Nefise Cagla Tarhan 1,2 Christine. Chung 1 urea Valeria Rosa Mohana-orges 1 Tudor Hughes 1 Donald Resnick 1 Received September 30, 2003; accepted after revision February 2, 2004. 1 Department of Radiology, V Medical Center, University of California San Diego, 3350 La Jolla Village Dr., San Diego, C 92161. ddress correspondence to C.. Chung. 2 Present address: Department of Radiology, askent University, Faculty of Medicine, nkara, Turkey. JR 2004;183:9 15 0361 3X/04/1831 9 merican Roentgen Ray Society Meniscal Tears: Role of xial MRI lone and in Combination with Other Imaging Planes OJECTIVE. The purpose of this study was to determine the reliability of standard axial MR images alone in the diagnosis of meniscal tears of the knee and in combination with other imaging planes. MTERILS ND METHODS. Sixty-two patients (55 men, seven women; age range, 23 68 years) with a prior MRI examination who underwent arthroscopic surgery of the knee during a 1-year period were included in the study group. Images were independently reviewed for identification of meniscal tears by two musculoskeletal radiologists blinded to arthroscopic findings. Sequences for meniscal evaluation included axial fat-saturated fast spin-echo proton density, coronal fat-saturated fast spin-echo proton density, and sagittal fast spin-echo proton density with 4- to 5-mm slice thicknesses. Imaging groups for evaluation were axial, coronal, sagittal, axial and sagittal, axial and coronal, and coronal and sagittal. Observers reported a confidence level for the presence or absence of meniscal tear in all imaging groups based on a 5-point scale. Statistical analysis considered medial and lateral menisci separately. RESULTS. Forty patients had medial meniscal tears, and 16 had lateral meniscal tears at arthroscopy. For medial and lateral meniscal tears, the accuracy (79% and 71%, respectively) of imaging in the axial plane was comparable to other imaging groups but the mean confidence levels (2. and 3.00, respectively) were low. In one patient, the axial plane alone correctly showed that no tear was present. No statistically significant difference was observed between imaging plane groups of both menisci in the diagnosis of meniscal tears (p > 0.05). The axial plane increased the accuracy of sagittal and coronal planes of lateral meniscus when combined. CONCLUSION. In standard knee MRI examinations, the axial imaging plane may be valuable for the detection and characterization of meniscal tears. M RI is the preferred diagnostic method for evaluation of internal derangements of the knee. The suspicion of meniscal tear constitutes one of the leading indications for MRI examination of this articulation [1]. Imaging in the sagittal and coronal planes has been proven to be valuable in the MRI diagnosis of meniscal tears [2 4]. The use of the axial plane for meniscal evaluation has been investigated to a lesser degree but has been reported to be helpful in meniscal assessment, particularly with respect to characterization of tears [5 7]. No previous study has addressed the accuracy of the axial plane for the detection of meniscal disease. The aim of this study was to determine the value of standard axial MR images alone and in combination with other imaging planes in the diagnosis of meniscal tears of the knee. Materials and Methods fter approval from the investigational review board at our institution had been obtained, the medical records of all patients who had an MRI examination of the knee over a 14-month period were retrospectively reviewed to identify a subpopulation of patients who underwent arthroscopy. This review yielded 62 patients (55 men and seven women; age range, 23 68 years; mean age, 46.8 years). The time interval between the MRI examination and arthroscopy ranged between 2 weeks and 16 months (mean, 20.8 weeks). MRI Protocol ll MRI examinations were performed with a 1.5-T unit (Signa, General Electric Medical Systems). Images were obtained using a dedicated knee coil. Three imaging sequences (one for each plane) from the standard knee imaging protocol were selected for this study and included axial fatsaturated fast spin-echo proton density (TR range/ JR:183, July 2004 9
Tarhan et al. TE range, 2,650 2,717/12 54), coronal fat-saturated fast spin-echo proton density (3,000 3,400/ 15 65), and sagittal spin-echo proton density (TR range/te, 2,117 2,367/30). Other imaging parameters included the following: number of excitations, 2; matrix, 256 192 (axial and coronal) and 256 128 (sagittal); and field of view, 160 units (axial) and 140 units (coronal and sagittal). The sequences were obtained with a 4-mm slice thickness in 43 patients and with a 5-mm slice thickness in 19 patients because it was a retrospective review. Imaging nalysis The imaging sequences were separated into six evaluation groups as follows: I, axial; II, coronal; III, sagittal; IV, axial and sagittal; V, axial and coronal; and VI, coronal and sagittal. To control bias associated with case recognition, we randomly distributed the studies within each group according to sequence, and interpretation sessions were separated by at least 1 day. Image groups were independently reviewed by two musculoskeletal radiologists with different levels of expertise (observer 1, 1 year 6 months of fellowship training; and observer 2, 4 years of experience as a musculoskeletal radiologist). Reviewers were blinded to arthroscopic findings. The data acquired were evaluated for the presence or absence of meniscal tear. Meniscal tears were characterized as increased signal intensity on MRI extending to the articular surface of the meniscus or abnormal morphology, or both [1]. Observers were requested to determine the confidence level for the di- agnosis of meniscal tear in all image groups based on a 5-point scale: 1, completely uncertain of tear; 2, small likelihood of tear; 3, equivocal for tear; 4, probable tear; and 5, very certain of tear. similar 5-point scale was used for the diagnosis of no tear. Medial and lateral menisci were evaluated as separate interpretations. ll discordant results were reviewed by consensus. MRI findings were then compared with those of arthroscopy. MRI reports of all patients were also reviewed by a radiologist not involved in imaging interpretation, and other associated findings were evaluated. Findings recorded were ligamentous tears; osteoarthritic changes, including osteochondral defects and osseous abnormalities; and the presence or absence of meniscal cysts, intraarticular bodies, and joint effusion. Statistical nalysis Sensitivity, specificity, and accuracy for the detection of meniscal tears were calculated for each observer separately and for interpretations in consensus. Statistical analysis was performed with the McNemar test to interpret the differences in diagnosis between the six imaging groups. Mean confidence levels for every group of interpretation were calculated for all observers in consensus, and the difference between the confidence levels was analyzed by the Wilcoxon s signed rank test. Interobserver agreement for detection of meniscal tears for each plane group was calculated using kappa values. Kappa values were interpreted as follows: 0, poor; 0.01 0.20, slight; 0.21 0.40, fair; 0.41 0.60, moderate; 0.61 0., high; and 0. 1.00, almost perfect. Results t arthroscopy, 40 medial menisci and 16 lateral menisci were diagnosed as torn. The first observer correctly diagnosed 27 of 40 medial meniscal and four of 16 lateral meniscal tears, and the second observer correctly diagnosed 32 of 40 medial meniscal and seven of 16 lateral meniscal tears using the axial plane alone (Figs. 1 and 2). In one patient with no tear of the lateral meniscus, presence of no tear was correctly diagnosed on the basis of the axial plane. The findings of all other plane groups were false-positive. In 16 patients, on the basis of the axial plane alone the presence or absence of tear of the menisci was correctly diagnosed, whereas false-positive or false-negative findings based on other plane groups were recorded. In 14 of these patients, the lateral meniscus was involved; and in two, the medial meniscus was involved. The tears that were most often diagnosed with the axial plane alone were radial or complex tears of the posterior horn. ll findings of sensitivity, specificity, and accuracy are summarized in Table 1 for each imaging group and for each observer. In the evaluation of the medial meniscus, the highest sensitivity (95%) was observed in imaging group VI (coronal and sagittal planes). Sensitivity in the axial plane (%) was Fig. 1. 66-year-old man with radial tear of lateral meniscus., xial fat-saturated fast spin-echo proton density weighted image (TR/TE, 2,717/54; slice thickness, 4 mm) of right knee shows well-defined linear increased intensity (arrow) in body of lateral meniscus starting from free edge., Coronal fat-saturated fast spin-echo proton density weighted image (3,750/60; slice thickness, 4 mm) shows tear as vertical defect in body of lateral meniscus (arrow). 10 JR:183, July 2004
MRI of Meniscal Tears Fig. 2. 62-year-old man with bucket-handle tear of medial meniscus., xial fat-saturated fast spin-echo proton density weighted image (TR/TE, 2,717/54; slice thickness, 4 mm) of right knee reveals thinning and irregularity of medial meniscus body with bucket-handle component (arrows) extending from anterior to posterior., ucket-handle tear (arrow) is shown on coronal fat-saturated fast spin-echo proton density weighted image (3,267/60; slice thickness, 4 mm). comparable to other imaging plane groups. Specificity was increased with group V (axial and coronal planes) when compared with group I (axial plane) and group III (coronal plane). Highest accuracy (92%) was obtained in group III (coronal plane) and group TLE 1 V (sagittal and coronal planes). ccuracy in the axial plane alone (79%) was comparable to that of other imaging plane groups. The axial plane when combined with other imaging planes did not result in a significant change in overall accuracy. In the evaluation of the lateral meniscus, highest sensitivities were found in group II (sagittal plane), group III (coronal plane), and group IV (axial and sagittal planes) (%). Sensitivity was lower (37.5%) in the axial plane when compared with the other MRI and rthroscopic Correlation for Each Observer First Observer Plane Second Observer In Consensus Group Sensitivity Specificity ccuracy Sensitivity Specificity ccuracy Sensitivity Specificity ccuracy Medial meniscus xial Sagittal Coronal xial and sagittal xial and coronal Coronal and sagittal I II III IV V VI 67.5.5.5 92.5 71.5.5 95 79 92 92.5.5 95 91 91 79 92 92 Lateral meniscus xial Sagittal Coronal xial and sagittal xial and coronal Coronal and sagittal I II III IV V VI 25 69 56 62.5 96 98 91 73 43 56 56 69 70 63 79 37.5 62.5 75 83 98 71 94 Note. ll numbers are percentages. p > 0.05. JR:183, July 2004 11
Tarhan et al. planes. Specificity was also increased in group V (axial and coronal planes) when compared with group I (axial plane) or with group III (coronal plane) for the lateral meniscus. The highest accuracy was observed in group III (coronal plane) and in group IV (axial and sagittal planes) (% and 94%, respectively). ccuracy in the axial plane alone for the lateral meniscal tears was also comparable to other imaging plane groups (71%). The axial plane when combined with other planes increased the accuracy of the sagittal plane for the first observer and the accuracy of the coronal plane for the second observer. Group IV (axial and sagittal planes) had the highest accuracy (94%) of any plane group with regard to consensus interpretation, including group VI (coronal and sagittal planes) (%). When diagnosis of meniscal tears was compared between the plane groups, no statistically significant difference was observed in assessment of the medial or lateral meniscus (p > 0.05). The mean confidence levels of each imaging group are summarized for both menisci in Table 2. The axial plane had the lowest mean confidence levels for both the medial and lateral menisci. These levels proved significantly lower for this imaging plane than all other imaging plane groups (p < 0.001). Group IV (axial and sagittal planes) had significantly higher mean confidence levels compared with group V (axial and coronal planes) and group III (coronal plane) for the TLE 2 second observer with regard to the lateral meniscus (p < 0.05). The kappa values are given in Table 3 for the assessment of both menisci. Interobserver agreement was high for the axial plane with regard to the medial meniscus (κ = 0.64) and moderate for the axial plane with regard to the lateral meniscus (κ = 0.41). The lowest kappa value was obtained for group III (coronal plane) with regard to the lateral meniscus (κ = 0.35, fair) and highest for group VI (sagittal and coronal planes) with regard to the medial meniscus (κ = 0., almost perfect). ssociated MRI findings that were identified included osteoarthritic changes (n = 39), anterior cruciate ligament tear (n = 12), medial collateral ligament tear (n = 5), posterior cruciate ligament tear (n = 1), patellar tendon rupture (n = 1), popliteus muscle injury (n = 1), semimembranosus tendon avulsion (n = 1), joint effusion (large, n = 10; small, n = 5), meniscal cyst (n = 5), intraarticular bodies (n = 5), and osteochondritis dissecans (n = 1). In three patients, either motion or metallic artifact was present leading to degradation of the MRI study. In 31 patients with erroneous interpretation of meniscal disease, osteoarthritic changes were noted (Fig. 3), including secondary osteoarthritis in one patient with osteochondritis dissecans. In nine of 12 patients with anterior cruciate ligament tears, meniscal disease was misinterpreted. In six of these patients, false-positive interpretation of a lateral meniscus tear was made (Fig. 4). Incorrect diagnosis Mean Confidence Levels of Each Plane for the First and Second Observers Plane Group First Observer Second Observer Medial meniscus xial I 2. 2.95 Sagittal II 4.50 4.53 Coronal III 4.48 4.59 xial and sagittal IV 4.53 4.53 xial and coronal V 4.45 4.56 Coronal and sagittal VI 4.54 4.59 Lateral meniscus xial I 2. 2.53 Sagittal II 4.29 4.35 Coronal III 4.25 4.20 xial and sagittal IV 4.20 4.43 xial and coronal V 4.30 4.25 Coronal and sagittal VI 4.48 4.22 Note. p < 0.01 for axial plane compared with other planes. occurred in four patients with variable amounts of joint effusion and in three patients with tendon injuries. In four patients who had only fraying of the lateral meniscus noted at arthroscopy, false-positive interpretations were rendered by both observers. In one patient with a small medial meniscal tear that was stable arthroscopically, both observers rendered falsenegative interpretations. Discussion In the evaluation of the menisci of the knee, MRI has been found to be highly accurate for the diagnosis of meniscal tears [1]. Many investigators have suggested that the sagittal imaging plane is the most useful for the detection of meniscal disease and that the coronal plane improves the overall diagnostic accuracy [2 4]. For most institutions, the axial fast spin-echo fat-saturated proton density sequence is the only technique used for evaluation in the axial plane [7]. There have been reports on thin-section axial 3D gradient-echo sequences in the evaluation of the menisci [5, 6]. These studies have shown that the extent and severity of meniscal tears were better delineated compared with coronal and sagittal images, especially when disease existed in the periphery of the meniscus. Small radial tears of the free edge and displaced fragments have also been reported to be better visualized in this plane [5, 6]. These sequences, however, are not routinely used in clinical practice, primarily related to con- TLE 3 Interobserver greement for Each Meniscus Plane Group κ Medial meniscus xial I 0.64 Sagittal II 0. Coronal III 0. xial and sagittal IV 0. xial and coronal V 0. Coronal and sagittal VI 0. Lateral meniscus xial I 0.41 Sagittal II 0.66 Coronal III 0.35 xial and sagittal IV 0.63 xial and coronal V 0.58 Coronal and sagittal VI 0.47 12 JR:183, July 2004
MRI of Meniscal Tears cerns regarding the study acquisition time. preliminary report [7] suggested that most of the different types of meniscal tears have discrete appearances in axial fat-saturated fast spin-echo proton density images that provide additional information regarding lesion characterization. Their results, however, were not compared with those of arthroscopy. ccuracy for the diagnosis of meniscal tears with MRI has been studied by different authors, and several different sensitivity, specificity, and accuracy levels have been found [1, 8 10]. Sensitivity and specificity ranged between % and 97% and % and 91% for the medial meniscus and between C D Fig. 3. 54-year-old man with osteoarthritis and chondromalacia with misinterpretation of axial images for lateral and medial menisci. and, Sequential axial fat-saturated fast spin-echo proton density weighted images (TR/TE, 2,700/54; slice thickness, 4 mm) of right knee show medial and lateral menisci as intact. C, Coronal fat-saturated fast spin-echo proton density weighted image (3,767/60; slice thickness, 4 mm) shows tear (arrow) of lateral meniscus as tapering and increased intensity. D, Sagittal spin-echo proton density weighted image (2,167/30; slice thickness, 4 mm) shows tear (arrow) on body of medial meniscus as vertical defect. JR:183, July 2004 13
Tarhan et al. Fig. 4. 45-year-old woman with anterior cruciate ligament tear with false-positive interpretation of lateral meniscus tear., xial fat-saturated fast spin-echo proton density weighted image (TR/TE, 2,717/54; slice thickness, 4 mm) of right knee shows linear increased signal intensity (arrow) in anterior horn of lateral meniscus thought to be tear., Coronal fat-saturated fast spin-echo proton density weighted image (3,750/60; slice thickness, 4 mm) shows meniscus is intact. 69% and 92% and 91% and 98% for the lateral meniscus, respectively, in these previous studies [1, 8, 9]. Our findings were similar to the highest level of these ranges for both sensitivity and specificity. In these previous studies, only coronal and sagittal planes were used together with no consideration given to the value of the axial plane. Our study showed high accuracy levels for the detection of meniscal disease using the axial plane alone. Confidence levels, however, were low with regard to the assessment of both the medial and lateral menisci. In sagittal and coronal orientations, menisci are seen in almost 13 or 14 images, but in the axial plane with 4- to 5-mm slice thickness, they are seen only in two or three images. This is one potential reason for low confidence levels related to meniscal tears associated with assessment of the axial plane. We evaluated axial images that were characterized by a 4- to 5-mm slice thickness because such images are used in routine MRI protocols of the knee. lthough spatial resolution decreases with 5-mm-thick sections compared with thinner slices, signal-to-noise ratio increases and this may compensate, to some degree, for the lower spatial resolution. It has been shown that 5-mm-thick images 14 do not have lower diagnostic accuracy compared with that of thinner slices used for MRI studies of the knee [2]. With respect to the diagnosis of meniscal tears in axial images, confidence intervals may also reflect a relative lack of experience of the observers. s previously mentioned, the sagittal and coronal images are generally relied on for the diagnosis of meniscal disease. ecause no statistically significant difference was evident in the diagnosis of meniscal tears when comparing the axial imaging plane alone and all other imaging groups, it can be postulated that the axial plane is also valuable in the evaluation of meniscal tears. Our observations indicate that the overall accuracy for the diagnosis of medial meniscus tear was not increased by the addition of the axial plane to other imaging planes. ccuracy and specificity were, however, improved by the addition of the axial plane to other imaging planes for the lateral meniscus. The highest accuracy was observed in group IV (axial and sagittal planes) for assessment of the lateral meniscus, and mean confidence levels were significantly higher for this imaging group. Most of the true-positive and true-negative evaluations compared with other plane groups were found for the lateral meniscus. Therefore, the axial plane was especially helpful when interpreted with the sagittal plane in the evaluation of the lateral meniscus. We have found that most of the correctly diagnosed tears on the axial plane alone were radial or complex tears of the posterior horn. These findings will lead us to future studies of the accuracy of the characterization of meniscal tears with the addition of the standard axial plane to other planes. We believe that the low kappa values, particularly with regard to the lateral meniscus, may have been related to patient selection. Most of the patients who underwent arthroscopy had cartilage abnormalities and osteoarthritis (n = 39), which contributed to false-positive and false-negative evaluations when only one plane was used for assessment. We also had a significant number of patients with tears of the anterior cruciate ligament who had false-positive interpretations of meniscal disease. n anterior cruciate ligament tear is especially important in patients with false-positive diagnoses of lateral meniscal tears. Fraying and small stable tears of the menisci were also mistaken during our evaluation. In addition, a long interval (> 1 year) between MRI and arthroscopy JR:183, July 2004
MRI of Meniscal Tears in four patients may account for this finding as a result of interval healing or interval development of lesions or both. This study provided accuracy levels for the assessment of both meniscal tears in the axial plane alone and in combination with other imaging planes. The high levels of accuracy for detection of medial meniscal disease and increased levels of accuracy for detection of lateral meniscal disease with the addition of axial images shown in this study suggest that the axial imaging plane may prove helpful for the detection of meniscal tears in routine studies of the knee. References 1. Rubin D. MR imaging of the knee menisci. Radiol Clin North m 1997;35:21 44 2. Herman LJ, eltran J. Pitfalls in MR imaging of the knee. Radiology 1988;167:5 7 3. Tyrrell RL, Gluckert K, Pathria M, Modic MT. Fast three-dimensional MR imaging of the knee: comparison with arthroscopy. Radiology 1988;166:5 2 4. Reeder JD, Matz SO, ecker L, ndelman SM. MR imaging of the knee in the sagittal projection: comparison of three-dimensional gradient-echo and spin-echo sequences. JR 19;153:537 540 5. raki Y, Ootani F, Tsukaguchi I, et al. MR diagnosis of meniscal tears of the knee: value of axial three-dimensional Fourier transformation GRSS images. JR 1992;158:5 5 6. ubel S, Heyd RL, Thaete FL, Wozney P. MR knee imaging: axial 3DFT GRSS pulse sequence versus spin-echo imaging for detecting meniscal tears. Magn Reson Imaging 1992;10:531 539 7. Lee JHE, Singh TT, olton G. xial fat-saturated FSE imaging of the knee: appearance of meniscal tears. Skeletal Radiol 2002;31:3 395 8. Fischer SP, Fox JM, Del Pizzo W, Friedman MJ, Snyder SJ, Ferkel RD. ccuracy of diagnoses from magnetic resonance imaging of the knee: a multi-center analysis of one thousand and fourteen patients. J one Joint Surg m 1991;73:2 10 9. Crues JV, Mink J, Levy TL, Lotysch M, Stoller DW. Meniscal tears of the knee: accuracy of MR imaging. Radiology 19;164:445 448 10. oeve F, Davidson R, Staab EV. Magnetic resonance imaging in the evaluation of knee injuries. South Med J 1991;:1123 1127 The full text and images from the merican Journal of Roentgenology may also be viewed online at www.arrs.org or www.ajronline.org. JR:183, July 2004 15