| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Clinical Observations |
1 Neuroskeletal Imaging, 1344 S Apollo Blvd., Ste. 406, Melbourne, FL
32901.
2 Neuroskeletal Imaging, Orlando, FL.
3 Murrah Orthopedics, Orlando, FL.
Received March 25, 2005;
accepted after revision July 16, 2005.
Address correspondence to R. R. Ramnath
(rramnath{at}post.harvard.edu).
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CONCLUSION. We concluded from this study that 3-T MRI using fast spin-echo sequences is highly accurate in the detection of medial and lateral meniscal tears of the knee.
Keywords: fast spin echo • knee • meniscal tear • meniscus • MRI • MR technique • musculoskeletal imaging
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MRI
All patients were scanned on a 3-T short-bore Excite MRI system (GE
Healthcare). One of two extremity coils were used in the study: either a
quadrature knee coil with a chimney component (IGC-Medical Advances) or a
phased-array knee coil with either 4 or 8 channels (MRI Devices). Sequences
consisted of sagittal proton density fast spin echo (TR/TE, 3,600/16;
echo-train length, 6; 2-mm slice thickness; 0.2-mm gap; matrix, 416 x
256; auto-zero-fill interpolation [ZIP], 512; number of excitations [NEX], 2;
15-cm field of view; time, 5 minutes 48 seconds), sagittal fat-saturated
T2-weighted fast spin echo (TR/TE, 3,700/55; echo-train length, 12; 3-mm slice
thickness; 1-mm gap; matrix, 384 x 320; auto-ZIP, 512; NEX, 2; 15-cm
field of view; time, 4 minutes 22 seconds), coronal T1-weighted fast spin echo
(TR/TE, 750/15; echo-train length, 3; 4-mm slice thickness; 1-mm gap; matrix,
384 x 320; auto-ZIP, 512; NEX, 1; 16-cm field of view; time, 2 minutes
31 seconds), coronal fat-saturated T2-weighted fast spin echo (TR/TE,
3,400/55; echo-train length, 12; 4-mm slice thickness; 1-mm gap; matrix, 384
x 320; auto-ZIP, 512; NEX, 2; 16-cm field of view; time, 3 minutes 43
seconds), and axial fat-saturated T2-weighted fast spin echo (TR/TE, 4,200/55;
echo-train length, 12; 4-mm slice thickness; 1-mm gap; matrix, 384 x
320; auto-ZIP, 512; NEX, 2; 16-cm field of view; time, 3 minutes 43 seconds).
The total average sequence scanning time for each knee examination was
approximately 20 minutes with an allotted MRI time slot of 30 minutes. The
auto-ZIP feature on the GE Healthcare Excite platform automatically zero
interpolates matrix values greater than 256 to a value of 512.
|
|
|
Statistical Analysis
Statistical accuracy of MRI for detection of meniscal tears was calculated
and included determination of accuracy, sensitivity, specificity, positive
predictve value (PPV), and negative predictive value (NPV). Accuracy profiles
were determined for all tears (both medial and lateral menisci), and specific
profiles were calculated for medial menisci and lateral menisci. Accuracy is
defined as (true-positive [TP] + true-negative [TN]) / (TP + false-positive
[FP] + TN + false-negative [FN]). Sensitivity is defined as TP / (TP + FN),
specificity is TN / (FP + TN), PPV is TP / (TP + FP), and NPV is TN / (TN +
FN).
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Lateral Meniscus
Seventeen lateral meniscal tears were described at surgery (17/34, 50%).
MRI detected 16 of the 17 tears. Figure 4 shows a complex tear in the
posterior horn of the lateral meniscus, which was seen on MRI and at surgery.
One tear found at surgery was not seen on the MRI interpretation (Fig. 5), and
one tear found on MRI was not seen at surgery (Fig. 6). The results included
16 true-positives, one false-positive, 16 true-negatives, and one
false-negative. Therefore, the accuracy for the detection of lateral meniscal
tears was 94%, the sensitivity was 94%, the specificity was 94%, the PPV was
94%, and the NPV was 94%.
Both Menisci
Forty-two medial and lateral meniscal tears were seen at surgery (42/68,
62%), and MRI detected 40 of them. Two tears found on MRI were not seen at
surgery. The results included 40 true-positives, two false-positives, 24
true-negatives, and two false-negatives. Therefore, the overall accuracy for
the detection of medial and lateral meniscal tears was 94%, the sensitivity
was 95%, the specificity was 92%, the PPV was 95%, and the NPV was 92%.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The evaluation of menisci with MRI is a highly accurate endeavor with sensitivities ranging from 40-100% [1, 2, 4-7, 11, 12], specificities ranging from 72-100% [1, 2, 4-7], and accuracies ranging from 75-96% [1, 3, 6, 8, 9]. Our study produced results in the upper ends of these ranges (95% sensitivity, 92% specificity, and 94% accuracy for all menisci). However, several articles in the orthopedics literature suggest equal or better diagnostic accuracy with the use of a proper physical examination [6, 8, 13, 14]. Kocabey et al. [13] recently reported equivalent diagnostic accuracies of a physical examination by a skilled orthopedic surgeon for detecting medial and lateral meniscal tears. The article further advocated only obtaining an MRI in "more complicated and confusing" cases. Muellner et al. [6] showed a 95% accuracy, 97% sensitivity, and 87% specificity for the detection of meniscal tears with clinical examination alone. Stanitski [14] found a highly negative correlation between arthroscopic and MRI findings and stated, "MRI diagnoses added little guidance to patient management and at times provided spurious information." With the advent of 3-T MRI, the potential for greater spatial resolution and higher signal-to-noise ratio could be effectively used to improve the accuracy of 1.5-T systems and, thereby, further illustrate to orthopedic surgeons the value of MRI for routine assessment of meniscal abnormalities. Our results certainly showed a high diagnostic value of the 3-T MRI examination. However, whether the information gained on the MRI studies altered patient management was not assessed. A study comparing the clinical examination to 3-T MRI is warranted.
One potential limitation of our study was the exclusion of patients who did not proceed to surgery. Of the 140 patients who had MRI of the knee, only 34 proceeded to surgery in the time frame of this study; consequently, 106 MRIs were not evaluated. Presumably, most patients who underwent surgery did so because of a positive finding on the preoperative MRI. Therefore, our specificity and sensitivity data may fall short, especially because the true prevalence or absence of disease was not addressed in this study. In addition, the availability of the MRI reports to the surgeon before surgery may also result in some degree of surgical bias.
However, assuming that surgery is the gold standard for the assessment of meniscal abnormalities, one could argue that including only patients with operative correlation in our study provides the most accurate assessment of this MRI technique in evaluating menisci. An MRI interpretation of healthy or diseased menisci without surgical confirmation would simply be a theoretic negative or positive and not a true-negative or true-positive. In addition, because there were both healthy and diseased menisci found at surgery in our study, we were able to arrive at statistically justifiable sensitivity and specificity profiles. For instance, there were nine healthy medial menisci and 17 healthy lateral menisci found at surgery in our study, contributing to our statistical data. If the MRIs of patients who did not go to surgery were included in our study, without surgical confirmation we could not have truly assessed the accuracy of our results. Furthermore, the studies in the orthopedics literature assessing the accuracy of MRI generally only include patients with surgical correlation [6, 8, 13, 14].
An additional weakness in our study design was the use of consensus interpreting by three musculoskeletal radiologists. An alternative approach would have been to assess the accuracy of each individual interpreter with interobserver data available. This may have provided more clinically useful and practical information regarding radiologists' accuracies. However, because our goal was to assess the accuracy of a sequence design in lesion detection, we wanted to exclude the variability of radiologists' interpretive abilities and experience from the data analysis. By combining the expertise of all three radiologists in the study, we hoped to achieve a theoretic best-case scenario for the sequences evaluated and, thereby, eliminate interpretive inexperience by any one radiologist.
The tears missed on MRI in our study included a tear of the body of the lateral meniscus (Fig. 5), which was described as a "rim tear" on the surgical report. It is possible that a subtle truncation of the free edge of the meniscus seen on a single coronal T2-weighted image was difficult to detect even in retrospect. The second tear not detected on the initial MRI review involved the medial meniscus. In this case, the tear was not described in detail in the surgical report, but was called a tear involving the "posterior one-half" of the medial meniscus. Again, subtle blunting of the free edge on a single coronal T2-weighted image (Fig. 2B) and slight fraying of the free edge of the posterior horn on a single sagittal proton density-weighted image are shown (Fig. 2A). At best, meniscal free-edge fraying or possibly a small radial tear could be called in hindsight. Because the average length of time after the MRIs before surgery was 38 days, one could argue that this could explain any false-negatives and the occurrence of a tear after the MRI examination but before surgery. In the two false-negatives in this report, one patient waited 10 days until surgery and the second waited 28 days.
Two tears were seen on MRI, but not found during surgery. One tear had a vertical orientation extending to both superior and inferior articular surfaces of the anterior horn of the medial meniscus (Fig. 3). The other tear had a complex signal and appeared to extend to the superior articular surface of the anterior horn of the lateral meniscus (Fig. 6). Even on second look, these tears would likely be called prospectively on MRI.
The limitations of both MRI and arthroscopy have previously been implicated in the rates of false-positive and false-negative results. The importance of a high meniscal signal extending to an articular surface on more than one image has been proposed as the essential criterion for diagnosing a meniscal tear [19]. However, in that study, only 90% of menisci meeting this criterion showed a tear at arthroscopy, highlighting the potential for false-positives even in an accurately interpreted examination. One may argue that because we used the criterion of an abnormality extending to an articular surface seen on at least one image, we created the possibility for a high false-positive rate based on the De Smet et al. study [19]. However, both of our two false-positives were signal abnormalities extending to the surface on two or more consecutive images (Figs. 3 and 6), which would have qualified as tears based on the De Smet criterion.
Several confounding factors may have resulted in decreased accuracy of MRI in detecting meniscal tears, including the concomitant presence of anterior cruciate ligament (ACL) tears [20]. The two patients in our study in whom meniscal lesions were missed did not have ACL tears present. The anterior horns of the medial and lateral menisci have also been implicated in the elevation of false-positive rates [21]. Interestingly, both of our false-positive cases involved the anterior horns of medial and lateral menisci. In one study, an increased signal in the anterior horn of the lateral meniscus near its root was shown as a healthy finding [22]. Errors in MRI interpretation, the presence of healthy variants, and cases with subtle and equivocal findings have all been attributed as potential pitfalls [23]. However, the interpretations of the two false-negative and the two false-positive cases in our study would not have changed even on retrospective reassessment.
Several known surgical blind spots may have contributed to skewed results, one of which is the posterior meniscocapsular junction of the medial meniscus [24]. An incomplete arthroscopic examination may also have led to elevated false-positive rates [23, 25, 26]. In addition, one could argue that some of the tears visualized on higher-resolution images may be beyond the detection or diagnostic threshold of the arthroscopic examination. Surgical and radiologic terminology may also be implicated in discordant findings. For example, some orthopedists may use the term "fraying" and "tearing" interchangeably [25]. Therefore, because of the degree of subjectivity in describing subtle meniscal tears at arthroscopy, an abnormal meniscal contour may be called a "tear" by a radiologist but called "fraying" by the orthopedist, which would certainly alter statistical analysis. The converse would also be true.
In conclusion, high-resolution fast spin-echo 3-T MRI sequences are highly accurate and reliable in the detection of medial and lateral meniscal tears. Because of the inherent biases of a retrospective study, a prospective assessment of 3-T MRI using the fast spin-echo technique is warranted. Furthermore, even though the statistical values in our current study were within the upper range of those previously reported, a direct comparison between 3-T MRI and 1.5-T MRI images as part of a prospective study is needed to assess the theoretic improvement in diagnostic performance of 3-T MRI.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
