surface of the medial meniscus and the tibial joint surface was performed. Three-dimensional parameters were computed to describe the shape, position, and signal intensity of the entire meniscus and three subregions (body, anterior, and posterior horn). Key results included a greater size (i.e., volume, surface areas, and thickness), increased medial extrusion (i.e., greater extrusion distance, greater meniscal area uncovered by tibial surface), and elevated signal intensity of the medial meniscus in osteoarthritis than in nonosteoarthritis knees, particularly in the meniscus body. These results need to be confirmed in larger cohorts, preferably under weight-bearing conditions. Magn Reson Med 63:1162–1171, 2010. © 2010 Wiley-Liss, Inc.

The meniscus plays an important role in normal knee function by providing a higher degree of joint conformity (congruity) and by distributing loads over a wider area (1). Meniscal pathology (i.e., extrusion and tears) is frequent in the general population, even among asymptomatic individuals, and becomes more common with increasing age (2). Meniscal pathology has also been associated with knee pain (3–6) and osteoarthritis (OA) progression, specifically with increased rates of cartilage loss (7–12). These findings have generally relied on semiquantitative scoring of the meniscus (4, 13), and only few studies have utilized quantitative measures of meniscus position (subluxation) or shape (14–16).

Gale et al. (14) used coronal fat-saturated proton density and T2-weighted images to investigate the relationship between meniscal extrusion (subluxation) and joint space narrowing in 233 participants with symptomatic knee OA and in 58 asymptomatic controls. They determined the degree of extrusion to the nearest millimeter in the image where the greatest distance between the most peripheral aspect of the meniscus and the border of the tibia (excluding osteophytes) was observed. The OA participants displayed more extrusion of the medial meniscus (MM) than controls (5.1 versus 2.8mm; P = 0.001). Modest degrees of meniscal extrusion were common in controls, but severe degrees (>7mm) were unique to OA cases. Hunter et al. (15) explored the role of meniscal tears (semiquantitative scoring (13)), extrusion, and height as risk factors for cartilage loss in 257 subjects. Extrusion and height were measured quantitatively in the coronal MR image showing the maximal medial tibial spine volume and in two sagittal images, one through the medial and one through the lateral tibia. Meniscal coverage and height were smaller in knees with meniscal extrusion, and coverage, height, and extrusion were all associated with a higher risk of cartilage loss (13).

However, the above two-dimensional studies (14, 15) did not perform a full three-dimensional (3D) analysis of the meniscus. Two-dimensional measurements rely on the particular slice selected and its orientation, which may be difficult to reproduce in longitudinal trials. Also, two-dimensional measurements do not permit full analysis of defined meniscal subregions, such as the anterior and posterior horn and the body. With regard to articular cartilage, recent research reported 3D, quantitative outcomes of cartilage morphology to be more effective in describing longitudinal progression and the relationship with other OA risk factors than semiquantitative cartilage readings (12). Given the important role of the meniscus for symptoms and structural progression in OA, the objective was to develop a technique for fully quantitative 3D analysis of meniscus shape, position, and signal intensity (SI), including the body, the anterior horn, and the posterior horn. In this pilot analysis, we also explored whether such quantitative measures differ significantly between knees with radiographic OA and healthy knees.


Study Participants

This study was approved by the local committee on human research. Participants were recruited by committee on human research–approved flyers and newspaper advertisements. Inclusion criteria were female gender, age >40 years, and a body mass index of 25 to 35 kg/m2. Anterior-posterior extended weight-bearing radiographs were taken. Inclusion criteria for OA participants were self-reported pain, aching, or stiffness most days of a month during the past year; a Kellgren Lawrence grade (KLG) 2 or 3 of the study knee (readings performed by T.M.L.); the same degree of, less severe, or no OA in the contralateral knee; a medial joint space width >2mm; and a medial joint space width smaller than the lateral joint space width. Inclusion criteria for normal controls were self-reported infrequent (or no) knee pain, aching, or stiffness during the past year and no radiographic evidence of OA on either knee (KLG0). General exclusion criteria were MRI contraindications, a history of knee disease other than OA or knee surgery (including meniscus surgery), and intra-articular steroid injection. Finally, 31 women (age 55.3 ± 6.0 years [mean ± standard deviation], height 1.62 ± 0.07 m, body mass index 28.0 ± 2.4 kg/m2) were included, 11 with healthy knees (KLG0) and 20 with radiographic evidence of knee OA (10 KLG2, 10 KLG3). The OA participants were somewhat older than the controls (57.0 ± 4.7 versus 52.0 ± 6.9 years; P < 0.05), but height and body mass index were similar in both groups (P = 0.65).

Imaging Procedures

MR images were acquired using a proton density–weighted fat-suppressed coronal fast spin echo sequence with 2mm slice thickness (no interslice gap) and an in-plane resolution of 0.42mm × 0.83mm (interpolated to 0.31mm × 0.31mm, repetition time 3000 ms, echo time 10.2 ms, matrix 384 × 192 interpolated to 512 × 512, field of view = 16 cm, number of excitations = 2, acquisition time = 4:44 min), using a 3-T scanner (Signa HDx; General Electric, Waukesha, WI) and an eight-channel phased-array transmit-receive knee coil. The patient was positioned supine in the scanner; consistent alignment was ensured by positioning the knee in the center of the coil and the foot on a footplate with 10° external rotation.

Image Analysis

Manual segmentation was performed by a single reader (R.B.F.) who was blinded to KLGs. Segmentation software previously applied to the MR-based analysis of cartilage morphology (Chondrometrics GmbH, Ainring, Germany) (17–19) was extended for the segmentation and the computation of quantitative, 3D measures of meniscus shape and position. Although the current study focuses on the MM, the software allows the analysis of both menisci. Segmentation of MM was subdivided into four contours (Fig. 1):

Figure 1.

Coronal MR images without (left) and with segmentation (right) of the inferior surface of the MM (TA, green) directed toward the MT, the superior surface (FA, magenta) directed toward the medial femoral condyle (cMF), the external surface (EA, turquoise) of the MM the joint surface of the MT (MT.ACdAB, blue), excluding the part of the surface made up by peripheral osteophytes. The image shows a healthy knee (top row) and a knee with OA (bottom row).

  • 1the inferior side of MM, directed toward the tibial cartilage = tibial area (TA),
  • 2the superior side of MM, directed toward the femoral cartilage = femoral area (FA),
  • 3the external side of MM, directed toward medial = external area (EA),
  • 4the joint surface of the medial tibia (MT), excluding peripheral osteophytes (Fig. 1).

Since the total area of subchondral bone may not always be fully covered with cartilage (AC) (20), the entire contact surface with the meniscus was segmented as one contour (ACdAB) (20) (Fig. 1). Anteriorly and posteriorly, segmentation of the MM was only performed in slices where the MT cartilage surface was visible because partial-volume effects were too high to support segmentation beyond these slices. All segmentations were quality controlled (F.E.); adjustments were based on consensus of both readers.

Six knees (2 KLG0, 2 KLG2, and 2 KLG3) were reanalyzed 6 months later by the same reader (R.B.F.), with blinding to KLGs and the first segmentation, to assess the resegmentation precision. The system for naming quantitative measures of the meniscus, using an anatomic label first (MT or MM), structural feature label second (e.g., Th = thickness), and a statistical label last (e.g., Me = mean) was adopted from a previously published nomenclature for quantitative measurements of articular cartilage (20).

Shape Analysis

The size of the surface areas of MM (TA, FA, EA) and the joint surface of MT (ACdAB) was computed after 3D reconstruction (21) (Table 1; Fig. 2a). The anterior and posterior ends of EA were approximated by filling the gap between the TA and FA in the most anterior and posterior slice, in which MM was segmented. The volume of MM (MM.V) was calculated by numerical integration of all voxels encompassed by TA, FA, and EA. The mean (MM.Th.Me) and maximal thickness of the meniscus (MM.Th.Max, also termed “meniscal height” (15)) were determined by averaging two bidirectional Euclidean distance transforms (22) between TA and FA, respectively (Fig. 2b). The mean and maximal width of the meniscus (MM.Wid.Me and MM.Wid.Max) were determined in 3D, as the average or maximal distance between (a) the intersection of TA and FA, and (b) the EA (Fig. 2c). The mean bulging (MM.Bul.Me) of the EA (MM.EA) was determined slice by slice as the distance between EA and a straight line connecting (a) the intersection of TA and EA with (b) the intersection of FA and EA (Fig. 2d). These measurements were averaged across all segmented slices.

Figure 2.

3D reconstruction of the MM. a: MM shown on top of the joint surface of the MT (MT.ACdAB, blue) from anterior, with the MM being cut open in the middle for visualization purposes. The superior surface of MM (MM.FA) is shown in magenta, the inferior surface of MM (MM.TA) green, and the external surface of MM (MM.EA) in turquoise. The gray frame indicates the detailed view used in subsequent parts of this figure. b: Detailed view showing the bidirectional measurement of the thickness of the MM (MM.Th, yellow arrows). The meniscal surfaces are displayed in the same colors as in (a). c: Detailed view showing the measurement of the meniscal width (MM.Wid, yellow arrows). The overlap between MM.TA and MT.ACdAB was used to determine the coverage of the surfaces (MM.TA. Cov and MT.ACdAB.Cov in cm2), the overlap distance (MM.OvD inmm) and the extrusion (MM.Ex inmm). d: Detailed view showing the measurement of meniscus bulging (MM.Bul, yellow arrows).

Table 1. Labels (and Measurement Units) Used to Describe 3D Parameters of Meniscus Shape, Position (Relative to the Tibia), and SI*
  • *

    The system for naming quantitative measures of meniscus shape was adopted from a previously published nomenclature for quantitative measures of articular cartilage (20).

ACdABJoint surface of the tibia, consisting of the area of the cartilage surface (AC), and denuded areas of subchondral bone (dAB) if applicablecm2
ACdAB.CovACdAB covered by the meniscuscm2
ACdAB.Cov%Percentage of ACdAB covered by MM.TA%
ACdAB.UncovACdAB not covered by the meniscuscm2
ACdAB.Uncov%Percentage of ACdAB not covered by MM.TA%
TAInferiorly located surface of the meniscus (oriented toward the tibia)cm2
FASuperiorly located surface of the meniscus (oriented toward the femur)cm2
EAEA of the meniscus (connection to the medial collateral ligament)cm2
Th.MeMean thickness of the meniscusmm
Th.MaxMaximum thickness of the meniscusmm
VVolume of the meniscusmm3
Bul.MeMean bulging of the meniscus over all slices (two-dimensional measurement)mm
Wid.MeMean width of the meniscusmm
Wid.MaxMaximum width of the meniscusmm
TA.CovMM.TA covered by the tibiacm2
TA.Cov%Percentage of MM.TA covered by the tibia%
TA.UncovMM.TA not covered by the tibiacm2
TA.Uncov%Percentage of MM.TA not covered by the tibia%
mEx.MeMean medial extrusion of the meniscusmm
mEx.MinMinimum medial extrusion of the meniscusmm
mEx.MaxMaximum medial extrusion of the meniscusmm
OvD.MeMean overlap distance = distance between internal border of MM and external border of MT.ACdABmm
OvD.MinMinimal overlap distancemm
OvD.MaxMaximum overlap distancemm
SI.MeMean SI of the meniscus voxels 
SI.MinMinimum of the SI 
SI.MaxMaximum of the SI 
SI.SDStandard deviation of the SI 

Positional Analysis

The overlap between MT.ACdAB and MM.TA was used to indirectly describe the relative position of the meniscus on the tibial plateau. To determine the directional vector running orthogonal to the tibial surface, a plane minimizing the least-squares distance was fitted to MT.ACdAB. By intersecting the set of lines through each of the vertices of MT.ACdAB and MM.TA (and parallel to the directional vector) with each of the triangles of the other 3D mesh, the overlap status of the vertices was determined. Surface triangles with three overlapping vertices were labeled as “covered” (MT.ACdAB.Cov and MM.TA.Cov), while the remaining triangles were marked as “uncovered” (MT.ACdAB.Uncov and MM.TA.Uncov; Table 1; Fig. 2c).

For computing meniscal extrusion in medial direction (MM.mEx), the external border of MT.ACdAB (Fig. 2c) was extracted as a list of consecutive vertices. A series of planes was constructed from (a) the vectors connecting two consecutive (external) border points of MT.ACdAB and (b) the vector orthogonal to the MT surface (see above). The shortest distance to this series of planes was evaluated for all vertices of MM.EA (Fig. 2c) to determine the mean (MM.mEx.Me), minimum (MM.mEx.Min), and maximum medial extrusion (MM.mEx.Max). Additionally, we computed the relative distance of the internal margin of MM (intersection of TA and FA) to the external border of MT.ACdAB (Fig. 2c), termed the overlap distance (OvD). The mean (MM.OvD.Me), minimum (MM.OvD.Min), and maximum (MM.OvD.Max) were determined in 3D.

SI Analysis

To measure the SI of MM, the mean (MM.SI.Me), minimum (MM.SI.Min), maximum (MM.SI.Max), and the standard deviation (MM.SI.SD) were derived by extracting the SI across all MM voxels, and after excluding two border voxels.

Regional Analysis

Shape, positional, and SI parameters were determined separately for the body of MM (bMM), as well as in the anterior (aMM) and posterior horns (pMM; Fig. 3). While published approaches for regional cartilage analysis rely on the subchondral bone (18), regional subdivision of the meniscus should not rely on the meniscus or its position relative to other structures, because both depend on the knee flexion angle (23, 24). Because the meniscal roots at the intercondylar area were considered the most stable anatomic landmarks, we determined a point (A, Fig. 3) located halfway between the most internal aspects of the anterior and posterior horns (at the interface of TA and FA). These were considered as an approximation of the meniscal roots (B and C, Fig. 3). We then computed the medial axis of MM.EA (excluding the most anterior and posterior parts that were reconstructed by filling the gap between MM.TA and MM.FA), to obtain an ordered sequence of points describing the anterior-posterior shape of EA. Two points located at 25% (D) and at 75% (E) of the length of the medial EA axis were determined. Two clipping planes were then constructed perpendicular to MM.ACdAB, one through points A and D, and one through points A and E (Fig. 3). The two clipping planes were used to assign all segmented voxels and triangles of the reconstruction of MM to one of the three subregions (aMM, bMM, pMM). Correct labeling of the subregions was ensured by computing the distance between the subregions and the position of the fibula, which was marked manually during the segmentation process. Visual inspection of the 3D reconstruction confirmed that the anatomic labeling was correct in all cases. All quantitative parameters described previously were evaluated within the subregions (Table 1), with the exception of MM.OvD and MM.Wid, as these were based on relatively few measurement points (slice thickness = 2mm).

Figure 3.

Subdivision of the MM into the anterior horn (aMM, yellow), the body segment (bMM, red), and the posterior horn (pMM, magenta). Two clipping planes were defined by the point A (located halfway between the estimated meniscal roots (B, C) of the meniscal horns, two points dividing the medial axis of MM.EA in three parts (D, E), and the vector perpendicular to the tibial joint surface (MT.ACdAB, blue).

Statistical Analysis

Resegmentation precision errors were computed as the root mean square standard deviation (25) for each parameter analyzed. A Mann-Whitney U test was applied to test whether the parameter differed significantly (P < 0.05) between non-OA and OA knees (KLG2 or 3). No adjustment for multiple testing was performed in this exploratory study.


Resegmentation Precision

The test-retest error for resegmentation conditions (6 months later) is reported in Tables 2 to 4. The errors were small for measures of SI and meniscus thickness; larger (but still smaller than the intersubject variability) for surface measures, volume, and bulging; and larger than the intersubject differences for meniscus width and the minimal/maximal (albeit not the mean) medial extrusion.

Table 2. 3D Measures of Meniscal Shape: Mean and Standard Deviation (SD) in Healthy Knees (KLG0) and in Knees With Radiographic OA (KLG2 or 3)*
 KLG0KLG2KLG3P value: OA vs non-OATest-retest (RMS SD)
  • *

    OA = KLG2 or 3, non-OA = KLG0, aMM = anterior horn, bMM = body of the meniscus, pMM = posterior horn; other abbreviations described in Table 1. Differences between OA and non-OA knees were tested for statistical significance using a Mann-Whitney U test; significant differences (P < 0.05) are marked in bold. The resegmentation precision is given as root mean square (RMS) SD of test-retest segmentations.


General Observations in Healthy Knees

The MM.FA was larger than the MM.TA, whereas the MM.EA represented the smallest surface (Table 2). The MM.V was 1770 ± 287mm3. The mean (∼2.5mm) and the maximal thickness of the meniscus (∼6mm) were similar in different parts of MM (Table 2). A 0.2 to 0.3mm bulging of the external surface (MM.Bul.Me) was observed (Table 2). 38.8% ± 5.1% of the tibial joint surface was covered by MM and 71.4% ± 6.0% of the MM.TA was covered by the MT.ACtAB (Table 3). A mean medial extrusion (MM.mEx.Me) of 1.6mm and a maximal medial extrusion (MM.mEx.Max) of 3.2mm were observed for the meniscus (MM.mEx.Me), with values highest in the anterior horn (aMM). The mean distance of the inner meniscal margin to the external tibial border (MM.OvD.Me) was 10.5mm.

Table 3. 3D Measures of Meniscus Position: Mean and Standard Deviation (SD) in Healthy Knees (KLG0) and in Knees With Radiographic OA (KLG2 or 3)*
 KLG0KLG2KLG3P value: OA vs non-OATest-retest (RMS SD)
  • *

    OA = KLG2 or 3, non-OA = KLG0, aMM = anterior horn, bMM = body of the meniscus, pMM = posterior horn; other abbreviations described in Table 1. Differences between OA and non-OA knees were tested for statistical significance using a Mann-Whitney U test; significant differences (P < 0.05) are marked in bold. The resegmentation precision is given as root mean square (RMS) SD of test-retest segmentations.


Differences Between OA and non-OA Knees

No significant differences in the medial tibial joint surface (MT.ACdAB) were observed between OA and non-OA knees, but all meniscal surfaces (TA, FA, and EA) were significantly greater in OA knees (P < 0.05; Table 2). These differences were significant for the entire meniscus and its body, but not for the anterior and posterior horns (Table 2). The meniscus also tended to be thicker in OA knees, but the difference only reached significance (P < 0.05) for Th.Me in the body and the posterior horn (Table 2). MM.V was greater in OA than in non-OA knees in the entire meniscus (P < 0.01), the body (bMM.V; P < 0.01), and the posterior horn (pMM.V; P < 0.05). Bulging of the external meniscal surface was somewhat greater in OA than in non-OA knees in the body (P < 0.05).

No significant differences between OA and non-OA knees were noted in the coverage/uncoverage of the tibial surface areas or in coverage of the tibial meniscal surface (MM.TA.Cov; Table 3). Similarly, the distance between the inner margin of the meniscus and the external border of the tibial surface (MM.OvD) did not differ significantly. However, the size of MM.TA uncovered by the joint surface (MM.TA.Uncov) was significantly larger in OA than in non-OA knees (P < 0.05; Table 3), consistent with a significantly larger MM.TA noted above. The mean (P < 0.01) and maximal (P < 0.05) medial extrusion (MM.mEx.Me, MM.mEx.Max) were significantly greater in OA versus non-OA knees, particularly in the body (Table 3).

The mean SI of the entire meniscus (MM.SI.Me; P < 0.05), the body (bMM.Si.Me; P < 0.01), and the posterior horn (pMM.SI.Me; P < 0.01) was significantly elevated in OA versus non-OA knees (Table 4). The minimal SI (SI.Min) was significantly increased in OA knees in all meniscal subregions, whereas the maximum (SI.Max) only differed significantly in the body and only when excluding the border voxels (Table 4).

Table 4. Measures of Meniscus SI, Including (Top) and Excluding (Bottom) the Border Voxels: Mean and Standard Deviation (SD) in Healthy Knees (KLG0) and in Knees With Radiographic OA (KLG2 or 3)
 KLG0KLG2KLG3P value: OA vs non-OATest-retest (RMS SD)
  1. OA = KLG2 or 3, non-OA = KLG0, aMM = anterior horn, bMM = body of the meniscus, pMM = posterior horn; other abbreviations described in Table 1. Differences between OA and non-OA knees were tested for statistical significance using a Mann-Whitney U test; significant differences (P < 0.05) are marked in bold. The resegmentation precision is given as root mean square (RMS) SD of test-retest segmentations.

Across all voxels
Without border voxels


Here we propose a technique for fully quantitative 3D analysis of meniscus shape, position, and SI, including the body, as well as the anterior and posterior horns. Few studies dealt quantitatively with the meniscus. In a cadaver study, Bowers et al. (26) reported MRI-based meniscus segmentation to provide volume estimates within 5% of the true values (determined by water displacement), with a test-retest error of 4% (26). Our work extends previous quantitative investigations on meniscus shape and position in OA (14, 15) and is the first to apply quantitative measurement technology to full 3D analysis. A high spatial resolution was required and contiguous 2mm slices were used, whereas previous work in OA was based on 3mm slices with 1mm gaps (14, 15). To compensate for the lower signal- and contrast-to-noise ratios in the higher-resolution images, two excitations were averaged, and the acquisition time of 4:44 min was well tolerated by all participants. The test-retest (resegmentation) precision was satisfactory for most parameters: for the tibial surface and the meniscus thickness, it was similar to that for resegmentation of the joint surface and cartilage thickness using cartilage-specific high-resolution sequences reported previously (27). In contrast, the meniscal width and the minimal and maximal medial extrusion distance displayed errors larger than the intersubject variability. This is plausible as these latter measures were based on few measurement points only, which are sensitive to small inconsistencies in segmentation. A limitation of the study is that test-retest precision was not determined for repositioning of the knee or between different observers.

Other limitations of the study include the small sample size and the fact that only women were studied, the use of coronal slices only, and that measurements of meniscal shape and position were made without joint loading, while the meniscus performs its mechanical function primarily under load bearing. Nevertheless, meniscal surfaces, volume, and thickness were significantly greater in OA than in non-OA knees, although body height and joint surface areas of MT were similar between the groups and although participants with meniscus surgery were excluded from the study. A potential explanation is that the MM may undergo swelling with degeneration and tears, which are known to be more frequent in OA than in healthy knees. Interestingly, meniscal hypertrophy was recently also reported by Jung et al. (16) in late-stage OA, with two-dimensional measurements of meniscal height being similar to our 3D measurements of maximal meniscus thickness.

The medial extrusion (MM.mEx.Me and MM.mEx.Max) was elevated in OA versus non-OA knees, our 3D results being quantitatively similar to previous two-dimensional measurements (14). The fact that the distance of the internal meniscus margin to the external border of MT.ACdAB (MM.OvD), the area of MT.ACdAB covered by the MM, and the area of MM covered by MT.ACdAB did not differ between OA and non-OA knees indicates that the relative position of the internal margin of MM to the tibial surface was similar in OA and non-OA knees. These observations are explained by the larger size of the tibial surface (TA), which results in the EA being farther away from the external border of ACdAB. Hence, the medial extrusion (MM.mEx.Me) and uncovered areas of the tibial meniscal surface (MM.TA.Uncov) were significantly greater in OA than in non-OA knees, whereas MM.TA.Cov (and MT.ACdAB.Cov and MT.ACdAB.Uncov) were unchanged. Differences in the medial extrusion (MM.mEx.Me) between OA and non-OA knees were most obvious in the meniscus body.

The higher signal of MM in OA knees was expected since meniscus damage is characterized by increased signal, mostly in the posterior horn (13). A limitation of our approach is that the signal was not normalized to an external calibration standard or to another tissue in the MR images. Further work is necessary to test whether quantitative analysis of SI change is potentially superior to semiquantitative scoring (13, 28).

The observations from this study need to be confirmed in larger cohorts, including men, since the current pilot study only involved a small number of female participants. Larger cohorts will also be required to establish normal values and the normal variation in young healthy participants. Semiquantitative readings (T.M.L.) of the MR images of the KLG0 participants revealed that none of the KLG0 knees had cartilage changes in the medial femorotibial compartment (where the meniscus was measured), but one participant had a cartilage defect of <50% (thickness) in the lateral femoral condyle, one had a >50% defect in the lateral tibia, and seven had cartilage defects in the femoropatellar joint (four full thickness, three <50%). It is unclear how these defects (in other compartments) affect the status of the MM. Another limitation is that the imaging protocol used (coronal images) did not allow us to measure anterior extrusion of the meniscus, as described previously (15) in sagittal slices. As the slice thickness of the spin-echo sequence used here was 2mm, the spatial resolution for providing measurements of anterior subluxation was insufficient to provide this measurement. A viable solution is to analyze radially orientated images rotated around the center of the tibial plateau, to use isotropic imaging, or to perform additional analyses on sagittal images. It has been shown that semiquantitative MR readings of meniscal pathology are associated with knee pain (3–6) and OA progression (i.e., cartilage loss) (7–12). Future applications of the technique presented here will have to test whether the quantitative endpoints proposed display stronger relationships with clinical (pain and function) and structural outcomes (cartilage loss) than previous semiquantitative MRI readings, and these analyses may also include the lateral meniscus. In this context, longitudinal studies are needed as the current cross-sectional analysis cannot reveal whether quantitative differences of MM are causes or consequences of OA.

The measurement technology proposed is not confined to OA. For example, Kessler et al. (29) reported a 10% decrease in meniscus volume after a 20-km distance running, and recovery within 1 h (30), and our 3D technology can be used to measure the impact of load bearing and various physical activities on 3D parameters of meniscus shape and position, including subregions. Similarly, shape and position of the meniscus could be determined in variable functional load-bearing positions of the knee (24), particularly when using open magnets (23).

In conclusion, this study presents a technique for fully quantitative 3D analysis of meniscus shape, position, and SI, including the body segment, as well as the anterior and posterior horns. The presented pilot data indicated a greater size, increased medial extrusion, and elevated SI of the MM in OA compared with non-OA knees, particularly in the body. The quantitative measurement technology proposed may be used to further explore the relationship of meniscus shape, position, and signal with symptoms and progression of OA and to elucidate the function of the meniscus during load bearing and motion.


We thank Pfizer Inc. for funding this study, Choong-soo Shin for the assistance with the data acquisition, Thelma Munoz for the subject recruitment and management, Karen Huerta for the x-ray acquisitions, and Martin Hudelmaier for the quality control of the MR images and the data conversion.