Annals ofthe Rhewmatic Diseases 1990; 49: 672-675

Articular cartilage defects of the knee: correlationbetween magnetic resonance imaging and gross

pathologyRobert L Karvonen, William G Negendank, Susan M Fraser, Maureen D Mayes, Teisa An,Felix Fernandez-Madrid

AbstractMagnetic resonance imaging (MRI) of theknee articular cartilage is possible owing tothe contrast provided by different signalintensities of adjacent menisci and sub-chondral bone. The objective of this studywas to determine the accuracy of MRI inquantitatively detecting thinning and focaldefects of articular cartilage in vivo. Highresolution MRI was performed foliowed bydissection of the knee within one hour ofamputations above the knee of eight patients(62-89 years) with peripheral vascular disease.Articular cartilage was examined for erosions,surface irregularities, and appearance. Meanthicknesses of femoral and tibial articularcartilage sagittal sections from MRI were

statistically indistinguishable from matchedgross thicknesses. In those joints in whichcartilage erosions, thinning, or irregularitieswere detected by MRI the same defects were

apparent by gross examination. Cartilage thatappeared normal by MRI had a normal grossappearance by gross examination. Thus highresolution MRI can accurately predict grossarticular cartilage appearance and thickness,allowing an objective, quantitative, non-

invasive assessment of eroded cartilage.

Department of InternalMedicine, Wayne StateUniversity School ofMedicine, Detroit,Michigan, USAR L KarvonenW G NegendankS M FraserM D MayesF Fernandez-MadridDepartent of Pathologyof Harper Hospital,Detroit, Michigan, USAT AnCorrespondence to:Dr Robert L Karvonen,Hutzel Hospital,Center for RheumaticDiseases, 4707 St Antoine,5 West, Detroit, MI 48201,USA.Accepted for publication19 October 1989

Conventional radiographs are not helpfulin evaluating cartilage erosions. Magneticresonance imaging (MRI) is a promising non-

invasive means of evaluating joint integrity.Magnetic resonance imaging of the knee definesarticular cartilage clearly owing to the contrastprovided by different signal intensities ofadjacent menisci and subchondral bone. Itsusefulness in depicting the normal anatomicallandmarks of the knee as well as changes inmenisci, ligaments, and periarticular structureshas been reported.`8 In addition, Stoller et alfound an excellent correlation betweenabnormal meniscal signals in MRI and theextent of mucinous degeneration in histo-pathology of knees obtained from necropsy or

after amputation.9 Little work has been done toascertain the value of MRI in assessing discretearticular cartilage defects, though its potentialhas been recognised in studies of the knee andwrist.2 'o Intra-articular injection of contrastagents can help in the detection of cartilagedefects,12 but its invasive nature and the need torepeat this procedure to assess the progressionof erosions make this approach less desirable.Wojtys et al have reported that MRI can detectarticular cartilage defects disclosed by arthro-tomic or arthroscopic examination.13 The pur-

pose of this study was to determine whetherMRI can be used to detect and measure changesin the thickness and focal architecture ofarticular cartilage. Preliminary results of thisstudy have been reported in abstract form.'4

Patients and methodsTable 1 gives details of the patients studied.

Magnetic resonance imaging was performedon extended knee joints obtained within onehour of amputations above the knee in eightpatients with peripheral vascular disease. Highresolution, TI weighted, spin echo coronal andsagittal images were obtained with a 256x256matrix in a field of view of 24 cm (0-95mm/pixel) in either a 1 0 tesla (five knees) or a15 tesla (three knees) Siemen's instrumentusing opposed surface coils in a Helmholtzconfiguration as the receiver. Repetition time(TR) was 600 or 700 ms and echo time (TE) was22 or 17 ms, respectively. Multiple sagittalimages 3 mm thick were obtained with a gap of15 mm-that is, at 4-5 mm intervals. Thefrequency encoding axis was perpendicular tothe articular cartilage on the tibial plateau.Chemical shift artefact-that is, overlap ofsignals from the marrow fat-was not sufficientto interfere with visualisation of the tibial orfemoral articular cartilage because the thicknessof subchondral bone (which has very low signal)separated the fatty marrow from the articularcartilage.

In our experience T1 weighted imagesproduced the best contrast between subchondralbone, articular cartilage, and the menisci, andtherefore only TI weighted images were usedfor the analysis of articular cartilage thicknesses.Window settings were adjusted to optimise thecontrast between articular cartilage and sub-chondral bone or meniscal cartilage. It shouldbe emphasised that window settings optimisedfor this purpose are not necessarily the best forobtaining a good photographic reproduction,and thus the quality of the original magnetic

Tabk I Patient data

Patient Age Sex RaceNo

1 62 F W2 79 F B3 89 F B4 66 F W5 84 F W6 75 M B7 65 M W8 74 F B

Mean age was 74 (SD 9 6) years; F:M=6:2; W:B=4:4.

672

on July 7, 2020 by guest. Protected by copyright.

http://ard.bmj.com

/A

nn Rheum

Dis: first published as 10.1136/ard.49.9.672 on 1 S

eptember 1990. D

ownloaded from

Aricular cartilage defects ofthe knee

resonance image was not fully reproducible inphotographs.

Imaging of the intact knee took about 15minutes and was immediately followed byarthrotomy and pathological examination of thecartilage surfaces and menisci. To provide a

quantitative comparison four knee joints(patients 5-8) were sectioned and articularcartilage thicknesses measured. Because thejoints were sectioned and measured immediatelythere were no effects due to freezing or dryingof the cartilage. None of the joints studied hadan abnormal amount of fluid in the joint. Thearticular cartilage was examined for erosionscharacteristic of degenerative joint disease andfor surface irregularities and thickness. Menisciwere examined for surface irregularities andappearance.The tibial plateaux and femoral condyles of

four of the patients (5-8) were cut into severalsagittal sections immediately after MRI (fig 1).In the coronal plane the position of each grosssagittal section was measured, as well as itsdistance from the centre of the knee. Also in thecoronal plane the largest bone width across bothcondyles was measured, and the distance ofeach sagittal section from both ends of thelargest bone width was recorded. Each grosssagittal section was traced on graph paper. Forthe femoral condyles an arbitrary starting pointwas noted on each section and its graph papertracing, near the posterior aspect of the knee.For the tibial cartilage the flatter surface of theplateau allowed the starting point to be placedexactly at the centre of each side of the plateau.An overlay of parallel lines 1 cm apart was

placed on each sagittal section using the arbitrarypoint to align the first parallel line, and thearticular cartilage thickness was determined towithin 0 1 mm, where each line intersected thecartilage surface. The gross sagittal sectionswere then matched to the corresponding MRIsections using the coronal plane gross measure-ments and the coronal MRI. The tracings ofeach gross sagittal section were entered into a

computer data base and an overlay image of thegross section matching the distance scale of theMRI was generated on plastic. The overlay withparallel lines 1 cm apart was placed over the

sagittal magnetic resonance images of the cor-

responding slice, positioned for an exact match,and the cartilage thickness determined for eachof the points corresponding to the grossmeasurements. Magnetic resonance imagingand gross measurements were performed

separately at different times to eliminate biasedmeasurements. Magnetic resonance imagingand gross thickness measurements were madewith a dissecting microscope and a transparencywith parallel lines 0 1 mm apart, laid over theMR image or the specimen respectively.Measurements in this study were obtained atcartilage sites where partial volume averagingdid not occur. Matched cartilage thicknesseswere determined for several dozen differentarticular cartilage sites of each individual knee.Intraobserver variability in measuring articularcartilage thickness for two different observerswas 3-2% and 10-5% respectively (table 2).Interobserver variability averaged 17-8% (table3). Although two observers in table 2 and threeobservers in table 3 were used to test vari-ability, only one observer (with the smallestintraobserver variability, see table 2) was usedin all subsequent measurements.

Cartilage and synovial membrane for histologywere fixed in 10% buffered formalin, embeddedin paraplast (Fisher Scientific, Livonia, MI),sectioned, and stained with haematoxylin andeosin.

Paired Student's t tests and the standarderrors for the differences were used to deter-mine if the matched sets of thickness measure-ments were different.

Results and discussionThe articular cartilage specimens showedanatomical changes ranging from normal toseverely eroded cartilage with no histologicalevidence of inflammation. A typical MRI sec-tion through normal cartilage shows an evenarticular surface consistent with the grossexamination of the specimen (fig 2). A typicalMRI section through a severely eroded femoralcondyle shows an articular cartilage surface withobvious abnormalities (fig 3). Figure 4 shows amagnetic resonance sagittal image through asection of thinning articular cartilage flanked bynormal regions. Figure 5 shows an example of asmall focal defect (about 1 mm in diameter

Tabk 2 Intraobserver variability in the measurement ofaricsdar cartilage thickness by magnetic resonance imaging

Observer Mean Coeffwient of nthickness (SD) variation (%)

1 1-91 (020) 105 42 1-41 (004) 3-2 5

n=number of measurements.

Figure I Femoralcondyles sectioned sagiuallyto allow gross measuremeofartukar cartilagethickness. The posteioraspect ofthe knee isdownard. (PaientNo 5.)

Table 3 Interobserver variability in the measurement ofaricular cartilage thickness by magnetic resonance imagingfor three observers

Site Mean Coefficient ofthickness (SD) variation (%)(mm)

1 1-63 (0-21) 12-82 1-69 (0 50) 29-73 2-09 (0-48) 23-24 1-94(0-10) 5-4

Average 17 8

Four sites were selected at random. Each observer made onemeasurement at each site.

673

on July 7, 2020 by guest. Protected by copyright.

http://ard.bmj.com

/A

nn Rheum

Dis: first published as 10.1136/ard.49.9.672 on 1 S

eptember 1990. D

ownloaded from

Karwnen, Negendank, Fraser, Mayes, An, Femandez-Madrid

Figure 2 Magneticresonance image ofnormalarticlr cartilage with theposterior aspect ofthe kneeto the right (A) andcorresponding gross femoralcartilage slice with kneeposterior aspect also to theright (B). (Patient No4.)

I

K-AV

:¢_~4,."Pl

Figure 3 Magneticresonance image ofseverelyerodedfemoral aricularcartilage with the posterioraspect ofthe knee to the right(A) and correspondingcarilage by grossexamination uith kneeposterior aspect downward(B). White arrows on thegross photograph (B)indicate the position ofthesagittal magnetic resonanceimage. (PatientNo 8.)

Figure 4 Magnetic resonance image ofthinning (arrow)femoral anicudar cartilage with the posterior aspect oftheknee to the right (A) and corresponding thinning (blackarrow) by gross examination with knee posterior aspectdownward (B). White arrows on the gross photograph (B)indicate the position ofthe sagittal magnetic resonance image.(PatientNo S.)

-I

extending down to the bone) detected on amagnetic resonance image and appearing onthe corresponding gross photograph. Histo-logical examination of the defect and its sur-rounding area showed the centre of the defectwith an absence of cartilage, many cell clusters,large erosions, and areas of 'fibrillation'.'5 At0-45 mm from the centre of the defect histo-logical examination showed fraying of the carti-lage and fibrillation. At 0-75 mm from thecentre the cartilage was somewhat frayed. At1-75 mm and 3-25 mm from the centre of thedefect the cartilage morphology was normalexcept for a rare cluster of cells at 1-75 mm.Thus histological analysis showed a narrow zoneof eroded and abnormal cartilage at and near thedefect, in agreement with MRI. In those joints

Figur S Magnetic soance inage ofa small defect(arrm ) i femoralariular cartilage with the postrior aspectofte kee to de right (A)and the same defect (black arrow)by gross tkne posterior aspectdowwrd(B). Whitearrows on t gross photograph (B) indicate thepsition ofthe sagital magntic resonance image. (PatientNo3.)

674

on July 7, 2020 by guest. Protected by copyright.

http://ard.bmj.com

/A

nn Rheum

Dis: first published as 10.1136/ard.49.9.672 on 1 S

eptember 1990. D

ownloaded from

Artcular carl age defects of the knee

Table 4 Comparison of human knee artciar cartilage thicknesses measured by magneticresonance imagig (MRI) and gross examinaton

Patent Cartilage Side n Mean thickness (SD) StandardNo in mm error

for theMRI Gross difference

5 Femoral M 18 1-3 (0-4) 1-3 (0-4) 0-1S Femoral L 18 1-9(0-5) 1-9 (0-5) 0-2S Tibial M 13 1-3 (0-5) 1-4-(0-4) 0-2S Tibial L 9 1-6 (0-3) 1-7 (0-4) 0-26 Femoral M 20 2-3 (1-0) 2-4 (1-2) 0-36 Femoral L 17 2-0 (0-5) 2-1 (0-6) 0-26 Tibial M 15 2-6 (0-8) 2-6 (0-8) 0-36 Tibial L 12 1-6 (0-4) 1-4 (0-5) 0-27 Femoral M 20 2-5 (0-8) 2-7 (0-6) 0-27 Femoral L 22 2-2 (0 4) 2-2 (0-3) 0-17 Tibial M 13 2-4 (0-9) 2-4 (0-7) 0-37 Tibial L 12 3-1 (1-4) 3-1 (1-5) 0-68 Femoral M 13 1-5 (0-9) 1 5 (0-9) 0-48 Femoral L 21 2-2 (0-8) 2-3 (0-8) 0-28 Tibial L 8 1-8 (0-4) 1-8 (0-5) 0-2

n=number of different sites of articular cardlage where thickness measurements were performed;M=the medial side; L=the lateral side.In every case the paired t test of the means indicated no significant difference between MRI andgross thicknesses (p>0 05).

in which cartilage erosions, th' ning, or ir-regularities were detected b the samedefects were apparent by gross exmination. Inthose joints in which abnormal signal intensitywas apparent in a portion of a meniscus by MRIthe specimen often showed abnormal grossstructural appearance. Cartilage and meniscithat appeared normal by MRI had a normalgross appearance by pathological eymination.There was no statistically signicant differ-

ence between the mean thic kess of all themeasured points from gross examination andthose obtained by MRI (table 4). Thus MRimages taken in 3 mm thick slices can detectvariations in cartilage thickness of the order of 1mm as well as focal defects of 1 mm diameter.The MRI signals observed in articular carti-

lage are of higher intensity than those ofsubchondral bone or menisci. The areas oflower signal intensity were usually thickerunder tibial than under femoral cartilage, and inthe tibial plateaux were often several milli-metres thick, especially in the central region ofeach side-for example, fig 5.The MRI sagittal sections were adjusted to

sample signal intensity in a 3 mm slice thickness.Cartilage thickness could not be obtained insections where the surface of the femoral

Fige 6 Magnegc reso e ime ofafemoral arcularcarte region (arrow) showigptl volume aveaging ofcartilage and bone signals. The posterior aspect ofthe knee isto the right. (PatientNo 7.)

articular cartilage crossed the section at an anglegreater than about 450 because partial volumeaveraging'6 of cartilage signals with those ofbone occurred. An example is shown in fig 6.The cartilage thickness could be measured if theslice through this region were obtained at adifferent angle where a perpendicular to thearticular surface would be parallel to the planeof the MRI section. Partial volume averagingdiffers from focal areas of denuded bone on thearticular surface, in that the latter are usuallyflanked by cartilage and do not produce anormal signal even if realigned. Tibial articularcartilage thicknesses can be determined over agreater percentage of the surface as the tibialplateau is relatively flat with perpendiculars tomore points in its plane parallel to the sagittalsections.Thus high resolution MRI can accurately

predict gross appearance and thickness ofarticular cartilage, allowing an objective,quantitative, non-invasive in vivo assessment.This non-invasive method may prove valuablein determining the rate of cartilage degenerationand repair, and in assessing the long termeffects of drug treatment on eroded articularcatage.

The authors wish to thank Neil Keller for technical assistance inMRI, Dr George Corondan for generous use of equipment, theHape Hospital pathology staff for their help in obtaining kneespecimens, and the Michigan Chapter Arthritis Foundation andthe Harper-Grace Hospital Radiation Oncology Department forpatial support of this proje. Magnetic resonance imaging wasperformed in the Vaitkevicius MR Center at Harper Hospital.

1 Beltran J, Noto A M, Mosure J C, Weiss K L, Zuelzer W,Christoforidis A J. The knee; surface-coil MR imaging at1-5 T. Radiology 1986; 159: 747-51.

2 Burk D L, Kanal E, Brunberg J A, Johnstone G F, SwensenH E, Wolf G L. 1-5 T surface-coil MRI of the knee. AJR1986; 147: 293-300.

3 Haggar A M, Froelich J W, Hearshen D 0, Sadasivan K.Meniscal abnormalities of the knee: 3DFT fast-scanGRASS MR imaging. AJR 1988; 150: 1341-4.

4 Jackson D W, Jennings L D, Maywood R M, Berger P E.Magnetic resonance imaging of the knee. AmJ Sor Med1988; 16: 29-37.

5 Manco L G, Lozman J, Coleman N D, Kavanaugh J H,Bilfield B S, Dougherty J. Noninvasive evaluation of kneemeniscal tears: preliminary comparison ofMR imaging andCT. Radiology 1987; 163: 727-30.

6 Mink J H, Levy T, Crues J V. Tears of the anterior cruciateligament and menisci of the knee: MR imagng evaluation.Radiology 1988; 167: 769-74.

7 Polly D W, Claghan J J, Sikes R A, McCabe J M,McMahon K, Savory C G. The accuracy of selectivemagnetic resonance imaging compared with the findings ofarthroscopy of the knee. J Bone joint Surg [Am] 1988; 70:192-8.

8 Reicher M A, Hartzman S, Bassett L W, Mandelbaum B,Duckwiler G, Gold R H. MR imaging of the knee. Part I.Traumatic disorders. Radiology 1987; 162: 547-51.

9 Stoller D W, Martin C, Crues J V, Kaplan L, Mink J H.Meniscal tears: pathologic correlation with MR g.Radiology 1987; 163: 731-5.

10 Koenig H, Lucas D, Meissner R. The wrist: prelimiaryreport on high-resolution MR imaging. Radiol 1986;160: 463-7.

11 Li D K B, Adams M E, McConkey J P. Magnetic resonanceimaging of the ligments and menisci of the knee. RadiolCin North Am 1986; 24: 209-27.

12 Gylys-Morin V M, Haiek P C, Sartoris D J, Resnick D.Articular cartilage defects: Detectability in cadaver kneeswith MR. AJR 1987; 148: 1153-7.

13 Woitys E, Wilson M, Buckwalter K, Braunstein E, MartelW. Magntc resonance imaging of knee hyaline carlageand intraarticular pathology. Am J Sports Med 1987; 15:455-63.

14 Karvonen R L, Negendank W G, Fernandez-Madrid F,Fraser S M. Ptholgic correlation of articular cartilagedefects in MR imaingn of the knee [Absract. ArthrisRh.e. 1988; 31 (sujipI): S19.

15 Mankin H J. Normal articular cartilage and the alteraions inosteoarthitis. In: Lombardino J G, ed. Nostroidala _tuunusao, ihwgs. New York: Wiley, 1985: 1-73.

16 Rich on ML, Helms C A. Artifai, normal variants, andimaging pitfalls of musculoskeletl nagnetic resonanceimaging. Radiol Clin North Am 1986; 24: 145-77.

675

on July 7, 2020 by guest. Protected by copyright.

http://ard.bmj.com

/A

nn Rheum

Dis: first published as 10.1136/ard.49.9.672 on 1 S

eptember 1990. D

ownloaded from