The authors measured the sensitivity, specificity, and accuracy of mid-field magnetic resonance imaging (MRI) evaluations of the knee, compared to actual pathology as observed by arthroscopy.
As the use of magnetic resonance imaging (MRI) is increasingly
diverse in diagnosing a variety of injuries, the value of studies
that analyze the clinical efficacy and further define the role of
MRl increases. The purpose of this research was to measure the
sensitivity, specificity, and accuracy of mid-field MRI evaluations
of the knee, compared to actual pathology observed by arthroscopy.
Seventy-four knees were evaluated, and arthroscopy was taken to be
the true diagnosis of the menisci and the anterior cruciate
ligament (ACL). Meniscal topography also was included secondarily
in the analysis of MRI diagnostic validity. This study showed good
overall validity with the use of mid-field MRI to diagnose knee
structures. Accuracy was highest in the detection of ACL tears
(96%) and lowest in finding medial meniscal tears (89%).
Sensitivity was lower in the lateral meniscus compared to the
medial meniscus (69% vs 94%, respectively), and lower specificity
occurred with the medial meniscus compared to the lateral meniscus
(88% vs 94%, respectively). MRI can be a valuable adjunct in
medical diagnoses, and the use of MRI will continue to evolve as an
important part of surgical planning.
Since the discovery of the use of nuclear magnetic resonance in
imaging by Lauterbur in 1973,1 magnetic resonance imaging (MRI) has
come to be accepted as a noninvasive modality to examine the
anatomy and pathology of various organs2,3 and soft tissues as well
as the musculoskeletal system.4,5 Other authors have reported on
the specific usefulness of MRI.6-10
The use of MRI offers many advantages, which include excellent
inherent soft-tissue contrast, multiplanar imaging ability, lack of
exposure to ionizing radiation, no need for intraarticular contrast
material, and the absence of artifact caused by the superimposition
of osseous structures. MRI also is painless, noninvasive, and
easily and rapidly performed.
Studies have been published that involve comparisons of the
findings of MRI with those of arthroscopy. In series ranging in
size from 25 to 1,014 patients,11-22 the accuracy of meniscal
imaging ranged from 45% to 98% and the accuracy of imaging the
anterior cruciate ligament (ACL) ranged from 92% to 100%.
The purpose of this study was to evaluate the sensitivity,
specificity, and accuracy of our local experience using mid-field
MRI evaluations of the knee, compared to actual pathology as
observed by arthroscopy.
Methods and materials
Seventy-four knees in 72 patients were evaluated with MRI
followed by arthroscopy. Data collected from a period of February
1991 to July 1994 were evaluated by an independent researcher. The
patient series represented a random sampling of the population, who
would be evaluated clinically for suspected acute or chronic
lesions of the meniscus and/or ACL. There were 40 male and 32
female patients involved in the study, and the average age was 34.7
years (standard deviation = 11.2 years), with a range of 13 to 65
years. The series included 41 right knees and 33 left knees.
Arthroscopy was performed from 1 day to 13 weeks after MRI
evaluation and was carried out under general anesthesia. Distention
of the joint cavity was achieved with an electrolyte-free solution
(saline). Thirty-degree rod-lens arthroscope and a probe were
introduced using anterolateral and anteromedial portals. All
arthroscopic findings and remedial procedures were recorded on
videotape. Only one arthroscopist performed all surgeries
MRI images were obtained from a 0.6T (Tesla) Health Images (HI)
standard magnet. A dedicated full-volume (360°) knee surface coil
was utilized. The patient's knee within the surface coil usually
was in approximately 10° of flexion and 10° to 15° of external
rotation, based primarily on the patient's comfort level.
The routine knee protocol included a T1-weighted transaxial and
coronal spin-echo sequence (TR 600-650; TE 30) with a field of view
(FOV) of 16,21 cm or 17,21 cm; YRES 160; 2 NSQ; 5-mm-thick slices
with a 20% gap. Spin-echo proton-density and T2-weighted sagittal
images (TR 2000; TE 30/90); FOV 16,21 cm or 17,21 cm; YRES 184; 2
NSQ; with 4-mm-thick slices and a 25% gap, also were performed. At
the technologist's discretion, if a normal ACL was not seen on the
routine sagittal images, an additional gradient-echo sagittal
sequence angled through the region of the ACL was performed with a
TR 440, TE 24 and a flip angle of 13°. The latter sequence utilized
an FOV of 16,21 cm or 17,21 cm; YRES 160; and a 3-mm slice
thickness with a 20% gap.
Arthroscopy was taken to be the true diagnosis to evaluate the
sensitivity, specificity, and accuracy of MRI findings. Sensitivity
was calculated using the number of true-positive results divided by
the sum of the true-positive and false-negative results.
Specificity was calculated using the number of true-negative
results divided by the sum of the true-negative results and
false-positive results. Accuracy was determined using the total
number of patients minus the sum of false-positive and
false-negative MRI results divided by the total number of
Predictive values also were calculated. The negative predictive
values were calculated by dividing the number of true-negative
results by the sum of true-negative results plus false-negative
results. Positive predictive values were determined by the number
of true-positive results divided by the sum of true-positive
results plus false-positive results.
MR images were interpreted primarily by one radiologist (SMS)
experienced in MR radiology. The criteria used to determine
meniscal tears were as follows.
* Criteria involved abnormal increase of intrameniscal signal
(best demonstrated on the proton-density sagittal sequences and the
next T1-weighted coronal sequence), which extended confidently into
an articular surface (again designated as a grade III tear).
* Abnormal increase of intrameniscal signal presenting as a
global (grade I) or linear (grade II) configuration without
definite extension into an articular surface was considered to
represent internal degeneration rather than a frank tear (figure
- Morphologic distortion, blunting, or truncation of the
meniscus also was used as a criterion in determining the presence
of a tear, often associated with the abnormal intrameniscal
signal described previously.
The ACL was considered present when the entire ligament could be
identified. Contiguous sagittal images were reviewed to verify the
intact length of the ACL. The ligament was considered torn when the
entire length could not be identified (figure 3).
Comparisons of the MR diagnoses and arthroscopy diagnoses are
summarized in tables 1 and 2. For the medial meniscus, 31
true-positive and 43 true-negative tears were revealed by
arthroscopy. Of the 74 patients, MR diagnoses resulted in 6
false-positive and 2 false-negative images (table 1). Accuracy was
89%; sensitivity was 94%; specificity was 88%; negative predictive
value (NPV) was 96%; and positive predictive value (PPV) was 84%
With regard to the lateral meniscus, arthroscopy identified 11
true-positive and 63 true-negative tears (table 1). Comparisons of
the findings of MRI with arthroscopy revealed one false-positive
and five false-negative images. Accuracy was 92%; sensitivity was
69%; specificity was 94%; NPV was 93%; and PPV was 92%, as shown in
For the ACL in table 1, there were 15 true-positive, 59
true-negative, 1 false-positive, and 2 false-negative results.
Accuracy, sensitivity, and specificity were 96%, 88%, and 98%,
respectively. NPV was 97%, and PPV was 94% (table 2).
Meniscal topography was included in the analysis of the MRI
diagnostic validity (figure 1). When each of the three segments of
the menisci were evaluated individually, the anterior aspect of the
medial and lateral meniscus had the lowest sensitivity values (67%
for both), compared to other segments. Also, somewhat low
sensitivity resulted in the middle and posterior aspects of the
lateral meniscus (78% and 71%, respectively). Overall, MRI results
were generally accurate in all meniscal components with very good
specificity (figure 1).
Diagnoses other than meniscal and ACL included Baker's and
popliteal cysts, articular cartilage and osteochondral defects,
bone tumors, meniscal cysts, extensor and patellar tracking
problems, subchondral sclerosis, tibial plateau fractures, patellar
tendinitis, and collateral ligament sprains and tears.
This study resulted in generally good validity in the use of MRI
to diagnose deranged knee structures. Accuracy was highest in
detecting ACL tears (96%) and lowest in detecting medial meniscus
(89%), as shown in table 2. Sensitivity was lower in the lateral
meniscus compared to the medial meniscus (69% vs 94%,
respectively), and lower specificity occurred with the medial
meniscus compared to the lateral meniscus (88% vs 94%,
Reduced specificity in MRI evaluations of the medial meniscus
appears to be a common tendency and has been reported in other
studies.12,18-21,23 It has been suggested that because medial
meniscal tears occur more often than lateral meniscal tears,
radiologists tend to expect medial meniscus tears and consequently
"over-read" images of the medial menisci. Conversely, lateral
meniscus images are "under-read" as having no tears.12 In this
series, table 1 shows that arthroscopy revealed almost three times
as many tears of the medial meniscus compared to the lateral
meniscus, with nearly orthogonal results in false-positive and
false-negative diagnoses of the menisci with MRI.
We considered the possibility of other factors that could have
acted as potential confounders in over- or under-reading meniscal
MR images. Factors such as the amount of effusion, grade or
location of chondromalacia, chronicity, accompanying bone injury,
or previous injury were reviewed and compared in patients with
false-positive and negative results versus patients with
true-positive and negative results. No specific or apparent trends
in the above factors were present.
Our algorithm for the use of MRI in the management of traumatic
knee injuries is similar to others that have been described.13 In
our clinic, a history and an objective and subjective clinical
evaluation are used to determine the functional status of the ACL.
MRI is an essential tool in a difficult-to-examine knee due to
pain, swelling, or muscle spasms; these complications occur in
about 20% of our patients. If the ACL is deficient with significant
laxity and the patient is a candidate for surgery based on age and
activity level, MRI can be used to determine the structural
integrity of the ACL and, consequently, augment a surgical
decision. In patients who present with painful joint range of
motion but minor joint instability, MRI can be useful in
determining the condition of the menisci. If the patient is not a
candidate for ACL surgical intervention and conservative,
nonoperative treatment is elected, MRI can define meniscal
pathology and possibly reduce the need for diagnostic
MRI is especially indicated when the clinician is doubtful
regarding the diagnosis and treatment or when an expeditious
diagnosis is required. Athletes and select self-employed
individuals particularly desire an immediate diagnosis and
typically are anxious to begin treatment and rehabilitation after
an injury. Those patients' situations are such that prolonged time
off could hinder their level of performance and/or place them in
Excellent results using MRI evaluations to help guide surgical
intervention decisions have been reported. In a series of 45
athletic patients, surgical treatment of all MRI grade II lesions
of the menisci resulted in the elimination of symptoms in almost
all patients in the series who previously had severe complaints and
disability.24 Although signal grade was not part of this
investigation, it appears to be another potentially useful
component in successful clinical outcomes.
The economic efficacy of diagnostic tests continues to emerge as
an issue in health care and medical management. We are challenged
to provide the highest level of care yet keep costs to a minimum.
The results of our local series reveal very good validity and
reliability using MRI to diagnose knee injuries. Total overall
combined accuracy, PPV, and NPV (calculated for the medial and
lateral meniscus and ACL) were 92%, 90%, and 95%, respectively;
these are very high validity rates compared to similar studies.
MRI has been instrumental in preventing unnecessary surgery when
there has been a reasonable clinical doubt and, in some cases, the
use of MRI has been important in surgical planning. MRI can be a
useful tool in the decision-making process to treat knee injuries
successfully and can help direct the efficient use of diagnostic
1. Lauterbur PC: Image formation by induced local interactions:
Examples employing nuclear magnetic resonances. Nature 242:190-191,
2. Hendee WR, Morgan CJ: Magnetic resonance imaging, part 2:
Clinical applications. West J Med 141:638-648, 1984.
3. Fitzgerald RH, Berquist TH: Magnetic resonance imaging. J
Bone Joint Surg 68A:799-801, 1986.
4. Kean DM, Worthington BS, Preston BJ, et al: Role of NMR in
the evaluation of musculoskeletal disease. Appl Radiol
5. Moon KL, Genant HK, Helms CA, et al: Musculoskeletal
applications of nuclear magnetic resonance. Radiology 1147:161-171,
6. Kean DM, Worthington BS, Preston BJ, et al: NMR imaging of
the knees: Example of normal anatomy and pathology. Br J Radiol
7. Reicher MA, Rauschning W, Gold RH, et al: High-resolution
magnetic resonance imaging of the knee joint: Normal anatomy. Am J
Radiol 145:895-902, 1985.
8. Gallimore GW, Herms SE: Knee injuries: High-resolution MR
imaging. Radiology 160:457-461, 1986.
9. Stoller DW, Martin C, Crues JV III, et al: Meniscal tears:
Pathologic correlation with MR imaging. Radiology 163:731,
10. Crues JV III, Ryu R, Morgan F: Meniscal pathology: The
expanding role of magnetic resonance imaging. Clin Orthop Rel Res
11. Lee JK, Yao L, Phelps CT, et al: Anterior cruciate tears: MR
imaging compared with arthroscopy and clinical tests. Radiology
12. Fischer SP, Fox JM, Del Pizzo W, et al: Accuracy of
diagnosis from magnetic resonance imaging of the knee. J Bone Joint
Surg 73A(1):2-10, 1991.
13. Mandelbaum BR, Finerman GA, Reicher MA, et al: Magnetic
resonance imaging as a tool for evaluation of traumatic knee
injuries: Anatomical and pathoanatomical correlations. Am J Sports
Med 14(5):361-370, 1986.
14. Polly DW, Callaghan JJ, Sikes RL, et al:
The accuracy of selective magnetic resonance imaging compared
with the findings of arthroscopy of the knee. J Bone Joint Surg
15. Silva I, Silver DM: Tears of the meniscus as revealed by
magnetic resonance imaging. J Bone Joint Surg 70A(2):199-202,
16. Reicher MA, Hartzman S, Duckwiler GR, et al: Meniscal
injuries: Detection using MR imaging. Radiology 159(3):753-759,
17. Crues JV III, Mink V, Levy TL, et al: Meniscal tears of the
knee: Accuracy of MR imaging. Radiology 160(3):445-448, 1987.
18. Raunest J, Oberle K, Loehnert J, et al: The clinical value
of magnetic resonance imaging in the evaluation of meniscal
disorders. J Bone Joint Surg 73A(1):11-16, 1991.
19. Jackson DW, Jennings LD, Maywood RM, et al: Magnetic
resonance imaging of the knee. Am J Sports Med 16(1):29-37,
20. Ng J, Baron M, Ng AC, et al: Traumatic knee injuries: The
accuracy of MRI compared with arthroscopy. Indiana Med 11:886-890,
21. Boeree NR, Watkinson AF, Ackroyd CE, et al: Magnetic
resonance imaging of meniscal and cruciate injuries of the knee. J
Bone Joint Surg 73B(3):452-457, 1991.
22. Sanchis-Alfonso V, Martinez-SanJuan V, Gastaldi-Orquin Z:
The value of MRI in the evaluation of the ACL-deficient knee in the
postoperative evaluation after ACL reconstruction. Europ J Radiol
23. Barrion AD, Zoltan JD, Bucon KA: Magnetic resonance imaging
of the knee: Correlation with arthroscopy. J Arthroscop Rel Surg
24. Biedert RM: Intrasubstance meniscal tears: Clinical aspects
of the role of MRI. Arch Orthop Trauma Surg 112:142-147, 1993.