Technical Note

Use of 3T MRI and an Unspoiled 3D Fast GradientEcho Sequence for Porcine Knee CartilageVolumetry: Preliminary Findings

Megan S. Cromer, BAppSc,1,2* Sheryl L. Foster, MHSc(MRS),1,2

Roger M. Bourne, PhD,2 Marlene Fransen, PhD,3 Roger Fulton, PhD,2,4,5

and Shih-Chang Wang, BSc(Med)1,6

Purpose: To assess the utility of knee cartilage volumetryusing an unspoiled fat-suppressed 3D fast gradient echo(FGRE) sequence at 3T.

Materials and Methods: Sagittal magnetic resonance(MR) images were obtained with an unspoiled fat-sup-pressed 3D FGRE sequence in eight porcine knees. Man-ual segmentation was used to derive the cartilage volume.This volume was compared to a volume measurement ofcartilage scraping specimens obtained by water displace-ment. Imaging was repeated five times in four ofthe knees to assess interscan volume measurement repro-ducibility and calculate precision error. A single 3Ddataset was manually segmented five times at weeklyintervals to assess intraobserver volume measurementreproducibility.

Results: Total cartilage volume obtained from MRI andwater displacement correlated well (r ¼ 0.75). The inter-scan reproducibility of total volume measurements,expressed as the coefficient of variation (CV), was 4.2%,and the precision error (root mean square [RMS] CV) was4.1%. The CV of intraobserver estimates of total cartilagevolume by MRI was 3.6%.

Conclusion: Interscan reproducibility of quantification oftotal cartilage volume and reproducibility of the manualsegmentation technique were both high (>95%). Accurateand reproducible cartilage volumetry can be obtained byusing a clinical unspoiled fat-suppressed 3D FGREacquired at 3T MRI.

Key Words: 3T; cartilage quantification; magnetic reso-nance imaging; porcine kneeJ. Magn. Reson. Imaging 2013;38:245–250.VC 2012 Wiley Periodicals, Inc.

OSTEOARTHRITIS (OA) is characterized as a multifac-torial process, which involves all the tissues withinthe entire joint (1). Articular cartilage loss, however, isone of the key disease processes in osteoarthritis.Knee OA is highly prevalent within the aging commu-nity (�50 years) worldwide (2). For clinical trials eval-uating potential disease-modifying interventions forknee OA the measure of minimum joint space widthis currently the only structural endpoint accepted byregulatory bodies (3,4). This measurement is obtainedfrom weight-bearing plain radiographs; however, it isindirectly related to cartilage thickness (5), and isthus only a surrogate and relatively late marker ofcartilage loss. It is sensitive to variations in position-ing techniques and is only accurate for the medialcompartment of the knee (6). A further confoundingfactor is that the presence of meniscal pathology hasbeen shown to be a major contributor to change injoint space width measurements (7,8). It has beenestimated that around 13% of the total cartilage vol-ume has been lost before narrowing of the joint spaceis radiographically visible (9).

Magnetic resonance imaging (MRI) has distinctadvantages for knee OA imaging, as it allows for anassessment of the entire joint, is both tomographicand volumetric, and is the imaging modality thatholds the most promise for the advancement of knowl-edge into OA disease processes and for the evaluationof novel treatments (10). Higher field strength systems(3T and above) offer increased signal-to-noise andcontrast-to-noise ratios compared to imaging at 1.5T(11), thus enabling higher spatial resolution imaging.MRI is already widely used to qualitatively image thechanges of OA due to its excellent soft tissue contrastand high spatial resolution (12). MRI directly depictscartilage in vivo and can offer a noninvasive imaging

1Department of Radiology, Westmead Hospital, Sydney, Australia.2Discipline of Medical Radiation Sciences, Faculty of Health Sciences,University of Sydney, Sydney, Australia.3Clinical and Rehabilitation Sciences Research Group, Faculty ofHealth Sciences, University of Sydney, Sydney, Australia.4Department of Medical Physics, Westmead Hospital, Sydney,Australia.5School of Physics, University of Sydney, Sydney, Australia.6Discipline of Imaging, Sydney Medical School, Sydney, Australia.

*Address reprint requests to: M.C., Westmead Hospital, Cnr Darcy &Hawkesbury Roads, Westmead NSW 2145, Australia.E-mail: megan.cromer@swahs.health.nsw.gov.au

Received January 29, 2012; Accepted September 18, 2012.

DOI 10.1002/jmri.23898View this article online at wileyonlinelibrary.com.

JOURNAL OF MAGNETIC RESONANCE IMAGING 38:245–250 (2013)

CME

VC 2012 Wiley Periodicals, Inc. 245

tool to monitor and quantify cartilage loss (13,14).MRI also allows for 3D volume acquisitions in clini-cally viable imaging times of �4–5 minutes per kneejoint.

Thin-section volumetric gradient echo sequenceshave previously been shown to provide accuratedepiction of the articular cartilage due to theincreased contrast-to-noise ratio between the highsignal of the articular cartilage and the adjacent lowsignal structures of the bone and associated soft-tissue structures (15).

Previously, a fat-suppressed T1-weighted 3D spoiledgradient echo (FS T1-W 3D SPGR) sequence has beendescribed as a sequence of choice in quantitativeimaging for volumetric analyses. This type of gradientecho sequence has been shown to be accurate for car-tilage volume quantification when compared to vol-ume estimation by a water displacement method usedin cadavers (15,16) and pigs (11). A similar fastspoiled gradient sequence (FS T1-W 3D FSPGR),which allows for a decreased acquisition time (around6 minutes compared to 12 minutes) has been shownto have comparable accuracy and reproducibility ratesfor cartilage volume quantification compared to thepreviously validated FS T1-W 3D SPGR (16). Thesequence investigated in this study, a fat-suppressed3D fast gradient echo (FS 3D FGRE) without a spoilerpulse, allows for shorter repetition times (TR), result-ing in a further reduction in acquisition times (toaround 4 minutes) applicable for clinical use.

It is paramount that quantitative measurementsfrom longitudinal imaging are accurate, precise, andreproducible if measurements derived from MRI tech-niques are to be used in clinical trials or for clinicalassessments. The aim of this study was to assess thefeasibility of using 3T MRI and an unspoiled fat-sup-pressed 3D FGRE sequence for knee volumetry.

MATERIALS AND METHODS

Eight disarticulated frozen porcine hind limbs (por-cine knees) were obtained from young adult Westranpigs (age 10–12 months) that had been euthanized aspart of an unrelated research study. The knees werethawed overnight in a saline bath. Volume measure-ments of cartilage were obtained by two methods:water displacement of harvested cartilage and manualsegmentation of 3D MRI datasets.

MRI

All imaging was completed in one session. Each por-cine knee was imaged on a clinical 3T Signa ExciteMRI scanner (General Electric Medical Systems, Mil-waukee, WI) with an eight-channel transmit/receiveknee coil by Invivo (Gainesville, FL) The imagingexaminations consisted of a 3-plane localizer, axialcalibration scan, and a sagittal fat-suppressed 3DFGRE dataset (Fig. 1). The 3D datasets (320 � 320matrix [ZIP 512], 2.2 mm slice thickness [ZIP 2], zerogap, TR 7.6 msec, TE 2.3 msec, flip angle 30�, 2 num-ber of excitations [NEX], 160 mm field of view [FOV])

were acquired in 4 minutes 23 seconds. Zerofill inter-polation (ZIP) is a postprocessing technique that isapplied to the acquired raw data in k-space toincrease through-plane resolution. ZIP effectively cre-ates additional images between each acquired sliceresulting in twice as many slices per acquisition, eachwith half the thickness of the acquired images (17).

One 3D dataset was acquired from each of the eightporcine knees. To assess imaging and segmentationreproducibility the imaging sequence was repeatedfive times on four out of the eight porcine knees. Thefour randomly chosen porcine knees were removed,flexed three times, and repositioned between imagingsets to include normal sources of error due to noni-dentical repositioning in longitudinal assessments.

Volume Measurement by Displacement

Cartilage harvest was performed (by two authors) onthe day when imaging was completed by the methodof Dupuy et al (15). The knees were stored in a salinebath to await harvest. All soft tissue was removed toexpose the joint capsule and the joint was then disar-ticulated to expose the cartilage surfaces. The hyalinecartilage was scraped from each of the three surfacesof the joint (patella, tibial plateaus, and femoral con-dyles) with a sharp scalpel. Scrapings were placedinto a graduated cylinder (0.01 mL increments) con-taining a defined volume of saline. The volume of thecartilage was then calculated as the amount of waterthat was displaced. Cartilage volume was measuredseparately for each of the three cartilage surfaces ineach knee. Postharvest photographs were taken ofeach exposed subchondral surface, providing a

Figure 1. Representative sagittal FS 3D FGRE porcine kneeimage. P, Patella; T, Tibia; F, Femur.

246 Cromer et al.

pictorial record that all cartilage had been removed(Fig. 2).

MRI Volume Measurement

Cartilage volume was calculated from the sagittal FS3D FGRE volume datasets using manual segmenta-tion (3D Slicer3, v. 3.4). The segmentation processinvolved defining the cartilage of the medial and lat-eral tibia, medial and lateral femoral condyles, andpatella plates between the bone/cartilage interfaceand cartilage surface on each slice on the sagittal vol-ume dataset. The cartilage borders were manuallytraced and checked on each slice of the dataset. Allcartilage segmentation was performed by one authora radiography researcher experienced in viewing mus-culoskeletal MRI. Segmentation of each dataset took1 and a half to 2 hours.

Statistical Analysis

MRI and displacement volumes were compared usinglinear regression analysis. Reproducibility of volumemeasurement by MRI was assessed with segmentationof the repeated datasets (five repeat acquisitions)acquired on four of the porcine knees. Diagnosticreproducibility was quantified as the precision errorcalculated as the root mean square (RMS) of coeffi-

cient of variation (CV) for the repeated measures, asdescribed by Gluer et al (18). To assess intraobserverreproducibility (19) of the manual segmentationprocess, a single dataset was segmented five times at1-week intervals by one author.

RESULTS

Correlation Analysis

A summary of cartilage volume measurements calcu-lated from both methods is shown in Table 1. Thetotal cartilage volume measured by water displace-ment was strongly correlated (r ¼ 0.75) with MRI vol-ume. The agreement between the two measurementmethods is displayed in a Bland–Altman plot (Fig. 3).The MRI cartilage volumes ranged from 0.5–1.2 mL inthe patellar, 0.7–1.2 mL in the tibial, and 2.3–3.0 mLin the femoral cartilage plates. Displacement volumemeasurements had a similar range: 0.5–1.1 mL in thepatellar, 0.9–1.5 mL in the tibial, and 2.0–3.0 mL inthe femoral cartilage plates. The mean (6SD) of theabsolute differences between the measurements was0.33 6 0.18 mL. A two-tailed Student’s t-test showedthere was no significant difference between the mea-surement methods for total cartilage volume (P ¼0.28) (20).

Figure 2. Photograph of one porcine joint post cartilage har-vest. P, Patella; T, Tibia; F, Femur.

Table 1

Comparison of Cartilage Volume Measurements

Cartilage volume (mL)

Patellar Tibial Femoral Total

Knee WD MR WD MR WD MR WD MR

1 0.5 0.7 1.2 1.1 3 2.8 4.7 4.6

2 0.5 0.6 0.9 0.7 2 2.4 3.4 3.7

3 0.5 0.7 1.5 1.2 2.7 3.0 4.7 4.9

4 0.5 0.5 1.0 0.9 2.6 2.3 4.1 3.7

5 1.1 1.2 1.0 1.1 2.4 2.9 4.5 5.2

6 0.8 0.8 1.0 0.7 2.4 2.5 4.2 4

7 0.9 0.7 0.9 1.1 2.2 2.5 4 4.3

8 0.8 0.6 0.9 1.2 2.4 2.7 4.1 4.5

WD, volume from water displacement; MR, volume from magnetic

resonance imaging data.

Figure 3. Bland–Altman plot ofmeasurement methods.

Reproducibility of Cartilage Volume Using MRI 247

Interscan Reproducibility

Results for the five repeated MRI measurements offour of the knees are summarized in Table 2. The MRIvolume calculations from each dataset are shown inFig. 4. The differences between the MRI volume anddisplacement volume ranged from 0.1–0.2 mL in thepatellar, 0.1–0.3 mL in the tibial, and 0.1–0.5 mL inthe femoral plates. CV ranged from 4.2% (total vol-ume) to 12% (patellar volume). The precision error forthe total volume was 4.1% (Table 3). The interscanreproducibility for total cartilage volume was greaterthan 95%.

Intraobserver Reproducibility

The results for the resegmentation of the data are pre-sented in Table 4. CVs ranged from 3.5% (tibial vol-ume) to 4.8% (patellar volume). CV for total volumewas 3.6%.

DISCUSSION

Accurate and reliable measurement techniques arerequired for the longitudinal study of OA disease pro-gression. A T1-weighted spoiled gradient sequencehas previously been most commonly used for quanti-tative analyses (21). This article presents data usingan unspoiled gradient echo sequence for quantitativeanalysis. Recently, sequences using a water excitationtechnique have been released for clinical use by somevendors. Xing et al (22), using a model similar to thatused in our experiment, published data comparingthe accuracy and reproducibility of FLASHwe (fastlow-angle shot with water excitation) and MEDICwe(multiple-echo data image combination water excita-tion) sequences for use in quantitative cartilage analy-

ses. However, we chose an unspoiled gradient echoclinical sequence for investigation due to its reasona-ble acquisition time and wide availability across allvendors.

Cartilage volumes obtained via the displacementmethod were similar to the MRI volumes. Althoughthere were differences between the volume measure-ments from the two methods, these differences weresmall (range, 0.1–0.7 mL for total cartilage volume)and there was good correlation between the two meth-ods (r ¼ 0.75) for total cartilage volume.

The volume calculations from the segmented dataon the repeated datasets demonstrated a high degreeof reproducibility. Our CVs are similar to otherreported studies (15). Dupuy et al (15), in quantifyingcadaver cartilage, reported repeated measure reprodu-cibility CVs of 7.0% and 9.7%. The exception, how-ever, was our much higher CV of 12% for patellameasurements. This result may be due to the use ofthe sagittal acquisition plane, even though thin sliceimages were acquired. Sagittal datasets are not theoptimal plane of choice for segmentation of the patellacartilage, as the cartilage is only seen in profile andpartial volume effects can lead to inaccurate segmen-tation (19). Protocol designs for longitudinal studiesshould carefully consider the selection of the appro-priate imaging plane based on which cartilage platesare to be quantified.

We found a precision error of 4.1% (RMS CV) fortotal volume calculations using our techniques. Thisis slightly higher than found by Bauer et al (11), whoexamined measurements of porcine cartilage at 3Tusing two different sequences. These authors reportedprecision errors of 2.7% for fat-suppressed SPGR and4.0% for a water excitation sequence. Their sequencewas run at a longer acquisition time of around 6minutes with higher in-plane spatial resolution of0.19 � 0.38 mm2, with slice thickness of 1.0 mm. Our

Table 2

Pooled Results of Volumes of Repeat Datasets

Cartilage plate Mean (mL) SD (mL) CV (%)

Patellar volume 0.6 0.1 12.0

Tibial volume 1.1 0.1 6.6

Femoral volume 2.8 0.1 4.2

Total volume 4.5 0.2 4.2

Figure 4. Reproducibility chart. Volumes for patellar, tibial,and femoral cartilage plates segmented from the repeatdatasets.

Table 3

Precision Errors for Total and Tibial Cartilage Volume

Total cartilage volume (mL) Tibial volume (mL)

Knee Mean SD CV% Knee Mean SD CV%

1 4.7 0.2 4.6 1 1.2 0.1 6.1

3 4.8 0.2 4.4 3 1.1 0.1 9.0

7 4.2 0.2 4.1 7 1.0 0.1 6.8

8 4.4 0.1 3.0 8 1.3 0.0 3.2

RMS 0.2 4.1 RMS 0.1 6.6

RMS, root mean square of coefficient of variation.

Table 4

Intraobserver Reproducibility of Manual Segmentation Technique

Cartilage plate volume (mL)

Segmentation

timepoint Patellar Tibial Femoral Total

1 0.6 0.8 2.4 3.8

2 0.6 0.7 2.2 3.5

3 0.5 0.8 2.1 3.4

4 0.6 0.8 2.2 3.6

5 0.5 0.8 2.3 3.6

CV (%) 4.8 3.5 4.6 3.6

248 Cromer et al.

sequence was faster at around 4 minutes acquisitiontime, with lower in-plane spatial resolution, 0.31 mm� 0.31 mm (ZIP 512) with an acquired slice thicknessof 2.2 mm with ZIP2 applied. A key benefit of ZIP forclinical studies is that it provides an increase in effec-tive resolution with minimal time penalty. The goal ofusing ZIP in this study was to reduce effectivethrough-plane slice thickness in order to decreasepartial volume effects on measurement accuracy.Marshall et al (23) reported an increase in measure-ment variability with increasing acquisition slicethickness. They reported mean percentage error dif-ferences of 1.7 6 1.3% for 1.0 mm slices to 22.7 6

1.9% for 3.0 mm slices, with 1.5 mm slices 3.5 6

2.5%. Compared to these results, the increased errorseen in the sagittal plane measurements of patellarcartilage in this study suggests that ZIP may not offerthe same degree of benefit in reducing partial volumeerrors as directly acquired thin slice volumes. A fur-ther study could examine whether acquiring an imag-ing volume with thinner prescribed slice thickness(resulting in a longer acquisition time), rather thaninterpolation of the data may allow for greater accu-racy and an improvement in precision. However, thesequence examined in this study was chosen for itsviability as a clinical sequence with a reasonable ac-quisition time, applicable for use in clinical applica-tions and research trials.

Segmentation technique can also be a source ofmeasurement error. Inaccurately defining the cartilageborders during the segmentation process will lead toinexact volume measurements. The intraobserverreproducibility CV in this study was relatively low at3.6%. It is poorer but within a similar range to otherreported studies. Marshall et al (23) reported intraob-server variability (CV%) for inexperienced observers at2.8% (range 1.4%–4.2%) and experienced observers at0.8% (range 1.0–1.8%). Practice and operator trainingwith the software analysis tool would improve the op-erator expertise, thus reducing the variation of themeasurements and time taken for the segmentationprocess. This would ensure that the technique couldbe used in a clinical environment. Operator experi-ence with the manual segmentation technique isessential to ensure longitudinal differences in carti-lage volumetry are due to disease progression and noterrors in segmentation technique. Bauer et al (11)used a semiautomatic segmentation technique whichmay have aided in better cartilage border delineationleading to more accurate results. At this time, fullyautomated segmentation techniques which, in theory,should have minimal variation between measure-ments of the same dataset are not widely available.

Measurement differences found in our study aresmall compared to the expected cartilage loss in OAprogression. Published rates of cartilage loss arevaried, as cartilage loss itself is not a linear process.A meta-analysis of longitudinal OA studies performedby Eckstein et al (24) reported cartilage losses of4%–6% per annum in different compartments of theknee. Raynauld et al (25) studied 32 OA (KellgrenLawrence grade 2–3) patients over 2 years and at 12months reported a global cartilage volume loss of

0.473 mL 6 0.658 mL (472.7 6 658 mm3) (mean 6

SD), which represents a 3.6 6 5.1% change frombaseline volume. In our current study the precisionerror for total volume measurements was 0.18 mL(180 mm3) (RMS SD from repeated measures), whichis smaller than cartilage losses in knee OA patientsreported by Raynauld et al (25).

This study had several limitations. There was only arelatively small sample of eight porcine knees. Thevolume of porcine knee cartilage is considerablysmaller than the volume of adult human knee carti-lage and although porcine anatomy is similar tohuman anatomy, the shape of the cartilage plates isnot an exact match (11). The use of porcine knees didfacilitate the ability to harvest and directly measurethe cartilage and allow for the correlation to the MRI-derived volume estimation. The correlation found inthis study was r ¼ 0.75. Higher correlations werefound in other experiments using volume compari-sons from porcine knees (r ¼ 0.99) (11) and humancadavers (r ¼ 0.99) (15). While great care was takenduring the cartilage harvest process it is possible thatthere was incomplete removal of cartilage or the inclu-sion of some other tissues. In the previous morehighly correlated experiments, Dupuy et al (15)enlisted a bone pathologist to inspect the joints post-harvest to gauge if all cartilage was harvested.

The cartilage plates of porcine knees used in thisstudy were all disease-free. The absolute volume andthickness values of these normal porcine cartilageplates correspond to diseased rather than healthystates in human knees. However, it is important tonote that segmentation of cartilage borders doesbecome more difficult as OA progresses due to emer-gent cartilage surface irregularities and associatedsubchondral bony changes. Consequently, precisionerrors tend to be higher as the absolute volume of themeasured cartilage plates decreases (26,27). Ecksteinet al (26) found larger precision errors in OA patientswith Kellgren Lawrence grade (KLG) 3 than for KLG 0and 2. Glaser et al (27) reported precision errors (fem-oral cartilage volume and thickness) for patients withsevere OA of 3.0%–7.0%.

The interscan reproducibility of quantification oftotal cartilage volume and reproducibility of the man-ual segmentation technique were both high (>95%).This finding demonstrates that the measurementsproduced from the sagittal 3D MRI datasets are bothaccurate and reproducible and that the potentialerrors from repositioning the knee, which may occurin longitudinal assessment of patients, are tolerable.

In conclusion, an unspoiled fat-suppressed 3DFGRE sequence acquired at 3T in combination with amanual segmentation technique can be used to pro-vide accurate and reproducible datasets for knee car-tilage volumetry.

ACKNOWLEDGMENTS

We thank Dr Anthony Peduto for direction with thecartilage harvest, Ms. Su Liew for assisting in thepreparation of the porcine joints for cartilage harvest,

Reproducibility of Cartilage Volume Using MRI 249

and Ms. Kylie Walters for assistance in acquiringimages of the porcine joints.

REFERENCES

1. Conaghan PG, Felson D, Gold G, Lohmander S, Totterman S, Alt-man R. MRI and non-cartilaginous structures in knee osteoar-thritis. Osteoarthritis Cartilage 2006;14(Suppl 1):87–94.

2. Busija L, Bridgett L, Williams SRM, et al. Osteoarthritis. Bail-lieres Best Pract Res Clin Rheumatol 2010;24:757–768.

3. Eckstein F, Mosher T, Hunter D. Imaging of knee osteoarthritis:data beyond the beauty. Curr Opin Rheumatol 2007;19:435–443.

4. Hunter DJ, Le Graverand M-PH, Eckstein F. Radiologic markersof osteoarthritis progression. Curr Opin Rheumatol 2009;21:110–117.

5. Adams JG, McAlindon T, Dimasi M, Carey J, Eustace S. Contri-bution of meniscal extrusion and cartilage loss to joint space nar-rowing in osteoarthritis. Clin Radiol 1999;54:502–506.

6. Buckland-Wright JC, Macfarlane DG, Lynch JA, Jasani MK,Bradshaw CR. Joint space width measures cartilage thickness inosteoarthritis of the knee: high resolution plain film and doublecontrast macroradiographic investigation. Ann Rheum Dis 1995;54:263–268.

7. Hunter DJ, Zhang YQ, Tu X, et al. Change in joint space width:hyaline articular cartilage loss or alteration in meniscus? Arthri-tis Rheum 2006;54:2488–2495.

8. Madan-Sharma R, Kloppenburg M, Kornaat PR, et al. Do MRIfeatures at baseline predict radiographic joint space narrowing inthe medial compartment of the osteoarthritic knee 2 years later?Skeletal Radiol 2008;37:805–811.

9. Jones G, Ding C, Scott F, Glisson M, Cicuttini F. Early radio-graphic osteoarthritis is associated with substantial changes incartilage volume and tibial bone surface area in both males andfemales. Osteoarthritis Cartilage 2004;12:169–174.

10. Guermazi A, Hunter DJ, Roemer FW. Plain radiography and mag-netic resonance imaging diagnostics in osteoarthritis: validatedstaging and scoring. J Bone Joint Surg Am 2009;91(Suppl 1):54–62.

11. Bauer JS, Krause SJ, Ross CJ, et al. Volumetric cartilage meas-urements of porcine knee at 1.5-T and 3.0-T MR imaging: evalua-tion of precision accuracy. Radiology 2006;241:399–406.

12. Jeffrey DR, Watt I. Imaging hyaline cartilage. Br J Radiol 2003;76:777–787.

13. Carballido-Gamio J, Bauer J, Stahl R, et al. Inter-subject com-parison of MRI knee cartilage thickness. Med Image Anal 2008;12:120–135.

14. Burgkart R, Glaser C, Hyhlik-Durr A, Englmeier KH, Reiser M,Eckstein F. Magnetic resonance imaging-based assessment of

cartilage loss in severe osteoarthritis: accuracy, precision, anddiagnostic value. Arthritis Rheum 2001;44:2072–2077.

15. Dupuy DE, Spillane RM, Rosol MS, et al. Quantification of articu-lar cartilage in the knee with three-dimensional MR imaging.Acad Radiol 1996;3:919–924.

16. Cicuttini F, Forbes A, Asbeutah A, Morris K, Stuckey S. Compari-son and reproducibility of fast and conventional spoiled gradient-echo magnetic resonance sequences in the determination of kneecartilage volume. J Orthop Res 2000;18:580–584.

17. GE HDx Users Manual, Revision 2. Milwaukee, WI: GE; 2005.Chapter 10, p 76.

18. Gluer CC, Blake G, Lu Y, Blunt BA, Jergas M, Genant HK. Accu-rate assessment of precision errors: how to measure the reprodu-cibility of bone densitometry techniques. Osteoporos Int 1995;5:262–270.

19. Eckstein F, Charles HC, Buck RJ, et al. Accuracy and precisionof quantitative assessment of cartilage morphology by magneticresonance imaging at 3.0T. Arthritis Rheum 2005;52:3132–3136.

20. VassarStats: Statistical Computation Web Site (Richard Lowry1998–2012). Available at: http://faculty.vassar.edu/lowry/Vas-sarStats.html. Last accessed: July 27, 2010.

21. Eckstein F, Burstein D, Link TM. Quantitative MRI of cartilageand bone: degenerative changes in osteoarthritis. NMR Biomed2006;19:822–854.

22. Xing W, Sheng J, Chen W-h, Tian J-m, Zhang L-r, Wang D-q.Reproducibility and accuracy of quantitative assessment of artic-ular cartilage volume measurements with 3.0 Tesla magnetic res-onance imaging. Chin Med J 2011;124:1251–1256.

23. Marshall KW, Mikulis DJ, Guthrie BM. Quantitation of articularcartilage using magnetic resonance imaging and three-dimen-sional reconstruction. J Orthop Res 1995;13:814–823.

24. Eckstein F, Cicuttini F, Raynauld JP, Waterton JC, Peterfy C.Magnetic resonance imaging (MRI) of articular cartilage in kneeosteoarthritis (OA): morphological assessment. OsteoarthritisCartilage 2006;14(Suppl A):A46–75.

25. Raynauld J-P, Martel-Pelletier J, Berthiaume M-J, et al. Quanti-tative magnetic resonance imaging evaluation of knee osteoarthri-tis progression over two years and correlation with clinicalsymptoms and radiologic changes. Arthritis Rheum 2004;50:476–487.

26. Eckstein F, Buck RJ, Burstein D, et al. Precision of 3.0 Teslaquantitative magnetic resonance imaging of cartilage morphologyin a multicentre clinical trial. Ann Rheum Dis 2008;67:1683–1688.

27. Glaser C, Burgkart R, Kutschera A, Englmeier K-H, Reiser M,Eckstein F. Femoro-tibial cartilage metrics from coronal MRimage data: technique, test-retest reproducibility, and findings inosteoarthritis. Magn Reson Med 2003;50:1229–1236.

250 Cromer et al.