410
Injury (1992) 23, (6), 410-412
Printed in Great Britain
Magnetic resonance imaging of the knee: initial experience in a district general hospital M. P. Grevitt, C. J. F. Pool, R. N. Bodley and P. E. Savage Stoke Mandeville
Hospital, Aylesbury,
Bucks, UK
We present the results of a prospecfi,,e sfudy comparing fk accuracy of magnetic resonance imaging wifh arthroscopy in fk assessmenf of knee complaints. L&g a 1owfieM s frengfh tnagnef, MRl was able to achievea high diamiic accuracy within fk sefting of a district general hospital.
Introduction Since 1973, when Lauterbur first proposed that nuclear magnetic resonance might be used to generate images, magnetic resonance imaging (MRI) has been employed in the investigation of the soft tissue anatomy and pathology of various organs (Lauterbur, 1973). The provision of good tissue discrimination without the need for intra-articular contrast, multi-planar imaging capability and lack of exposure to ionizing radiation renders MRI particularly suitable for the non-invasive investigation of knee complaints. The advent of close fitting surface coils (thereby increasing the signal-to-noise ratio) has allowed direct visualization of menisci and ligaments. Published reports from North America and Australia have compared the accuracy of MRI with arthroscopy in the diagnosis of meniscal and ligamentous injuries. Accuracy in identifying tears of the meniscus has ranged from 64 to 95 per cent (Reicher et al., 1986; Crues et al., 1987, Stoller et al., 1987). The published experience with MRI of the musculoskeleta1 system in Great Britain has been limited. The aim of this study was to establish how quickly high diagnostic accuracy in assessing knee complaints could be achieved within the setting of a district general hospital, using a low field strength magnet.
Patients and methods Stoke Mandeville Hospital began operating a MRI scanner in July 1990. Between !%ptember 1990 and June 1991 55 consecutive patients scheduled to undergo arthroscopy of a knee had preoperative MRI scans of the affected joint. There were 38 males and 17 females; the average age was 36 years (range 17-65 years). The principal indications for arthroscopy were suspected meniscal tears, disruption of the cruciate ligaments or degenerative arthritis. All the scans were made with a 0.2 tesla permanent magnet (Hitachi MRP-20) using an extremity coil with a 16 0 1992 Butterworth-Heinemann 002&1383/92/06041(rO3
Ltd
cm diameter aperture, with the knee in slight flexion and the leg 15” externally rotated. Contiguous 5 mm sagittal and coronal sections were taken using a spin-echo sequence with relaxation times of 1500-1800ms, excitation times of 25 ms, and two averages (excitations) in a 22 cm field of view, to give T,-proton density anatomical images. The matrix size was 256 x 256 and the pixel resolution was 0.5 mm The scanning time was 11 min in each plane. Abnormal meniscal appearances were graded 1 to 3 after the method of Stoller (Stoller et al., 1987). Only a grade 3 signal, i.e. an intrameniscal linear signal intensity extending to an articular surface was classified as a men&al tear (Figure la,b). Other criteria for diagnosing a meniscal tear were disruption of the normal men&al contour with foreshortening or complete absence of any meniscal structure (Figzues 2, 3u,b). The anterior cruciate ligament was considered intact if a homogeneous dark signal could be traced in continuity through the long axis of the ligament from its origin to its insertion. The appearance of a high signal intensity in any portion of the ligament was interpreted as a disruption of the ligament. Comment was also made as to the state and integrity of the articular cartilage and extra-articular structures. All MRI scans were reported by two radiologists with little prior experience of MRI. No clinical or radiographic information was provided before their interpretation of the scans and their reports were collected on a standard proforma. Arthroscopies were performed on average 8 weeks after a MRI examination. No patient had a further injury in the intervening period. The results of the MRI scans were not known by the surgeon performing the arthroscopy. All the arthroscopies were performed or validated by the senior author (CJP). A tom meniscus was diagnosed if a cleavage in the cartilage produced a mechanical abnormality sufficient to allow a portion of the meniscus to be displaced from its normal position by the examining probe. Cruciate ligament tears were diagnosed if a disruption was seen at either of its attachments. Arthroscopic findings were recorded on a standard form that detailed the internal anatomy of the knee joint. The results of the MRI scans and arthroscopic examinations were collated by an independent observer (MPG). In order to facilitate the learning process of interpretation
Grevitt et al.: MRI of the knee
411
Table I. Results of magnetic resonance imaging comparedwith arthropscopy Results True
Medial meniscus Lateral meniscus Anterior cruciate ligament Chondromalacia syndromes
False
Positive
Negative
Positive
23
27 46 43 119
3 1 4 3
: 6
Negative 2 1 37
Sensitivity
Specificity
PPV
NPV
Accuracy
92 88 100 14
90 98 91 98
88 88 66 66
93 98 100 76
91 96 93 76
PPV, positive predictive value. NPV, negative predictive value.
of knee MRI, the arthroscopic findings were discussed with the radiologists after each group of 10 patients. Using arthroscopy as the ‘gold standard, sensitivity was calculated as the number of true-positive results divided by the number true-positive plus the number of false-negative results. Specificity was calculated as the number of truenegative results divided by the number of true-negative plus false-positive results. The positive predictive value was determined by the number of true-positive results divided by the the number of true-positive plus false-positive results. The negative predictive value was determined by the number of true-negative results divided by the number of true-negative plus false-negative results. Accuracy was defined as the number of true-positive results plus truenegative results divided by the total number of examinations.
Results (~ibk I). There were 25 tears of the medial meniscus found at arthroscopy of which 23 had been identified by MRI. The remainder of the medial menisci were judged arthroscopically normal. Of this latter group, three had been reported as complete tears by the radiologists. Thus, there were three false-positive and two false-negative interpretations. This produced a sensitivity of 92 per cent, specificity of 90 per cent, positive predictive value (PPV) 88 per cent, negative predictive value (NPV) 93 per cent and overall accuracy of 91 per cent. Nine tears of the lateral meniscus were identified by arthroscopy, of which eight had been correctly predicted by MRI. One of the normal lateral menisci had been interpreted on the MRI scan as a tear. This resulted in one false-positive and one false-negative test. The sensitivity was calculated as 88 per cent, specificity 98 per cent, PPV 88 per cent, NPV 93 per cent and accuracy 96 per cent. The overall values for the meniscal group combined were sensitivity 91 per cent, specificity 95 per cent, PPV 89 per cent, NPV 96 per cent and accuracy 94 per cent. With regard to tears of the anterior cruciate ligament, there were eight complete disruptions identified by arthroscopy and MRI. Four ligaments that were stable to manipulation with the examining probe at arthroscopy were interpreted as complete tears on MRI. The sensitivity of MRI for this group was 100 per cent, specificity 91 per cent, PPV 66 per cent, NPV 100 per cent and accuracy 93 per cent. No tears of the posterior cmciate ligament were found at arthroscopy and therefore no comparison was made with MRI.
When analysing the accuracy of MRI in detecting lesions of the articular cartilage, the retropatellar, femoral and tibia1 condylar surfaces were recorded separately. There were 43 articular defects noted at arthroscopy of which three were correctly identified by MRI. Three articular defects predicted by MRI were not subsequently confirmed at arthroscopy. Thus, there were three false-positive and 37 false-negative results. The calculated values for the test in this group were sensitivity 14 per cent, specificity 98 per cent, PPV 66 per cent, NPV 76 per cent and accuracy 76 per cent.
Discussion The majority of reports comparing the accuracy of MRI with knee arthroscopy have been from specialist centres. These have demonstrated that most intra-articular pathology can be well visualized by MRI. The earlier studies were performed using superconducting magnets ranging in strength from 0.35 to 1.5 tesla. In our study there were four false-positive and three false-negative results relating to men&al tears. These errors occurred in the first 20 patients in the series. The falsepositive tests were the result of over-interpretation of areas of chondromyxoid degeneration (grade 2). The falsenegative diagnoses were within the anterior horns of the menisci (and were not likely to be missed by the arthroscopist). In this region the transverse geniculate or meniscotibial ligaments may be diagnosed erroneously as tears (Watanabe et al., 1989). The four false-positive anterior cruciate tears may have arisen because of the partial volume effect producing a discontinuity in the MRI appearance of the ligament. Alternatively, the areas of high signal intensity may represent intraligamentous myxomatous degeneration. Although MRI in this study was not helpful in identifying the chondromalacia syndromes, this failure occurred in other series where more powerful magnets were employed (Polly et al., 1988; Barronian et al., 1989). This failure is a function of the scanning technique used; axial T, sections with use of gradient-echo fast scans (which produce greater contrast between synovial fluid and hyaline cartilage) may correct this deficiency. This study has demonstrated that high diagnostic accuracy can be achieved provided that regular audit of results is performed. Furthermore, the use of a low field strength magnet does not reduce the accuracy. By analogy, with the introduction of computed tomography the availability of cheaper MRI scanners (with lower running costs) is likely to increase. Orthopaedic surgeons may then have to
Injury: the British Journal of Accident Surgery (1992) Vol. 23/No. 6
412
decide how best to utilize this imaging technique. It is dificult to justify MRI on patients in whom the need for therapeutic arthroscopy is clear on clinical grounds, but we envisage several indications for its use: I. Equivocal or unusual symptoms. symptoms after a negative 3. The high surgical risk patient. 2. Persistent
arthroscopy.
Conclusion This study shows that with regular audit the ‘learning curve’ can be steepened appreciably and that relatively inexpensive MRI using a low field strength magnet in a district general hospital has a high accuracy rate.
Crues J. V., Mink J., Levy T. L. et al. (1987) Meniscal tears of the knee, accuracy of MR imaging. Radio& 164,445. Lauterbur P. C. (1973) Image formation by induced local- interactions: examples employing nuclear magnetic resonance. Nafure 242,190. Polly D. W, CaJlaghan J. J., Sikes R. A. et al. (1988) The accuracy of selective magnetic resonance imaging compared with the findings of arthroscopy of the knee. J. Bone]oint 50-g 7OA, 192. Reicher M. A., Hartzman S., Duckwiler G. R. et al. (1986) Meniscal injuries: detection using MR imaging. Radiology 159,753. Stoller D. W., Martin C., Crues J. V. etal. (1987) Meniscal tears, pathologic correlation with MR imaging. Radiology 163, 731. Watanabe A. T., Carter B. C., Tentelbaum G. P. et al. (1989) Common pitfalls in magnetic resonance imaging of the knee. J. BoneJoint Surg 7lA, 857.
Acknowledgements We wish to thank all the radiographers MRI examinations.
who performed
the Paper accepted 4 December
1991.
References Barronian A. D., Zoltan J. D., Bucon K. A. (1989) Magnetic resonance imaging of the knee: correlation with arthroscopy. Arthroscopy 71A, 187.
Requests for reprints shouM be adksed to: Mr M. Grevitt, 43 Aldershot Road, Fleet, Hampshire, GU13 9NT.
Injury (1992) 23, (6), 410-412
Printed in Great Britain
Magnetic resonance imaging of the knee: initial experience in a district general hospital M. P. Grevitt, C. J. F. Pool, R. N. Bodley and P. E. Savage Stoke Mandeville
Hospital, Aylesbury,
Bucks, UK
We present the results of a prospecfi,,e sfudy comparing fk accuracy of magnetic resonance imaging wifh arthroscopy in fk assessmenf of knee complaints. L&g a 1owfieM s frengfh tnagnef, MRl was able to achievea high diamiic accuracy within fk sefting of a district general hospital.
Introduction Since 1973, when Lauterbur first proposed that nuclear magnetic resonance might be used to generate images, magnetic resonance imaging (MRI) has been employed in the investigation of the soft tissue anatomy and pathology of various organs (Lauterbur, 1973). The provision of good tissue discrimination without the need for intra-articular contrast, multi-planar imaging capability and lack of exposure to ionizing radiation renders MRI particularly suitable for the non-invasive investigation of knee complaints. The advent of close fitting surface coils (thereby increasing the signal-to-noise ratio) has allowed direct visualization of menisci and ligaments. Published reports from North America and Australia have compared the accuracy of MRI with arthroscopy in the diagnosis of meniscal and ligamentous injuries. Accuracy in identifying tears of the meniscus has ranged from 64 to 95 per cent (Reicher et al., 1986; Crues et al., 1987, Stoller et al., 1987). The published experience with MRI of the musculoskeleta1 system in Great Britain has been limited. The aim of this study was to establish how quickly high diagnostic accuracy in assessing knee complaints could be achieved within the setting of a district general hospital, using a low field strength magnet.
Patients and methods Stoke Mandeville Hospital began operating a MRI scanner in July 1990. Between !%ptember 1990 and June 1991 55 consecutive patients scheduled to undergo arthroscopy of a knee had preoperative MRI scans of the affected joint. There were 38 males and 17 females; the average age was 36 years (range 17-65 years). The principal indications for arthroscopy were suspected meniscal tears, disruption of the cruciate ligaments or degenerative arthritis. All the scans were made with a 0.2 tesla permanent magnet (Hitachi MRP-20) using an extremity coil with a 16 0 1992 Butterworth-Heinemann 002&1383/92/06041(rO3
Ltd
cm diameter aperture, with the knee in slight flexion and the leg 15” externally rotated. Contiguous 5 mm sagittal and coronal sections were taken using a spin-echo sequence with relaxation times of 1500-1800ms, excitation times of 25 ms, and two averages (excitations) in a 22 cm field of view, to give T,-proton density anatomical images. The matrix size was 256 x 256 and the pixel resolution was 0.5 mm The scanning time was 11 min in each plane. Abnormal meniscal appearances were graded 1 to 3 after the method of Stoller (Stoller et al., 1987). Only a grade 3 signal, i.e. an intrameniscal linear signal intensity extending to an articular surface was classified as a men&al tear (Figure la,b). Other criteria for diagnosing a meniscal tear were disruption of the normal men&al contour with foreshortening or complete absence of any meniscal structure (Figzues 2, 3u,b). The anterior cruciate ligament was considered intact if a homogeneous dark signal could be traced in continuity through the long axis of the ligament from its origin to its insertion. The appearance of a high signal intensity in any portion of the ligament was interpreted as a disruption of the ligament. Comment was also made as to the state and integrity of the articular cartilage and extra-articular structures. All MRI scans were reported by two radiologists with little prior experience of MRI. No clinical or radiographic information was provided before their interpretation of the scans and their reports were collected on a standard proforma. Arthroscopies were performed on average 8 weeks after a MRI examination. No patient had a further injury in the intervening period. The results of the MRI scans were not known by the surgeon performing the arthroscopy. All the arthroscopies were performed or validated by the senior author (CJP). A tom meniscus was diagnosed if a cleavage in the cartilage produced a mechanical abnormality sufficient to allow a portion of the meniscus to be displaced from its normal position by the examining probe. Cruciate ligament tears were diagnosed if a disruption was seen at either of its attachments. Arthroscopic findings were recorded on a standard form that detailed the internal anatomy of the knee joint. The results of the MRI scans and arthroscopic examinations were collated by an independent observer (MPG). In order to facilitate the learning process of interpretation
Grevitt et al.: MRI of the knee
411
Table I. Results of magnetic resonance imaging comparedwith arthropscopy Results True
Medial meniscus Lateral meniscus Anterior cruciate ligament Chondromalacia syndromes
False
Positive
Negative
Positive
23
27 46 43 119
3 1 4 3
: 6
Negative 2 1 37
Sensitivity
Specificity
PPV
NPV
Accuracy
92 88 100 14
90 98 91 98
88 88 66 66
93 98 100 76
91 96 93 76
PPV, positive predictive value. NPV, negative predictive value.
of knee MRI, the arthroscopic findings were discussed with the radiologists after each group of 10 patients. Using arthroscopy as the ‘gold standard, sensitivity was calculated as the number of true-positive results divided by the number true-positive plus the number of false-negative results. Specificity was calculated as the number of truenegative results divided by the number of true-negative plus false-positive results. The positive predictive value was determined by the number of true-positive results divided by the the number of true-positive plus false-positive results. The negative predictive value was determined by the number of true-negative results divided by the number of true-negative plus false-negative results. Accuracy was defined as the number of true-positive results plus truenegative results divided by the total number of examinations.
Results (~ibk I). There were 25 tears of the medial meniscus found at arthroscopy of which 23 had been identified by MRI. The remainder of the medial menisci were judged arthroscopically normal. Of this latter group, three had been reported as complete tears by the radiologists. Thus, there were three false-positive and two false-negative interpretations. This produced a sensitivity of 92 per cent, specificity of 90 per cent, positive predictive value (PPV) 88 per cent, negative predictive value (NPV) 93 per cent and overall accuracy of 91 per cent. Nine tears of the lateral meniscus were identified by arthroscopy, of which eight had been correctly predicted by MRI. One of the normal lateral menisci had been interpreted on the MRI scan as a tear. This resulted in one false-positive and one false-negative test. The sensitivity was calculated as 88 per cent, specificity 98 per cent, PPV 88 per cent, NPV 93 per cent and accuracy 96 per cent. The overall values for the meniscal group combined were sensitivity 91 per cent, specificity 95 per cent, PPV 89 per cent, NPV 96 per cent and accuracy 94 per cent. With regard to tears of the anterior cruciate ligament, there were eight complete disruptions identified by arthroscopy and MRI. Four ligaments that were stable to manipulation with the examining probe at arthroscopy were interpreted as complete tears on MRI. The sensitivity of MRI for this group was 100 per cent, specificity 91 per cent, PPV 66 per cent, NPV 100 per cent and accuracy 93 per cent. No tears of the posterior cmciate ligament were found at arthroscopy and therefore no comparison was made with MRI.
When analysing the accuracy of MRI in detecting lesions of the articular cartilage, the retropatellar, femoral and tibia1 condylar surfaces were recorded separately. There were 43 articular defects noted at arthroscopy of which three were correctly identified by MRI. Three articular defects predicted by MRI were not subsequently confirmed at arthroscopy. Thus, there were three false-positive and 37 false-negative results. The calculated values for the test in this group were sensitivity 14 per cent, specificity 98 per cent, PPV 66 per cent, NPV 76 per cent and accuracy 76 per cent.
Discussion The majority of reports comparing the accuracy of MRI with knee arthroscopy have been from specialist centres. These have demonstrated that most intra-articular pathology can be well visualized by MRI. The earlier studies were performed using superconducting magnets ranging in strength from 0.35 to 1.5 tesla. In our study there were four false-positive and three false-negative results relating to men&al tears. These errors occurred in the first 20 patients in the series. The falsepositive tests were the result of over-interpretation of areas of chondromyxoid degeneration (grade 2). The falsenegative diagnoses were within the anterior horns of the menisci (and were not likely to be missed by the arthroscopist). In this region the transverse geniculate or meniscotibial ligaments may be diagnosed erroneously as tears (Watanabe et al., 1989). The four false-positive anterior cruciate tears may have arisen because of the partial volume effect producing a discontinuity in the MRI appearance of the ligament. Alternatively, the areas of high signal intensity may represent intraligamentous myxomatous degeneration. Although MRI in this study was not helpful in identifying the chondromalacia syndromes, this failure occurred in other series where more powerful magnets were employed (Polly et al., 1988; Barronian et al., 1989). This failure is a function of the scanning technique used; axial T, sections with use of gradient-echo fast scans (which produce greater contrast between synovial fluid and hyaline cartilage) may correct this deficiency. This study has demonstrated that high diagnostic accuracy can be achieved provided that regular audit of results is performed. Furthermore, the use of a low field strength magnet does not reduce the accuracy. By analogy, with the introduction of computed tomography the availability of cheaper MRI scanners (with lower running costs) is likely to increase. Orthopaedic surgeons may then have to
Injury: the British Journal of Accident Surgery (1992) Vol. 23/No. 6
412
decide how best to utilize this imaging technique. It is dificult to justify MRI on patients in whom the need for therapeutic arthroscopy is clear on clinical grounds, but we envisage several indications for its use: I. Equivocal or unusual symptoms. symptoms after a negative 3. The high surgical risk patient. 2. Persistent
arthroscopy.
Conclusion This study shows that with regular audit the ‘learning curve’ can be steepened appreciably and that relatively inexpensive MRI using a low field strength magnet in a district general hospital has a high accuracy rate.
Crues J. V., Mink J., Levy T. L. et al. (1987) Meniscal tears of the knee, accuracy of MR imaging. Radio& 164,445. Lauterbur P. C. (1973) Image formation by induced local- interactions: examples employing nuclear magnetic resonance. Nafure 242,190. Polly D. W, CaJlaghan J. J., Sikes R. A. et al. (1988) The accuracy of selective magnetic resonance imaging compared with the findings of arthroscopy of the knee. J. Bone]oint 50-g 7OA, 192. Reicher M. A., Hartzman S., Duckwiler G. R. et al. (1986) Meniscal injuries: detection using MR imaging. Radiology 159,753. Stoller D. W., Martin C., Crues J. V. etal. (1987) Meniscal tears, pathologic correlation with MR imaging. Radiology 163, 731. Watanabe A. T., Carter B. C., Tentelbaum G. P. et al. (1989) Common pitfalls in magnetic resonance imaging of the knee. J. BoneJoint Surg 7lA, 857.
Acknowledgements We wish to thank all the radiographers MRI examinations.
who performed
the Paper accepted 4 December
1991.
References Barronian A. D., Zoltan J. D., Bucon K. A. (1989) Magnetic resonance imaging of the knee: correlation with arthroscopy. Arthroscopy 71A, 187.
Requests for reprints shouM be adksed to: Mr M. Grevitt, 43 Aldershot Road, Fleet, Hampshire, GU13 9NT.
