Osteoarthritis and Cartilage (1996) 4, 181-186 1996 Osteoarthritis Research Society 1063-4584/96/030181 + 06 $12.00/0 OSTEOARTHRITIS and CARTILAGE Cartilage thickness measurement in magnetic resonance imaging BERND KLADNY*, HERMANN BAIL*, BERND SWOBODA*, HEINZ SCHIWY-BOCHATt, WOLFGANG F. BEYER* AND GERD WEsELOH* Division of Orthopaedic Rheumatology, Department of Orthopaedics, University of Erlangen, Rathsberger Str. 57, 91054, Erlangen, and/'department of Forensic Medicine, University of Technology, Aachen, Germany Summary T0 assess the accuracy of cartilage thickness measurements in magnetic resonance imaging (MRI), we compared data ::o]~tained by cartilage thickness measurements in MRI with corresponding histological sections of 14 human proximal tibial articular surfaces. Each proximal tibial articular surface was cut into five medial and lateral slices and each of these slices was divided into three sectors providing 420 sectors, 406 of which were evaluated in our study. The overall correlation coefficient (r) was 0.96. Topographical differences were found. The lowest correlation coefficient in our series was observed in the anterior part of the medial tibial plateau (r = 0.88). Cartilage thickness measurements in MRI were more accurate in cartilage thicker than 2 mm (r = 0.94) than in thinner cartilage layers (r = 0.73). There were no significant differences in cartilage thickness measurements in different grades of osteoarthritis. However, the mean percentage difference between cartilage thickness in MRI and histology was about 10% in our series. Key words: Magnetic resonance imaging (MRI), Cartilage thickness, Knee joint, Osteoarthritis. Introduction ACCURATE DATA on the thickness of articular cartilage with imaging techniques are difficult to obtain in vivo. Techniques to assess the cartilage thickness by arthrography and X-ray measurements have been described [1, 2, 3]. Magnetic resonance imaging (MRI), with its high soft-tissue contrast, has the potential to detect the cartilage layer; it is non-invasive and lacks ionizing radiation. The use of specially designed surface coils and high field imagers gives high resolution of the investigated structures. With the help of three-dimensional gradient-echo sequences, thin contiguous slices in any plane can be obtained. Various pulse sequences have been used to assess articular cartilage, but there has not been any clear consensus on which technique is best [4]. The three-dimensional gradient-echo sequence FISP (fast imaging with steady-state precession) has proved to be very effective in the evaluation of knee disorders [5, 6] and in the visual representation of cartilage [7]. The following study presents our results comparing cartilage thickness measurements in MRI with Submitted 10 February 1995; accepted 22 January 1996. Address correspondence to Bernd Kladny. Funds were received in support of the research from the Deutsche Forschungsgemeinschaft (KI 871/1-1). corresponding histological sections in human tibial plateaus. Methods Fourteen proximal tibial articular surfaces were removed at autopsy. There were no specific selection criteria. The soft tissues were removed, with the exception of the menisci, and the proximal tibial articular surfaces were freshly frozen (-20 Celsius) after preparation, then thawed for the MRI investigation. We used a high field imaging system (1.5 T Magnetom SIEMENS~). The tibial plateaus were positioned in the center of a surface coil used routinely for MRI of the knee joint. The positioning was that usually employed for knee MRI. The tuberositas tibiae was placed anteriorly. The rotation was controlled by laying the posterior aspects of the tibial condyles parallel to the underlying surface. The articular surface was at right angle to the underlying surface The first measurement was a short Tl-weighted sequence in axial slice orientation to control the correct positioning of the rotation. The axial scout view was printed and used as a guide for positioning the proximal tibial articular surface for cutting. Then we used a three-dimensional gradient-echo sequence (FISP, TR = 30 ms, TE = 12 ms, field of view 181
182 Kladny et al.: Cartilage thickness measurement in MRI 170 170mm, flip angle 40 ) in coronal slice orientation with an effective slice thickness of 1.1ram. Each tibial plateau was cut into five anatomically defined coronal slices with a special sawing device (EXAKT ) which allowed precise cuts of the bone (Fig. 1). To control the rotation and the tilt of the proximal tibia, the proximal tibial end was positioned in the sawing device as previously described for the positioning in the surface coil. The cutting planes were parallel to the posterior part of the tibial condyles. The first cut was made through the anterior part of the menisci, the second plane was defined by the posterior margin of the anterior horn of the menisci at the transition to the uncovered cartilage surface. The third cutting plane was in the center of the proximal tibial articular surface. The fourth and fifth cutting planes were defined through the posterior horn of the menisci, similar to those used for the first and second planes. The band-saw had a 0.2 mm thick blade with a diamond surface. This allowed minimum tissue loss during the sectioning process. The thickness of each bone block was measured. Corresponding magnetic resonance images were selected by corresponding distances between the planes. In addition, the shape of the corresponding slices (MRI and histological section) were matched. Controlling the rotation and geometrical measurements as described previously ensured that the histological section was within the corresponding MRI slice of 1.1 mm. The bone blocks were embedded in methyl methacrylate [8]. From the embedded block a 6pm histological section was cut off with a microtom (Polycut; REICHART-JUNG ~) The section was then stained with safranin-o. The medial and lateral tibial plateaus were divided into three sectors of.the same size in MRI and the histological sections (Fig. 2). The cartilage thickness was measured in the middle of each sector. The MR images were shown on the monitor screen at twice magnification. Cartilage thickness was measured by measuring distances on the monitor screen in pixels. The measurements were made at a workstation (SIEMENS ) connected to the MR unit (1.5 T Magnetom, SIEMENS ) and the MRI data sets were stored on an optical disc. In MRI, cartilage was defined as an area of intermediate signal intensity. A video camera (K 30, SIEMENS ) was connected to a microscope with a 1.25 magnification lens (Planachromat, ZEISS ). The histological sections were inducted into an image analysing system (KONTRON ), where the cartilage thickness was measured interactively with a mouse board by determining the extension of the cartilage layer. Cartilage thickness was measured by two different observers in MRI and in histology. In addition, the grade of osteoarthritis of each sector was determined according to surface irregularities in histological preparations: Grade 0=normal cartilage, grade 1 = slight fibrillations, grade 2 = severe fibrillations not exceeding 50% of the cartilage layer, grade 3 = severe cartilage fibrillations exceeding 50% of the cartilage layer, and grade 4=full-thickness cartilage defect. The relation between MRI and anatomic measurements was determined by using the Pearson correlation coefficient. Yll ~I2 VI3 VI4 vi5 Fro. 1. Sketched view on the tibial head, which demonstrates the coronal cutting planes.
Osteoarthritis and Cartilage Vol. 4 No. 3 183 L1 L2 \ \ M2 L3 M3 L4 L5 C M4 M5 \ Fro. 2. Each medial and lateral tibial plateau was divided in three sectors of equal size. Results To obtain data on the reliability of our methods we made 50 random measurements. The mean difference of intraindividual measurements in MRI was 14.3% (mean 0.39 mm), with the image analysing system the intraindividual difference was 1.4% (mean 0.03mm). The values for the interindividual differences were 14.8% in MRI (mean 0.39 mm) and 1.7% with the image analysing system (mean 0.03 ram). Of the 420 sectors evaluated, nine had to be excluded because of artefacts in MRI or in the histological section and all five full-thickness cartilage defects were excluded. Normal cartilage represented nearly 70% of all locations (N=287) and grade 1 of osteoarthritis was found in 18. 5% (N= 77). There were only a few cases with grade 2 (N=33) and grade 3 (N=9) lesions. The cartilage thickness in all sectors ranged from 0.9-6.6 mm. The mean cartilage thickness of all sectors in histological sections was 2.58 mm (S.D. 1.01 mm). The overall correlation of cartilage thickness measured in MRI and histological sections was 0.96. The t-test for paired samples showed no differences for the matched sets (P < 0.01). The absolute difference ranged from 0.8-1.7 mm. The mean difference was 0.12 mm (S.D. 0.28 mm). The mean percentage difference between the two values was 8.4% and ranged from 0~58%. There was a tendency to have lower values in MRI (N= 242); the mean percentage difference for these cases was 10.5%. In 89 cases the value of cartilage thickness in MRI was higher than that measured in the histological sections. In these cases the mean percentage difference between the two values was 9.8%. In 75 cases cartilage thickness in MRI and histology reached the same value. There were topographical differences. The lowest correlation was found in the anterior part of the medial tibial plateau. In the other areas the correlation coefficient was >0.90 (Table I). The reliability of cartilage thickness measurements depended on the absolute thickness of the cartilage layer. We found a correlation of 0.73 for an absolute cartilage thickness less than 2 mm (IV= 142), while the mean percentage difference in MRI and histology was 9.3%. There was a clear increase in the correlation coefficient to 0.94 (N= 264) with an absolute cartilage thickness greater than 2 ram, while the mean percentage difference was 7.9% in these cases. There were no significant differences between the measurements of cartilage thickness in histological sections and those in MRI depending on the grade of osteoarthritis in corresponding sectors (Table II). Discussion In former studies that compared MRI with findings in arthroscopy, the authors emphasized the ability of MRI to represent the cartilage layer [4, 5, 9, 10]. In vitro studies showed the potential to detect articular cartilage defects in the range of 2-3 mm in MRI [11]. For the detection of cartilage
184 K l a d n y et al.: Cartilage t h i c k n e s s m e a s u r e m e n t in MRI defects and measurements of cartilage thickness, three-dimensional fast imaging sequences are superior to two-dimensional sequences [12]. There is no clear consensus about which, if any, pulse sequence is the best to assess articular cartilage [4]. The FISP sequence used in our study has good spatial resolution which enables the detection of small defects; it gives good subject and image contrast and enables a reliable distinction of cartilage from adjacent anatomical structures. The sequence meets the conditions specified by Hayes and Conway for accurate MR evaluation of a r t i c u l a r cartilage [13]. The purpose of this study was not to find an optimum pulse sequence for cartilage but, r a t h e r,, to evaluate a sequence usually employed to cover all types of intra-articular lesions (especially menisci and cruciate ligaments). Studies with animal models suggest t h a t a loss of proteoglycan in cartilage degeneration can be detected by MRI [14] and t h a t MR relaxation times and proton density values vary with the severity of osteoarthritis [15]. We observed signal changes within the cartilage layer, even t h o u g h it was not FIa. 3. (a) Gross anatomic section of proximal tibial articular surface (plane 3), normal cartilage in the lateral compartment (left side), slight fibrillations in the medial compartment (right side), only to be seen by magnification. (b) Corresponding histological section. (c) Corresponding magnetic resonance image.
Osteoarthritis and Cartilage Vol. 4 No. 3 185 Table I. Correlation coefficient of cartilage thickness in the magnetic resonar~ce image and histological sections depending on the topography Correlation coefficient Medial tibial Lateral tibial Plane plateau plateau Plane 1 0.94 (N= 39) 0.93 (N=42) Plane 2 0.93 (N= 42) 0.95 (N= 41) Plane 3 0.94 (N= 41) 0.98 (N= 41) Plane 4 0.93 (N= 42) 0.97 (N= 41) Plane 5 0.88 (IV= 42) 0.90 (N= 35) the purpose of this study to investigate signal changes in MRI relative to histologic parameters. Further investigations in this field are in progress. In our study, the pixel size used for measuring cartilage thickness was about 0.3 mm. The average cartilage thickness was 2.5 mm. This indicates an error in measurement of about 12%. The mean percentage error in our study was 8.4%. The dependency on pixel measurements in MRI accounts for the closer correlation and greater reliability in cartilage thickness in the thicker layers. In our opinion, the volumetric measurement of cartilage [16] is also pixel-dependent. In the course of osteoarthritis, topographic changes are significant, but no topographic classification is possible in studies which assess articular cartilage volume. However, the minimum interval change in cartilage volume that could be measured with MRI is about 10-12% of the normal volume of these cartilages [16]. The correlation in our series was better than expected. The cartilage surface in our study was easily detected because we investigated the proximal tibial articular surface without the femoral articular surface. This allowed accurate and precise segmentation of the cartilage. In a surgically-induced model of osteoarthritis in gouts, there was no consistent correlation between the cartilage thickness of the gross specimen and that in the MRI apparent [17] where cartilage surfaces were directly touching. Table II. Correlation coefficient of cartilage thickness in the magnetic resonance image and histological section depending on the grade of osteoarthritis Grade Correlation coefficient Grade 0 0.96 (N= 287) Grade 1 0.96 (N= 77) Grade 2 0.90 (N= 33) Grade 3 0.97 (N=9) Studies comparing histological sections and MRI are fraught with the potential for error due to imprecise duplication of the planes in the two investigations. In our study, we tried to eliminate this error by precise positioning and accurate geometrical measurements. In the course of processing, the samples may have been contaminated by artefacts. Information is available on the shrinking of lungs by 7-80% in formalin [18]. Using acrylic media, a shrinking effect of up to 25% in lienal tissue has been observed [19]. In contrast to these results with parenchymatic tissues, we did not observe any major influence on the evaluated parameters in bony tissues. By comparing gross sectional cuts and histologic sections, we used well established methods for the preservation and preparation of the specimens. However, detailed investigation of the effect of specimen processing on artefacts was not part of this study. Jonsson et al. [3] compared cartilage thickness measurements of the hip and knee joint in vivo with X-ray, MRI and ultrasonography. He stated that the most precise measuring method was plain film radiography. In our opinion, the measurements in X-rays depend on many factors, the most significant being the roentgenographic magnification. The measurement of the cartilage thickness depends on the positioning of the joint, and different degrees of flexion and extension may vary the location of measurement. Investigations of hip joints revealed that measurements of cartilage thickness in MR images are not sufficient to be of value in clinical practice [20]. The minor mean cartilage thickness in hip joints in comparison with the present study [20] and the totally different anatomical and geometrical structure of the hip joint may account for these differences. A reduction in articular cartilage thickness in rabbit knees, caused by an experimentally induced loss of proteoglycans, was detected in MR images [21]. Chandnani et al. [22] measured cartilage thickness in six cadaveric knees. He compared the cartilage thickness of MR images with gross sections by measurements with a magnifying glass in a standardized area of the medial and lateral femoral condyle, the medial and lateral tibial plateau, and the patella. In 30 measurements, the cartilage thickness in the gross section ranged from 0.5-4.0 mm and the standardized areas were judged to be most accurately matched with Tl-weighted hybrid fat suppression images. Differences ranged from 0.1-2.9 mm. Karvonen et al. [23] used a Tl-weighted sequence with a slice thickness of 3 mm and a gap of 1.5 mm for his studies on eight knees. The thickness was
186 Kladny et al.: Cartilage thickness measurement in MRI measured with a dissecting microscope and a transparency with parallel lines 0.1 mm apart laid over the MR image or the specimen. He compared thickness measurements in over 200 sites in four knees and found no significant differences between MRI and gross thicknesses. The results were not correlated with different grades of osteoarthritis or different locations. The given data in this study suggest that, with limitations, MRI is a reliable method for measurement of the cartilage thickness independent of the grade of osteoarthritis. This offers the possibility of in vivo measurements. However, the mean percentage difference between cartilage thickness in MRI and histology is about 10%. Improved ~ imaging techniques with a better resolution in the plane will improve the reliability of cartilage thickness measurements in MRI studies. Acknowledgments The authors thank Dr William J. Ledbetter and Dr Stanley Moody for their assistance in preparing the manuscript. References 1 Ekholm R, Ingelmark BE. Functional thickness variations of human articular cartilage, Acta Societatis Medicorum Upsaliensis 1957;57:39-59. 2 Hall FM, Wyshak G. Thickness of articular cartilage in the normal knee, JBJS (Am) 1980;62-A:408-413. 3 Jonsson K, Buckwalter K, Helvie M. Niklason L, Martel WAD. Precision of hyaline cartilage thickness measurements, Acta Radiol 1992;33: 234-239. 4 Broderick LS, Turner DA, Renfrew DL, Schnitzer TJ, Huff JP, Harris C. Severity of articular cartilage abnormality in patients with osteoarthritis: evaluation with fast spin-echo MR vs arthoscopy, Am J Rheum 1994;162:99-103. 5 Glfickert K, Kladny B, Blank-Sch~il A, Hofmann G. MRI of the knee joint with a 3-D gradient echo sequence---equivalent to diagnostic arthroscopy? Arch Orthop Trauma Surg 1992;112:5-14. 6 Tyrell RL, Glfickert K, Pathria M, Modic MT. Fast three-dimensional MR imaging of the knee: comparison with arthroscopy. Radiology 1988;166: 865-872. 7 Reiser MF, Bongratz G Erlemann R, Strobel M, Pauly T, Gaebert K, Stoeber U, Peters PE. Magnetic resonance in cartilaginous lesions of the knee with three-dimensional gradient-echo imaging, Skeletal Radiol 1988;17:465-471. 8 Delling, G, Simplified metacrylate embedding method for undecalcified bone sections, Beitr Path 1972; 145:100-105. 9: Heron CW, Calvert PTAD. Three-dimensional gradi- ent-echo MR imaging of the knee: comparison with arthroscopy in 100 patients, Radiology 1992;183: 839-844. 10 Gylys-Morin VM, Hajek PC, Sartoris DJ, Resnick DMJ. Articular cartilage defects: detectability in cadaver knees with MR. Am J Rheum 1987;148: 1153-1157. 12 Tervonen O, Dietz MJ, Carmichael SW, Ehman RL. MR imaging of hyaline cartilage: evaluation of twoand three-dimensional sequ'ences. J Magnnetic Resonance Imaging 1993;3:663-668. 13 Hayes CW~ Conway WE Evaluation of articular cartilage: radiographic and cross-sectional imaging techniques. Radiographics 1992;12:409-428. 14 Wilson D, Paul PK, Roberts ED, Blancuzzi V, Gronlund-Jacob J, Vosbeck KK, Di Pasquale G, O'Byrne EM. Magnetic resonance imaging and morphometric quantitation of cartilage histology after chronic infusion of interleukin 1 in rabbit knees. Proc Soc Exper Biol Med 1993;203:30-37. 15 Gahunia HK, Lemaire C, Cross AR, Babyn P, Kessler MJ, Pritzker KP. Osteoarthritis in rhesus macaques: assessment of cartilage matrix quality by quantitative magnetic resonance imaging. Agents Actions (Suppl.) 1993;39:255-259. 16 Peterfy CG, van Dijke CF, Janzen DL, Gltier CC, Namba R, Majumdar S, Lang P, Genant HK. Quantification of articular cartilage in the knee with pulsed saturation transfer substraction and fat-suppressed MR imaging: optimization and validation. Radiology 1994;192:485-491. 17 Ho C, Cervilla V, Kjellin I, Haghigi P, Amiel D, Trudell D, Resnick D. Magnetic resonance imaging in assessing cartilage changes in experimental osteoarthrosis of the knee. Invest Radiol 1992;27: 84-90. 18 Bancroft JD, Stevens A, Turner DR, Eds. Theory and Practice of Histological Techniques, 3rd edn. Edinburgh, London, Melbourne and New York: Churchill Livingstone 1990. 19 Burkhardt R. Prfiparative Voraussetzungen zur klinischen Histologie des menschlichen Knochenmarks. Blut 1966,13:337-357. 20 Hodler J, Trudell D, Pathria MN, Resnick D. Width of the articular cartilage of the hip: quantification by using fat-suppression spin-echo MR imaging in cadavers. AJR 1992,159:351-355. 21 Paul PK, O'Byrne E, Blancuzzi V, Wilson D, Gunson D, Douglas FL, Wang JZ, Mezrich RS. Magnetic resonance imaging reflects cartilage proteoglycan degradation in the rabbit knee. Skeletal Radiol 1991;20:31-36. 22 Chandnani VP, Ho C, Chu P, Trudell D, Resnick D. Knee hyaline cartilage evaluated with MR imaging: a cadaveric study involving multiple imaging sequences and intraarticular injection of gadolinium and saline solution. Radiology 1991;178: 557-561. 23 Karvonen RL, Negendank WG, Fraser SM, Mayes MD, An T, Fernandez-Madrid E Articular cartilage defects of the knee: correlation between magnetic resonance imaging and gross pathology. Ann Rheum Dis 1990;49:672~75.