Diurnal Variation in the Femoral Articular Cartilage of theKnee in Young Adult Humans

John C. Waterton,1* Stuart Solloway,2 John E. Foster,3,4 Michael C. Keen,3

Stephen Gandy,3,4 Brian J. Middleton,5 Rose A. Maciewicz,1 Iain Watt,6

Paul A. Dieppe,4 and Christopher J. Taylor2

Our objective was to test the hypothesis that diurnal changesoccur in thickness or volume of the femoral articular cartilage ofthe knee in asymptomatic young adults. Fat-suppressed three-dimensional (3D) spoiled gradient-echo magnetic resonanceimaging (MRI) was employed. Six volunteers each were scannedearly in the morning and at the end of a working day spentmainly standing. This protocol was repeated on 3 successiveweeks. Femoral cartilage volumes were obtained via semiauto-matic segmentation that employed a seeding algorithm. Thesesegmentations then were regridded onto a 500-pixel template,and differences in the resulting thickness maps were assessed.Analysis of variance showed no significant diurnal variation inoverall volume or thickness. The reproducibility for volume(test-retest coefficient of variation) was 1.6%. There were, how-ever, statistically-significant diurnal changes in the thicknessmaps. Cartilage thickness decreased by up to 0.6 mm during theday in each of the following three specific locations: thepatellofemoral compartment, the lateral tibiofemoral compart-ment, and the medial tibiofemoral compartment. Elsewhere,cartilage thickness was unchanged or increased by up to 0.5mm. We conclude that, in asymptomatic young adults, cartilagevolume does not change during the day; however, the cartilagedoes become thinner in locations that encounter the greatestbiomechanical force. Magn Reson Med 43:126–132, 2000.

r 2000 Wiley-Liss, Inc.

Key words: cartilage volume; cartilage thickness; knee; diurnalchange; image analysis

Articular cartilage provides a low-friction, low-wear bear-ing surface for articulating bones. It consists largely ofextracellular matrix, mainly composed of collagen and theaggregating proteoglycan, aggrecan. Embedded within thematrix are cells of only one major class, the chondrocytes.Despite this apparent simplicity, the structure of the carti-

lage varies, both radially and laterally, in histology, inmaterial properties, and in magnetic resonance imaging(MRI). Both cells and matrix differ in regions of elevatedbiomechanical stress, but it is not fully understood howthese structural variations help accommodate the differingpatterns of loading encountered in different parts of thejoint. Although the effects of pressure on cartilage can bestudied ex vivo, studying the living joint has considerableadvantages because the capsule remains intact, tendontension and muscle tone are maintained, and all the dailybiomechanical stresses in this complex joint are present.

Osteoarthritis (OA) is one of the principal causes ofdisability in elderly people. The disease is characterized byfocal structural changes in, and eventual loss of, articularcartilage. The knee joint is often most severely affected,and anatomic locations suffering the highest biomechani-cal force are most likely to exhibit cartilage damage (1).

Recently, it has become possible to measure accuratelyand precisely the volume of the articular cartilage withMRI (2–16). Fat-suppressed three-dimensional (3D) spoiledgradient-echo MRI has been widely adopted and providesgood contrast with reasonable scanning times. However,because of the shape of the cartilage, manual segmentationof the images is tedious, so that for large-scale studies, it isdesirable to develop semiautomatic techniques. It nowappears possible that MRI assessment of cartilage volumewill be accurate and precise enough to measure OA diseaseprogression and therapeutic intervention in small-scaletrials, unlike the X-ray assessment of joint-space narrowing(17), which has poor statistical power. Before MRI can beemployed in prospective controlled trials of structure-modifying (cartilage-preserving) drugs, however, it is essen-tial to understand potential confounding factors, such asdiurnal variation.

It is well-known that people are taller in the morning(18). During the day, when people are upright, there areconsiderable static and dynamic pressures on the interver-tebral discs, and water is expelled from the proteoglycanmatrix. At night, when the spine is horizontal, water istaken up again by the disc. The legs also experience greaterstatic and dynamic pressures during the day than duringthe night. Because the structure of hyaline articular carti-lage resembles that of the disc, we reasoned that similardiurnal changes might be detectable in the articular carti-lage of the knee.

The aims of this study were, therefore, to determinewhether diurnal variation can affect the precision of carti-lage volume measurement using MRI, and to characterizeany such diurnal variations.

1Cardiovascular, Metabolism & Musculoskeletal Research Department, Astra-Zeneca, Alderley Park, Macclesfield, Cheshire, UK.2Department of Medical Biophysics, University of Manchester, Manchester, UK.3Department of Medical Physics & Bioengineering, United Bristol HealthcareTrust, Bristol, UK.4Department of Rheumatology, University of Bristol, Bristol, UK.5Safety of Medicines Department, AstraZeneca, Alderley Park, Macclesfield,Cheshire, UK.6Department of Clinical Radiology, United Bristol Healthcare Trust, Bristol, UK.Presented in part at the 2nd Meeting of the British Chapter of ISMRM, London,1996; the 5th Meeting of ISMRM, Vancouver, 1997; and the 3rd Meeting of theBritish Chapter of ISMRM, Manchester, 1997.Grant sponsors: Medical Research Council; Engineering and Physical Sci-ences Research Council; AstraZeneca.John Foster is currently at the Western Infirmary, Glasgow, UK. Stuart Sollowayis currently at Kestra Ltd., Skipton, UK.*Correspondence to: John C. Waterton, AstraZeneca, Alderley Park, Maccles-field, Cheshire SK10 4TG, UK. E-mail: john.waterton@astrazeneca.comReceived 16 September 1998; revised 2 September 1999; accepted 20September 1999.

Magnetic Resonance in Medicine 43:126–132 (2000)

126r 2000 Wiley-Liss, Inc.

METHODS

Volunteers and Study Protocol

After local ethical committee approval, six volunteers(three of each gender) aged 21 to 25 years without symp-toms of any musculoskeletal disorder were imaged in themorning (AM) at 07:45, and at 16:45 near the end of theirworking day (PM). The volunteers lived within 3 km of theRadiology Department at the Bristol Royal Infirmary andgenerally walked to the Infirmary, although they occasion-ally used a car or bus. When they arrived at the RadiologyDepartment they were allowed to sit down and werescanned within 15 minutes of arriving at the Department.Their days were spent predominantly in the laboratory,mostly standing at a bench. The imaging protocol wasrepeated on 3 consecutive weeks, although data were notobtained from one of the volunteers at week 2, so that intotal 34 of the possible 36 scans were available for analysis.

MR Image Acquisition

The MR protocol was designed to ensure that high contrastwas achieved between cartilage and surrounding tissues,but the signal from the cartilage itself was fairly uniform, sothat semiautomatic segmentation could be employed. TheMR parameters were as follows: field strength 1.0T (SiemensImpact); sagittal 3D spoiled gradient echo; fat saturation;TR 47 msec; TE 11 msec; flip angle 40°; 192 phase-encoding steps; 64 slice partitions; matrix zero-filled to256 3 256 3 64; FOV 140 mm; no signal averaging, givingvoxel dimensions 0.55 3 0.55 3 1.56 mm. The totalimaging time was 9.6 minutes, similar to or faster thanprotocols in other centers (3,7,12,14). The gradient calibra-tion was checked monthly by Siemens engineers: theaccuracy was 0.39% and the reproducibility was ,0.8%.

Segmentation

Three broad approaches to segmentation of the articularcartilage have been described previously, namely, manualdrawing (2,3), semiautomatic data-driven methods (4–6,8,12–14,16), and shape modelling (19). Manual drawing can berather operator dependent and, because of the time-consuming nature of the operation, prone to errors causedby operator fatigue. Supervised semiautomatic data-drivenmethods may use a combination of thresholding, seeding,connectivity, and edge detection; such methods should befaster and less subjective than manual drawing, but, be-cause of the high surface area–to-volume ratio in thearticular cartilage and the consequent risk of partial vol-ume averaging, the accuracy of these methods is criticallydependent on the behavior at the cartilage boundary of thealgorithms employed. Shape modelling offers accurate andextremely rapid segmentations (19) provided that an ad-equate training set is available; however, shape-modellingmethods have not yet been validated in 3D. This studyemployed data-driven segmentation (‘‘Tosca’’ version 2.3)(IBM, Winchester, UK) (20) giving rapid and precise mea-surements of volume and thickness assessment at subvoxelresolution. All segmentations were performed by the sameoperator (J.E.F.). All femoral condyle cartilage was in-cluded in the analysis, in contradistinction to some otherstudies (4,5). To segment the femoral articular cartilage, aseed-point was placed within the area to be ana-

lyzed. A volume-growing algorithm (21) was then startedusing interactively determined values for the cartilage. Theresulting region of interest was enclosed by a contour. Theprocedure was repeated for each slice in the data set. Toensure anatomic accuracy, minor adjustments to the seg-mented regions were required occasionally. A proportionof the segmentations was checked by a musculoskeletalradiologist (I.W.). J.E.F.’s intraobserver coefficient of varia-tion with this approach is 1.5% for independent examina-tions (16). This semiautomatic technique was considerablyfaster than manual segmentation. Typically, a completesemiautomatic cartilage volume measurement was pro-duced in less than an hour.

Volume Measurement

Volume was calculated by counting those pixels whollyinside the segmentation contour together with those on thecontour. The method was validated using an anthropomor-phic phantom. A water phantom was constructed in theshape of femoral cartilage by molding two sheets of ex-truded polyvinyl chloride around the condyles of a femur.The molds were secured together to form a shell, whichwas subsequently filled with a measured volume of dopedwater. The root-mean-square (rms) error from four calibra-tions was 4.8%

Thickness Mapping

The 3D thickness distribution of the cartilage was obtainedby an automatic method, developed and implemented bytwo of the authors (S.S. and C.J.T.). Using the contoursgenerated by the semiautomatic volume segmentation, anautomatic landmark generation technique was used tocreate sets of corresponding landmarks on the cartilagesurface. This is a generalization of a method previouslyused in two dimension (2D) (19). The medial axis of thepoints in each slice was found using the algorithm de-scribed by Shapiro and Sklansky (22). A set of equallyspaced points was generated along the medial axis. Foreach point, the thickness of the cartilage was given by thedistance between the intercepts of the normal to the axiswith the inner and outer surface. This slice-by-slice methodoverestimates thickness where the slice plane is ,90° tothe surface, e.g., between the condyles. However, becausethe thrust of our work is to measure differences in thick-ness, these small systematic errors will mostly cancel. Withmeasurement of the thickness with respect to the medialaxis, the measurements are robust to small indentations onthe cartilage surfaces. The resulting thickness distributionmaps are analogous to those obtained by other workersusing different algorithms (6,9,22–24), although some work-ers have used a fully 3D approach that is independent ofslice orientation and avoids the associated systematicerrors (9,10,25–29). Each 3D example was then regriddedonto a thickness map comprising 25 new sagittal slices,equally spaced between the extreme lateral and medialedges of the femoral cartilage. Each slice contained 20thickness values measured at each of the landmark points.Because the same number of landmark points was em-ployed in each slice, the spacing was denser at the patello-femoral surface and at the extreme medial edge. Thediurnal change for each subject on each day was repre-sented in the difference map, PM MINUS AM The popula-

Diurnal Variations in Knee Cartilage 127

tion mean diurnal difference was obtained from the meanof all 17 PM MINUS AM maps. Additionally, a meanthickness value was calculated from each of the 34 thick-ness maps.

Statistical Analysis of Global Changes

Repeated measures analysis of variance was used to testthe hypothesis that there was a date and diurnal change inthe volume or mean thickness, allowing for gender andusing the date and time of day as the repeated measures(30). The reproducibility of the method was assessed fromthe test-retest coefficient of variation (CoV). For eachsubject, i, the CoV is the standard deviation, si, for a seriesof measurements on that subject, divided by the meanvolume, µi, for the subject. The overall test-retest CoV for agroup of N subjects is then ÎSi (si/µi)2/N. An assessment ofthe ability of the method to measure small weekly changeswas obtained from the sum of the week-to-week term (i.e.,variance for a volunteer of a given gender on a given date)s2

subj (date3sex) and the residual term s2residual in the analysis of

variance: the CoV for a weekly difference is thenÎ2(s2

subj(date3sex) 1 sresidual2 )/(Siµi/N).

Statistical Analysis of Regional Changes

To test whether there was a statistically significant patternof diurnal redistribution of the cartilage thickness, a princi-pal component and linear discriminant analysis of the 34thickness maps was performed. This analysis was under-taken without any prior assumptions about which loca-tions might be most susceptible to diurnal change.

● With the assumption of identical covariance matricesfor the two groups, a principal component analysis onnormalized thickness vectors, t, was performed, giv-ing a set of eigenvectors, P.

● The original vectors were projected onto the eigenvec-tor basis P, giving tP. This is a data reduction step thathelps to reduce the complexity of the problem (31).For example, if there are 10 examples, each consistingof 500 thickness measurements, which produce 5significant eigenvectors, we can say that these eigen-vectors describe the underlying structure of the data.In addition, the vectors are orthogonal and can betreated as independent. We can therefore use the 10 35 matrix of eigenspace projections as input data for thefollowing step.

● A linear discriminant analysis was performed on theprojected vectors, using AM and PM thickness patternsas separate classes. This gave the vector, D, whichprovided the best discrimination between AM and PM.

● The eigenvectors, tP, were projected onto the discrim-inant vector D, giving tPD.

● The scalar values derived from this projection for eachthickness map were tested to see if there were signifi-cant differences between the AM and PM maps, usingthe two-tailed Student’s t-test.

RESULTS

Volume and Mean Thickness

Figure 1 shows a typical segmentation of the femoralarticular cartilage. Figure 2a shows the calculated femoral

cartilage volume for each individual at each examination.The mean volume was 14.8 ml for males and 11.0 ml forfemales, which compares with volumes of between 7.7 mland 19.3 ml for a total of 30 individuals (cadavers andvolunteers of either gender) reported from other centers(3,7,8,10–12,14). Analysis of variance (Table 1) showedthat most of the variation arose from gender and subject-to-subject variance. Males had significantly more cartilagethan females, and this gender effect was maintained if sixadditional values reported from young adults at othercenters (12,14) were included in the analysis. There was noevidence of diurnal variation in volume: the average (AMMINUS PM) difference was 0.025 ml (95% confidenceinterval 0.125 to 20.074 ml). The test-retest CoV forvolume was 1.6%, which is comparable with the intraob-server CoV. The CoV for the difference between twomeasurements in a single volunteer, obtained from theanalysis of variance, was 2.3%.

Figure 2b shows the average femoral cartilage thicknessfor each individual at each examination; again analysis ofvariance (Table 1) showed no evidence of diurnal variationin mean thickness. There was a trend toward thickercartilage in males than females, consistent with previousobservations from MRI in young people (32).

Spatial Dependence of Thickness

Although cartilage volume and overall mean thicknessremained unchanged during the day, it still remained totest the hypothesis that focal diurnal changes occur. Figure3 is a color map showing the mean of all 17 PM MINUS AMmaps, and an example of data from a single volunteer. Cartilagethickness decreased during the day in the following spe-cific locations: the patellofemoral compartment, the lateral

FIG. 1. 3D reconstruction of a typical segmentation of the articularcartilage of the femur. Right knee, posterior view. Patellofemoralcompartment is at top, medial tibiofemoral compartment is at bottomleft, and lateral tibiofemoral compartment is at bottom right.

128 Waterton et al.

tibiofemoral compartment, and the medial tibiofemoralcompartment. Elsewhere cartilage volume was unchangedor became thicker. In the patellofemoral compartment thegreatest decrease in thickness was 0.65 mm; in the lateraltibiofemoral compartment it was also 0.65 mm, and in themedial tibiofemoral compartment it was 0.59 mm. Outsidethese areas, the greatest increase in thickness was 0.50 mm.To test whether this pattern of diurnal redistribution of thecartilage mass was statistically significant, a discriminantanalysis was performed. The objective was to characterizethe difference between the AM and PM groups. Projectionof the original thickness vectors, t, onto the first twoprincipal components showed a clear separation betweenthe different subjects but did not separate the AM and PMimages. However, a significance test of the the projection tPof the measurement vector onto the discriminant mode,giving tPD, showed that this diurnal change in the distribu-tion of cartilage thickness was indeed statistically signifi-cant (p ,0.05). This discriminant vector is shown in fig. 4.

DISCUSSION

At the outset of this work, we reasoned that the increasedpressure on the cartilage during the working day wouldexpel water and cause the cartilage volume to decrease.This hypothesis is not supported by the data; instead, someareas become thinner, while elsewhere the cartilage thick-ens by an approximately compensating amount.

Why Does Cartilage Become Thinner in Some Regions ofthe Joint During the Day?

When uniform hydrostatic or osmotic pressure is appliedto the hyaline cartilage, water is expelled. Conversely,diminished pressure causes water to be taken up (e.g., fromthe synovial fluid), and the cartilage swells. When anindividual is standing, the weight of the body above theknees exerts pressure through the medial and lateraltibiofemoral compartments. In addition, during normaljoint flexion, much higher transient pressures are transmit-ted through these compartments, especially through thepatellofemoral compartment. Our data (Figs. 3 and 4),therefore, appear to show the movement of water out ofthose compartments that suffer the greatest pressure duringwalking and standing, i.e., the load-bearing compartments.

Why Do Some Non–Weight-Bearing Regions of theCartilage Actually Become Thicker During the Day?

In vivo, because of the elasticity of the capsule and the tonefrom the surrounding muscles, the intraarticular pressurein the normal joint at rest is slightly below atmospheric.During active extension of the normal knee, intraarticularpressure falls further (33). According to Jayson (33), ‘‘Quad-riceps contraction stretches the capsule and tends toenlarge the joint space producing the subatmosphericintra-articular pressure. An analogy may be drawn withsqueezing an empty toothpaste tube in an appropriatefashion when the paste is sucked back into the emptytube.’’ Our interpretation of the data (Figs. 3 and 4) is thusthat, in the non–load-bearing regions of the joint, thereduced intraarticular pressure during joint extension per-mits ingress of water to the cartilage. Alternative interpreta-tions are also possible; a referee has stated, ‘‘I do not agreewith [this] interpretation of cartilage swelling in the non-

FIG. 2. Measurements of volume (a) and mean thickness (b) duringthe study. The six different symbols each represent the samevolunteer in the two plots (males: ., ●, m; females: s, j, r). Solidsymbols: AM; open symbols: PM. The data from volunteer r at week3 are clearly outliers; however, those data were not excluded from theanalysis.

Diurnal Variations in Knee Cartilage 129

weight-bearing area. I believe that the interstitial fluid issimply displaced from the tissue under weight-bearingareas (compression), but I do not see why a mechanismsuch as negative intra-articular pressure should play a rolein this.’’

Response of Cartilage to Pressure

The shrinking and swelling of the articular cartilage ofdiarthrodial joints as a result of normal activities of dailylife has not previously been demonstrated. Changes in thepatellar cartilage thickness have been observed by MRI as aresult of externally applied pressure (34,35) or exercise

(28), and diurnal or exercise-induced changes in the inter-vertebral disc have been detected by MRI: the diurnalvariation in lumbar disc volume is about 20% (36). An MRIstudy (28) reported a 6% decrease in patellar articular cartilagevolume in response to vigorous physical exercise (knee bends).In that patellar protocol, the load is developed over most of thearticular surface, so it is perhaps not surprising to detect asignificant volume change, whereas in our femoral study, onlya proportion of the articular surface is exposed to substantialmechanical load, and the resulting focal thinning was accom-panied by focal thickening elsewhere.

Diurnal Contribution of the Articular Cartilage to Height

The total diurnal variation in human height (18) is about 17mm. In the exponential loss of body height after rising froma recumbent position, 30% of the change occurs in the firsthour, and curves of similar appearance have been reportedfor articular cartilage (28,37). Our data show a 0.6-mmchange in the thickness of the femoral cartilage in thelateral tibiofemoral compartment, which is about 4% of thetotal diurnal variation in height. The true contribution ofthe lower extremities to diurnal height variation must beconsiderably greater than this, both because subjects werescanned $30 minutes after waking, so short-term re-sponses of the articular cartilage to standing and walkingwould have been missed, and also because only one of thesix major articular surfaces of the diarthrodial joints of thelower extremity was measured.

Volumetric Assessment of Cartilage

For the clinical assessment of OA disease progression andtreatment, it is necessary to monitor (preferably noninva-sively) the amount of cartilage present and the extent towhich it has degenerated. Patients with OA lose cartilagerelatively slowly (17). If MRI measurements are to beemployed, we need to achieve adequate reproducibility,together with some understanding of the factors affectingprecision. Although this study protocol does not includethe extremes of loading possible for the knee, it does coverthe range likely to be encountered in practice. Our datafrom this series of 34 scans show good reproducibility forcartilage volume in the femur (CoV 1.6%). Comparablevalues reported from other centers are 1.4% to 3.5% (7) in aseries of eight volunteers, 6.4% in eight arthritic patients

FIG. 3. Overall cartilage thickness difference map, i.e., the averageof 17 individual PM MINUS AM difference maps. The scale repre-sents thickness change in mm. There are 500 points in the maprepresenting thickness changes varying from 20.65 to 10.50 mm:the 5th percentile of change is at 20.52 mm and the 95th percentile isat 10.27 mm. Patellofemoral compartment is at right, medial tibiofemo-ral compartment is bottom left, and lateral tibiofemoral compartmentis top left. Inset shows one individual’s difference maps on 3successive weeks: left panel , week1; right panel , week 3.

Table 1Analysis of Variance for Volume and for Mean Thickness

Term Cause of variation Volume (ml) Mean thickness (mm)

Fixed terms Effect (95% confidence interval)

Sex Male minus female 3.88 (6.26, 1.49) P 5 0.011 0.52 (1.07, 20.03) P 5 0.059Date Week 1 vs. week 2 vs. week 3 0.02 (20.26, 0.30) P . 0.05 0.11 (20.24, 0.36) P . 0.05Time AM minus PM 0.02 (0.12, 20.07) P . 0.05 0.02 (0.22, 20.17) P . 0.05

Random terms Contribution to total variance

Subj (sex) Volunteer of a given gender 96.5% P , 0.001 53.8% P 5 0.036Subj (date 3 sex) Volunteer of a given gender on a given date 2.3% P 5 0.016 25.5% P 5 0.006Subj (time 3 sex) Volunteer of a given gender at a given time of day 0.01% P . 0.05 12.4% P 5 0.020Residual 1.1% 8.3%

For the fixed term date, data shown are for week 1 minus week 2; for week 1 vs. week 3, and for week 2 vs. week 3, (not shown) the effectswere similar (all P . 0.05). Also included in the analysis of variance were the four cross terms (time 3 sex), (time 3 date), (date 3 sex) and(time 3 date 3 sex). These were fixed in the analyses, and none was significantly different from zero. P is the probability that a termcontributed to the residual a significant additional source of variance. Although the terms subj (date 3 sex) and subj (time 3 sex) appearsignificant, the significance would be lost if the outlying data from volunteer r at week 3 were omitted.

130 Waterton et al.

(3), and up to 9.7% in various smaller series (2–6,8,12–14,16). This degree of reproducibility is gratifying in viewof the very high surface area–to-volume ratio of the articu-lar cartilage and the corresponding difficulty of segmenta-tion. If such reproducibility can also be achieved in OApatients, the MRI assessment of cartilage volume mayprovide a valuable endpoint in the monitoring of diseaseprogression and drug efficacy.

Assessment of Cartilage Thickness

Patients with OA lose cartilage not uniformly but preferen-tially at certain locations (1), so that focal measurements ofthickness would provide additional information if goodstatistical power could be achieved. Point measurements ofthickness at specific anatomic locations have been made inseveral centers with varying success, but this approachdoes not appear to offer sufficient statistical power becauseof difficulty in returning to an identical location at fol-low-up and also because very few pixels define the tran-sect. The recent development of MRI cartilage thicknessmapping (14,25,38,39) allows the accurate and reproduc-ible assessment of regional variations in thickness(7,9,10,23,24). A limitation of the thickness mapping ap-proach is that it does not offer a single well definedendpoint and hence does not readily lend itself to objectivehypothesis testing in population studies. The methodemployed in this study allows multiple images from thesame or different subjects to be regridded onto a standardtemplate. Image analysis using this standard templateallowed an objective evaluation of regional thicknesschanges common to our population and allowed tinyphysiological changes in articular cartilage thickness to bedetected from data acquired using a standard clinical MRIprotocol. A strength of the approach is that the evaluationis not influenced by prior expectations about which loca-tions might be most likely to change. Nevertheless, diurnalcartilage thinning in this study was observed precisely atthose sites known to be most susceptible to OA (1).

Study Limitations

1. A finite TE of 11 msec was employed. Our MRI pulsesequence is similar to that used and validated at othercenters (2,9); however, published validation studies(2,16,38,40) use cadaveric material in which the intraar-ticular pressure, hydration, and T2 may be nonphysi-ological. At our TE, the very short T2 components in thearticular cartilage may exhibit low-signal. We cannotexclude the possibility that some of our diurnalvariation is in signal intensity rather than thickness.

2. Various methodological limitations in spatial resolu-tion, segmentation, and thickness determination havebeen discussed above. However we do not consider itlikely that these limitations would confound ourmain conclusions. Improved methods, which are likelyto improve the statistical power of the technique, areunder development in our laboratories and elsewhere.

3. This study measured only three articular surfaces, allon the femur, and only two diurnal time points.Further studies of the knee, hip, and ankle, withbetter time resolution, are necessary to separate theeffects of static loading, dynamic loading, and intraar-ticular pressure on the articular cartilage.

4. In this study, the knees of healthy young adults werestudied. Of greater clinical interest is the exploitationof these techniques in normal elderly individuals andpatients with early OA (whose cartilage may havefocal defects presenting additional challenges in seg-mentation). MRI of the knees in these groups mayreveal, as for whole-body height, less, or faster, defor-mation under load because of the inherent differencesin cartilage matrix composition as compared withyounger individuals, providing a noninvasive assess-ment of early cartilage degeneration.

CONCLUSION

In these young adults, no diurnal change in cartilagevolume could be detected. However, when thickness mapswere constructed and regridded onto a common template,statistically significant focal changes in cartilage thicknesswere detected in those locations expected to encounter thegreatest biomechanical force.

REFERENCES1. Bullough PG. Osteoarthritis and related disorders: pathology. In: Klip-

pel JH, Dieppe PA, editors. Rheumatology. Vol. 2. London: MosbyYearbook; 1998. p 8.8.1–8.8.8.

2. Peterfy CG, van Dijke CF, Lu Y, Nguyen A, Connick TJ, Kneeland JB,Tirman PFJ, Lang P, Dent S, Genant HK. Quantification of the volume ofarticular cartilage in the metacarpophalangeal joints of the hand:accuracy and precision of three-dimensional magnetic resonance imag-ing. Am J Roentgenol 1995;165:371–375.

3. Peterfy CG, van Dijke CF, Janzen DL, Gluer CC, Namba R, Majumdar S,Lang P, Genant HK. Quantification of the volume of articular cartilage inthe knee with pulsed saturation transfer subtraction and fat-suppressedmagnetic resonance imaging: optimisation and validation. Radiology1994;192:485–491.

4. Marshall KW, Mikulis DJ, Guthrie BM. Quantification of articularcartilage using magnetic resonance imaging and three-dimensionalreconstruction. J Orthop Res 1995;13:814–823.

5. Pilch L, Stewart C, Gordon D, Inman R, Parsons K, Pataki I, Stevens J.Assessment of cartilage volume in the femorotibial joint with magneticresonance imaging and three-dimensional computer reconstruction. JRheumatol 1994;21:2307–2319.

FIG. 4. Instances of the principal compo-nents of the diurnal thickness variation.Local changes to the cartilage thicknessare investigated by moving along thevector that best discriminates betweenAM and PM measurements in the spaceof the principal components. The twoimages show extremes in variation at 12and 22 standard deviations from themean of the cartilage thickness. The redareas are those that exhibit the greatestdeviations from the mean, whereas theblue areas exhibit the least variation.

Diurnal Variations in Knee Cartilage 131

6. Eckstein F, Sittek H, Milz S, Putz R, Reiser M. The morphology ofarticular cartilage assessed by magnetic resonance imaging. Surg RadiolAnat 1994;16:429–438.

7. Eckstein F, Westhoff J, Sittek H, Maag KP, Haubner M, Faber S,Englmeier KH, Reiser M. In vivo reproducibility of three-dimensionalcartilage volume and thickness measurements with MRI. Am J Roent-genol 1998;170:593–597.

8. Eckstein F, Sittek H, Gavazzeni A, Milz S, Kiefer B, Putz R, Reiser M. DerKniegelenksknorpel in der Magnetresonanztomographie: MR-Chondro-volumetrie (MR-CVM) mittels fettunterdruckter FLASH-3D-Sequenz.Radiologe 1995;35:87–93.

9. Sittek H, Eckstein F, Gavazzeni A, Milz S, Kiefer B, Schulte E, Reiser M.Assessment of normal patellar cartilage volume and thickness usingMRI: an analysis of currently available pulse sequences. Skeletal Radiol1996;25:55–62.

10. Eckstein F, Schnier M, Haubner M, Priebsch J, Glaser C, Englmeier KH,Reiser M. Accuracy of 3D knee joint cartilage volume and thicknessmeasurements with MRI. Clin Orthop Rel Res 1998;352:137–148.

11. Eckstein F, Winzheimer M, Westhoff J, Schnier M, Haubner M, En-glmeier KH, Reiser M, Putz R. Quantitative relationships of normalcartilage volumes of the human knee joint: assessment by MRI. AnatEmbryol 1998;197:383–390.

12. Dupuy DE, Spillane RM, Rosol MD, Rosenthal DI, Palmer WE, BurkeDW, Rosenberg AE. Quantification of articular cartilage in the knee withthree-dimensional magnetic resonance imaging. Academic Radiol 1996;11:919–924.

13. Piplani MA, Disler DG, McCauley TR, Holmes TJ, Cousins JP. Articularcartilage volume in the knee: semiautomated determination fromthree-dimensional reformations of MR images. Radiology 1996;198:855–859.

14. Kshirsagar AA, Watson PJ, Tyler JA, Hall LD. Measurement of localisedcartilage volume and thickness of human knee joints by computeranalysis of three-dimensional MRI. Invest Radiol 1998;33:289–299.

15. Tebben PJ, Pope TW, Hinson G, Batnitzky S, Wetzel LH, DePaolis DC,Munns S, Wilkinson SB, Gordon MA. Three dimensional computerizedreconstruction: illustration of incremental articular cartilage thinning.Invest Radiol 1997;32:475–484.

16. Foster JE, Dieppe PA, Maciewicz RA, Taberner J, Watt I, Waterton JC.Quantification of cartilage volume and visualisation of defects inosteoarthritis using a clinical MR system. Arthr Rheum 1996;39:s170.

17. Dieppe PA, Cushnaghan J, Shepstone L. The Bristol ‘OA500’ study:progression of osteoarthritis over three years and the relationshipbetween clinical and radiographic changes at the knee joint. Osteoarthri-tis Cartilage 1997;6:87–97.

18. Krag MH, Cohen MC, Haugh LD, Pope MH. Body height change duringupright and recumbent posture. Spine 1990;15:202–207.

19. Solloway S, Hutchinson CE, Waterton JC, Taylor CJ. The use of activeshape models for making thickness measurements of articular cartilagefrom MR images. Magn Reson Med 1997;37:943–952.

20. Elliott PJ, Diedrichsen J, Goodson K, Riste-Smith R, Sivewright GJ. Anobject oriented system for 3D medical image analysis. IBM Systems J1996;35:1–18.

21. Sivewright GJ, Elliott PJ. Interactive region and volume growing forsegmenting volumes in MR and CT images. Medical Inf (Lond) 1994;19:71–80.

22. Shapiro B, Sklansky J. Skeleton generation from x,y boundary se-quences. Comput Vision Graphics Image Processing 1981;15, 136–153.

23. Eckstein F, Sittek H, Milz S, Schulte E, Kiefer B, Reiser M, Putz R. Thepotential of MRI for quantifying articular cartilage thickness—a method-ological study. Clin Biomech 1995;10:434–440.

24. Eckstein F, Sittek H, Gavazzeni A, Schulte E, Milz S, Kiefer B, Reiser M,Putz R. Magnetic resonance chondro-crassometry (MR CCM): a methodfor accurate determination of articular cartilage thickness? Magn ResonMed 1996;35:89–96.

25. Losch A, Eckstein F, Haubner M, Englmeier KH. A non-invasivetechnique for three-dimensional assessment on articular cartilage thick-ness based on MRI. Part 1: development of a computational method.Magn Reson Imaging 1997;15:795–804.

26. Tieschky M, Faber S, Haubner M, Kolem H, Schulte E, Englmeier KH,Reiser M, Eckstein F. Repeatability of patellar cartilage thicknesspatterns in the living, using a fat-suppressed MRI sequence with shortacquisition time and 3D data processing. J Orthop Res 1997;15:808–813.

27. Eckstein F, Gavezzeni G, Sittek H, Haubner, Losch A, Milz S, EnglmeierKH, Schulte E, Putz R, Reiser M. Determination of knee joint cartilagethickness using three-dimensional magnetic resonance chondrocrassom-etry (3D MR-CCM). Magn Reson Med 1996;36:256–265.

28. Eckstein F, Tieschky M, Faber SC, Haubner M, Kolem H, Englmeier KH,Reiser M. Effect of physical exercise on cartilage volume and thicknessin vivo: an MRI study. Radiology 1998;207:243–248.

29. Trattnig S, Breitenseher MJ, Huber M, Zettl R, Rottman B, Haller J,Imhof H. Bestimmung der Knorpeldicke am oberen Sprunggelenk: eineMRT (1.5T)-anatomische Vergleichsstudie. Fortschr Rontgenstr 1997;166:303–306.

30. Winer BJ. Statistical principles in experimental design. London: Mc-Graw-Hill; 1971. 907 p.

31. Joliffe IT. Principal component analysis. New York: Springer-Verlag;1986. 271 p.

32. Cova M, Frezza F, Shariat-Razavi I, Ukmar M, Pozzi Mucelli RS, DallaPalma L. Valutazione con risonanza magnetica degli aspetti dellacartilagini ialine articolari del ginocchio in funzione dell’eta, del sessoe del peso corporeo. Radiol Med 1996;92:171–179.

33. Jayson MIV. Intra-articular pressure. Clin Rheum Dis 1981;7:149–166.34. Andresen R, Radmer S, Konig H, Wolf KJ. MRT-spezifishe Einteilung

der Chondromalacia patellae unter Zuhilfename eines speziellenKniekompressors: Gegenuberstellung mit dem arthroskopichen Be-fund. Fortschr Rontgenstr 1993;159:541–547.

35. Herberhold C, Stammberger T, Faber S, Putz R, Englmeier KH, Reiser M,Eckstein F. An MR-based technique for quantifying the deformation ofthe articular cartilage during mechanical loading in an intact cadaverjoint. Magn Reson Med 1998;39:843–850.

36. Botsford DJ, Esses SI, Ogilvie-Harris DJ. In vivo diurnal variation inintervertebral disk volume and morphology. Spine 1994;19:935–940.

37. Rubenstein JD, Kim JK, Henkelman RM. Effects of compression andrecovery on bovine articular cartilage: appearance on MRI. Radiology1996;201:843–850.

38. Haubner M, Eckstein F, Schnier M, Losch A, Sittek A, Becker C, KolemH, Reiser M, Englmeier KH. A non-invasive technique for 3-dimen-sional assessment of articular cartilage thickness based on MRI. Part 2:validation using CT arthrography. Magn Reson Imaging 1997;15:805–813.

39. McGibbon CA, Dupuy DE, Palmer WE, Krebs DE. Cartilage and subchon-dral bone thickness distribution with MRI. Acad Radiol 1998;5:20–25.

40. Eckstein F, Adam C, Sittek H, Becker C, Milz S, Schulte E, Reiser M,Putz R. Non-invasive determination of cartilage thickness throughoutjoint surfaces using magnetic resonance imaging. J Biomech 1997;30:285–289.

132 Waterton et al.