| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Radiology, University of Maryland, 22 S. Greene St., Baltimore,
MD 21201-1595.
2 Present address: Radiology Imaging Associates, 8200 E. Belleview Ave., Ste.
124, Greenwood Village, CO 80111.
3 Department of Radiology, University of Colorado School of Medicine, 4200 E.
Ninth St., Denver, CO 80262.
4 Present address: Department of Radiology (A-21), Cleveland Clinic Foundation,
9500 Euclid Ave., Cleveland, OH 44195-5021.
Received October 10, 2001;
accepted after revision May 7, 2002.
Presented at the annual meeting of the Radiological Society of North
America, Chicago, November 1999.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. We retrospectively reviewed MR images of the knee in 54 patients for whom arthroscopic results were available. All MR imaging studies included fast spin-echo proton density-weighted coronal and axial sequences as part of our routine protocol. Evaluation of the articular surfaces was performed by three independent observers who were unaware of the arthroscopic results. The cartilage surfaces were graded using a 3-point system, and results were compared with arthroscopic findings.
RESULTS. Of 324 cartilage surfaces evaluated, arthroscopy showed 241 surfaces as normal, 56 as containing partial-thickness defects, and 27 as containing full-thickness defects. Compared with arthroscopic data, sensitivity of MR imaging for the three reviewers was 59-73.5%; specificity, 86.7-90.5%; positive predictive value, 60.5-72.6%; negative predictive value, 86.0-90.8%; and accuracy, 79.6-86.1%. Interobserver variability for the presence of disease, which was measured using the kappa statistic, was 0.63.
CONCLUSION. Fast spin-echo proton density-weighted MR imaging sequences can be used to evaluate the cartilage of the knee with accuracy comparable to that of previously reported cartilage-specific sequences.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Because of inherent magnetization transfer effects caused by the multiple refocusing pulses used in fast spin-echo sequences, articular cartilage shows lower signal intensity on fast spin-echo sequences than on spin-echo sequences [21,22,23]. In fast spin-echo proton density-weighted images, the resulting tissue contrast between articular cartilage and adjacent fluid and cortical bone provides a useful window in which to visualize the integrity of the hyaline articular cartilage [24]. Our experience has shown proton density-weighted images to be useful in the evaluation of other structures in the knee such as menisci, ligaments, and tendons. A single pulse sequence accurate for studying these structures in addition to the hyaline cartilage surfaces of the joint would provide a powerful and efficient imaging tool in an environment of increasing pressure to maximize the use of scanner time. We therefore used a fast spin-echo proton density-weighted sequence to assess the morphology and thickness of the hyaline articular cartilage of the knee in a group of patients, using arthroscopy as the reference standard.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MR imaging was performed on a 1.5-T Signa 5.5 superconducting magnet (General Electric Medical Systems, Milwaukee, WI) using a quadrature array volume knee coil (Medical Advances, Milwaukee, WI). Scanning parameters for the coronal proton density-weighted sequence were TR/TEeff, 3500/19; echo train length, 8; field of view, 18 cm; slice thickness, 4 mm; interslice gap, 1 mm; matrix, 256 x 512 pixels; and 2 acquisitions, with a total imaging time of 3 min 51 sec. The phase encoding direction was left to right for this sequence. The axial proton density-weighted images were part of a dual-echo sequence, with parameters as follows: TR/first-echo TEeff, second-echo TEeff, 4000/12, 84; echo-train length, 8; field of view, 16 cm; slice thickness, 3 mm; interslice gap, 0.6 mm; matrix, 224 x 256 pixels; and 1 acquisition, with a total imaging time of 4 min. Phase encoding was in the anteroposterior direction for the axial images.
For the purposes of reviewing MR images, cartilage surfaces were divided into six regions—the patella, trochlea, medial and lateral femoral condyles, and medial and lateral tibial surfaces—for a total of 324 surfaces in our 54 patients. Retrospective evaluation of these articular surfaces on the coronal and axial fast spin-echo proton density-weighted images was performed by three experienced musculoskeletal radiologists acting as independent observers, each of whom was unaware of the arthroscopic results. The reviewers had different levels of experience using this particular sequence for characterizing hyaline cartilage; observer 1 had 5 years' experience in the technique, observer 2 had 2 years' experience, and observer 3 had no previous experience. A short training session preceded the study.
For a comparison of MR imaging findings with arthroscopic data, MR imaging and arthroscopic grading had to be compared using similar grading systems. Arthroscopic grading was accomplished in a manner similar to that described in previous reports [14,15,16,17,18, 25, 26] based on the system of Shahriaree [26], in which grade 0 indicates normal; grade 1, softening of hyaline cartilage without a morphologic defect; grade 2, shallow fibrillation, ulceration, or erosion composing less than 50% of the total thickness of the cartilage surface; grade 3, partial-thickness defect of more than 50% but less than 100% of the cartilage thickness; and grade 4, full-thickness defect. In our MR imaging grading system, we grouped grade 1 lesions with normal lesions because of the lack of morphologic abnormality in such cases, and we grouped all partial-thickness (grades 2 and 3) lesions together because we thought that the degree of partial-thickness defects was open to significant subjective variability on the part of both the arthroscopist and, to a lesser extent, the radiologist. Thus, our MR imaging grading system comprised three levels: normal, partial-thickness chondral defects, and full-thickness defects.
Defects were identified on the MR images on the basis of morphologic and signal changes in a defined cartilage surface that suggested either focal thinning or discontinuity of the cartilage or a lack of definable hyaline cartilage on the surface being evaluated. Abnormal signal in a cartilage surface with no visualized morphologic abnormality was scored as normal. Occasionally, less objective measurements were provided by the arthroscopist (e.g., "mild chondromalacia" was interpreted to mean a partial-thickness defect). If a surface was described with two grades (e.g., "grade 3-4"), the higher number (more severe grade) was recorded. Surfaces categorized as grade 1 by the arthroscopist (softening with no morphologic defect) were considered normal for the purpose of this study. If a joint surface was not specifically mentioned, it was assumed to be normal. Subchondral marrow changes were not specifically noted and were not used as prima facie evidence of chondral loss without a visualized cartilage defect overlying the area. Results of the MR imaging interpretations were subsequently compared with dictated arthroscopic results.
To determine sensitivity, specificity, positive and negative predictive values, and accuracy for each reviewer, we compressed imaging and surgical results into categories of positive or negative for disease (i.e., partial- and full-thickness defects were grouped together). The results of each reviewer were graded independently against the gold standard of the arthroscopy report. The McNemar test was applied to each reviewer's results to determine the presence and magnitude of systematic bias (i.e., the bias of any one reviewer to interpret all cases as a higher or a lower grade than that interpreted by the whole group), if any. Nonweighted two- and three-level kappa statistics were used to measure the degree of interobserver variability.
Arthroscopic examination of our patient population yielded 15 anterior cruciate ligament tears (28%; one partial), three posterior cruciate ligament tears (6%), 22 medial meniscus tears (41%), 16 lateral meniscus tears (30%; nine patients [17%] had both medial and lateral tears), and two tears (4%) each of the medial and lateral collateral ligament complexes.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
|
|
|
Surgical grading of the 324 cartilage surfaces evaluated was as follows: "normal" (grades 0 and 1), 241 (74.4%) (Figs. 1 and 2A,2B,2C); partial-thickness defects (grades 2 and 3), 56 (17.3%) (Fig. 3A,3B,3C); and full-thickness defects (grade 4), 27 (8.3%) (Figs. 2A,2B,2C and 3A,3B,3C). Of the partial-thickness defects, 29 (9.0% of the total) were grade 2 and 27 (8.3% of the total) were grade 3. The distribution of these lesions is illustrated in Table 1. Interpretation of the dictated arthroscopy reports was usually straightforward, with grades assigned to each cartilage surface according to the method of Shahriaree [26].
|
|
|
|
Sensitivity scores for the reviewers were 59.0-73.5%; specificity, 86.7-90.5%; positive predictive value, 60.5-72.6%; negative predictive value, 86.0-90.8%; and accuracy, 79.6-86.1%. To exclude reviewer bias, the McNemar test was performed for each reviewer, with no bias shown (Table 2). Sensitivity and specificity varied for the different cartilage surfaces. Sensitivity was greatest (80%) for the patella but was only 44% for lateral tibial plateau lesions. Specificity ranged from 75% for patellar lesions to 95% for lateral tibial plateau lesions (Table 3).
|
|
Variability among the three observers was examined using the kappa statistic in two ways. First, assuming two MR grading results (positive for disease versus negative) provided a two-level kappa coefficient of 0.63 [27]. Second, the kappa statistic was reapplied with MR grading results reflecting three levels (normal, partial-thickness, and full thickness) to more accurately measure the agreement between the three observers using the entire grading scale. This kappa value was 0.54 [27].
Because 324 cartilage surfaces were each evaluated by three reviewers, we had a total of 972 individual surface interpretations. Of the 104 undercalled interpretations among the three reviewers, 14 were full-thickness lesions at arthroscopy that were considered normal on MR imaging. Nine of these were caused by three defects (two trochlear, one medial femoral condyle) identified at arthroscopy that were considered normal by all three reviewers (Fig. 4A,4B). An additional 23 full-thickness defects at arthroscopy were graded as partial thickness on MR imaging (Table 4). All the remaining undercalled interpretations (n = 67) were for partial-thickness defects at arthroscopy that were considered normal on MR imaging (Table 4). Of these, 43 were grade 2 and 24 were grade 3 at arthroscopy.
|
|
|
Of the 106 overcalled interpretations among the three reviewers, six were interpreted as full thickness on MR imaging but were normal at arthroscopy. In each of these cases, only one of the three reviewers identified the lesion as being full thickness in nature (Fig. 5). An additional 24 partial-thickness lesions at arthroscopy were graded full thickness on MR imaging (Table 5). All the other overcalled interpretations (n = 76) were described as partial-thickness defects on MR imaging but were normal at arthroscopy (Fig. 6). Of these, 55 were grade 2 and 21 were grade 3 on MR imaging.
|
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Fat-suppressed three-dimensional spoiled gradient-recalled sequences have shown promising results, with sensitivities of 81-93% and specificities of 94-97% [13,14,15,16,17,18,19,20]. However, such sequences are time-consuming and require postprocessing, reformatting, and manipulation before interpretation. Furthermore, because of their inherently narrow spectrum of contrast resolution, the three-dimensional sequences are not useful for characterizing other intra- or extraarticular disorders.
Early reports of the use of spin-echo images in the evaluation of the patellofemoral joint noted that cartilage is rendered virtually devoid of signal on long-TE (T2-weighted) sequences and is thus indistinguishable from cortical bone [4, 8]. The addition of proton density—weighted images improved the contrast resolution between cartilage and bone, but such sequences showed lower accuracy for the detection of cartilage lesions, primarily because of the similarity in signal for fluid and for cartilage [4]. The inherent magnetization transfer contrast effect imparted to hyaline cartilage as a result of the multiple radiofrequency-refocusing pulses delivered in fast spin-echo proton density—weighted sequences provides the necessary contrast resolution to differentiate the articular cartilage from adjacent joint fluid or synovial tissue and cortical bone [21,22,23,24]. In fact, the presence of joint fluid is not necessary to visualize the borders of the individual cartilage surfaces with such a sequence (Fig. 2A,2B,2C). The combination of the established usefulness of fast spin-echo proton density—weighted sequences in the evaluation of ligamentous [30] (Fig. 1) and meniscal [31] (Fig. 3A,3B,3C) disorders with the ability to identify hyaline cartilage defects makes fast spin-echo proton density—weighted sequences a highly versatile diagnostic radiology tool that is widely available. Recent work by Potter et al. [24] showed the potential usefulness of this technique in the evaluation of articular cartilage.
Our maximum sensitivity of 73.5%, specificity range of 86.7-90.5%, and range in accuracy of 79.6-86.1% are similar to those in published reports of cartilage-specific sequences [15,16,17]. The moderate correlation coefficient among reviewers may be in part related to the significantly different degrees of experience with the technique among the reviewers. The relatively low sensitivity for lateral tibial plateau lesions is similar to the findings of others [15, 24]. We hypothesize that this low sensitivity may be due to the relatively thin cartilage on this surface. Certain locations, such as the extreme posterior aspects of the femoral condyles or the inferior aspects of the trochlear surfaces, were believed to be difficult to evaluate because of the inherent curvature of the articular surfaces. Previous studies [15, 17,18,19, 24] with a variety of imaging techniques have described using sagittal and axial planes of reconstruction for the determination of lesions, whereas our study used axial and coronal images. The lack of sagittal images may have impeded the detection of hyaline cartilage defects in these locations and thus affected the overall accuracy of the technique.
The appearance of diffuse thinning of a chondral surface caused an interpretive problem. Because the arthroscopist sees only the surface of the hyaline cartilage, he or she has no way to gauge the depth of the substance. MR imaging shows the true width of the cartilage layer; if a surface is substantially thinned compared with the perceived normal value but without a focal defect, the radiologist must consider what the arthroscopist will see in addition to the "true" nature of cartilage loss. The normal depth of patellar hyaline cartilage has been previously documented; it is typically 5-6 mm [32]. Such values may vary from individual to individual, however. In our study, the patellar surface was the most likely to be interpreted as diffusely thinned on MR imaging (Fig. 5). Although not statistically meaningful, the small number of such cases that we encountered were all graded normal at arthroscopy. Thus, we believe that a cartilage surface showing diffuse thinning on MR imaging, compared with subjective or objective normative values based both on the literature and on individual experience, may not correlate well with arthroscopic abnormalities.
Although several arthroscopic grading systems are in use, many surgeons use some variation of the system originally described by Outerbridge [25]. Our study was designed to evaluate mid- to high-grade cartilage lesions, which would correlate with grades 2-4 in the Shahriaree system [26]. On the basis of our previous experience, we thought that attempting to identify grade 1 lesions (softening of hyaline cartilage with no morphologic defect) would likely not be particularly accurate. This experience has been described by other authors [15,16,17, 24, 29]. Unlike those other authors, we chose to group all partial-thickness defects together because, in our experience, trying to distinguish between partial-thickness lesions of less than 50% of total chondral depth (Shahriaree grade 2) and those between 50% and 100% of total chondral depth (Shahriaree grade 3) is fraught with subjective variation. Most of the false-negative partial-thickness lesions in this study were grade 2 at arthroscopy.
Our study has several limitations. First, the arthroscopic gold standard we used is subject to operator error, and several cartilage lesions reported by all three reviewers that had imaging characteristics identical to surgically proven defects were considered normal at surgery; it is conceivable that some of these lesions represent surgical false-negatives. Second, because the orthopedic surgeons had the MR imaging reports available to them before surgery, this knowledge may have biased their decision to perform surgery or may have biased their arthroscopic findings. Third, the arthroscopy reports were sometimes less specific regarding the location or grade of cartilage defect than the retrospective MR image interpretations, so some generalization was required in the interpretation of the arthroscopic data. Fourth, the relatively low percentage (54/320, 16.9%) of patients imaged in our series who went on to arthroscopy may represent a source of selection bias. Finally, the lack of sagittal images could be related to lower sensitivity for lesions in certain areas.
The results of our study indicate the ability of fast spin-echo proton density—weighted MR imaging to depict the articular cartilage of the knee with accuracy comparable to that of previously described cartilage-specific sequences. Unlike cartilage-specific sequences, fast spin-echo proton density—weighted sequences require no additional processing or manipulation after acquisition, can provide valuable information about other knee structures such as ligaments and menisci [30, 31], and can be easily incorporated into a standard knee imaging protocol on most existing MR imaging units. Our use of an echo train of 8 for our proton density—weighted sequence provided a robust magnetization transfer effect that allowed the distinction of fluid from cartilage. However, high echo trains may decrease spatial resolution because of k-space filling limitations, and accuracy for detecting meniscal tears may consequently suffer [31]. Although the reliability of fast spin-echo proton density—weighted images in the evaluation of meniscal disorders has been a topic of continued debate and remains under scrutiny, it is clear that accurate results can be achieved with variations of the technique. In our experience, fast spin-echo proton density—weighted images can provide a comfortable and effective balance that allows accurate evaluation of hyaline articular cartilage while retaining diagnostic use for the description of other tissues in the knee.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  R. Kijowski, D. G. Blankenbaker, J. L. Klaers, K. Shinki, A. A. De Smet, and W. F. Block Vastly Undersampled Isotropic Projection Steady-State Free Precession Imaging of the Knee: Diagnostic Performance Compared with Conventional MR Radiology, April 1, 2009; 251(1): 185 - 194. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. Kijowski, D. G. Blankenbaker, K. W. Davis, K. Shinki, L. D. Kaplan, and A. A. De Smet Comparison of 1.5- and 3.0-T MR Imaging for Evaluating the Articular Cartilage of the Knee Joint Radiology, March 1, 2009; 250(3): 839 - 848. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. E. Gold, R. F. Busse, C. Beehler, E. Han, A. C. S. Brau, P. J. Beatty, and C. F. Beaulieu Isotropic MRI of the Knee with 3D Fast Spin-Echo Extended Echo-Train Acquisition (XETA): Initial Experience Am. J. Roentgenol., May 1, 2007; 188(5): 1287 - 1293. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. R. Duc, P. Koch, M. R. Schmid, W. Horger, J. Hodler, and C. W. A. Pfirrmann Diagnosis of Articular Cartilage Abnormalities of the Knee: Prospective Clinical Evaluation of a 3D Water-Excitation True FISP Sequence Radiology, May 1, 2007; 243(2): 475 - 482. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K.-P. Jungius, M. R. Schmid, M. Zanetti, J. Hodler, P. Koch, and C. W. A. Pfirrmann Cartilaginous Defects of the Femorotibial Joint: Accuracy of Coronal Short Inversion Time Inversion-Recovery MR Sequence Radiology, August 1, 2006; 240(2): 482 - 488. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. Kijowski, P. Stanton, J. Fine, and A. De Smet Subchondral Bone Marrow Edema in Patients with Degeneration of the Articular Cartilage of the Knee Joint Radiology, March 1, 2006; 238(3): 943 - 949. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. P. Recht, D. W. Goodwin, C. S. Winalski, and L. M. White MRI of Articular Cartilage: Revisiting Current Status and Future Directions Am. J. Roentgenol., October 1, 2005; 185(4): 899 - 914. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. N. Masi, C. A. Sell, C. Phan, E. Han, D. Newitt, L. Steinbach, S. Majumdar, and T. M. Link Cartilage MR Imaging at 3.0 versus That at 1.5 T: Preliminary Results in a Porcine Model Radiology, July 1, 2005; 236(1): 140 - 150. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. Magee and D. Williams Detection of Meniscal Tears and Marrow Lesions Using Coronal MRI Am. J. Roentgenol., November 1, 2004; 183(5): 1469 - 1473. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. Roth, J. Jacobson, D. Jamadar, E. Caoili, Y. Morag, and J. Housner Quadriceps Fat Pad Signal Intensity and Enlargement on MRI: Prevalence and Associated Findings Am. J. Roentgenol., June 1, 2004; 182(6): 1383 - 1387. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. H. Haims, A. E. Moore, M. E. Schweitzer, W. B. Morrison, D. Deely, R. W. Culp, and H. P. Forman MRI in the Diagnosis of Cartilage Injury in the Wrist Am. J. Roentgenol., May 1, 2004; 182(5): 1267 - 1270. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Mohr, F. W. Roemer, H. K. Genant, C. Liess, and A. Sonin Using Fat-Saturated Proton Density-Weighted MR Imaging to Evaluate Articular Cartilage Am. J. Roentgenol., July 1, 2003; 181(1): 280 - 282. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
