C184 Poster Presentations

a selective evaluation of the outer layer of the membrane wasdone. The ROIs were the lateral and medial suprapatellar bursaand the lateral and medial recess. A grading system was alsodeveloped: at each ROI, synovial thickness severity was gradedfrom 0 to 3, giving a global knee score of 0-12.A validation study was carried out with a cohort of 27 knee OApatients. MRIs were evaluated twice at two-week intervals, withthe randomization list changed each time.Results: The MRI sequence reading included a localization ofthe synovial liquid, in the images of a fast multiplanar spoiledgradient-recalled acquisition in the steady-state T2-weighted se-quence, allowing the localization and measurement of the syn-ovial membrane at the corresponding location in the images ofa GRE sequence. The acquisitions were performed in an axialplane without contrast agent.Evaluation of the synovial membrane thickness for the 27 OApatients showed a grade of 1.9±0.5 (mean±SD) with a scoreof 7.1±2.3, and an intrareader coefficient of 0.92 (p<0.001). Onthe medial compartment the grade was 1.9±0.6 and the score3.4±1.4, and on the lateral compartment, 2.0±0.7 and 3.7±1.5,respectively. The grade and score for the upper (suprapatellarbursa) and lower (recess) ROIs were 1.8±0.7 and 3.5±1.5, and2.1±0.5 and 3.9±0.9, respectively. No statistically significantdifferences in grade and score were observed between themedial and lateral compartments or between the upper andlower ROIs, indicating the homogeneity in the synovial membranethickness of the studied ROIs.Conclusions: Our newly developed method to assess synovialmembrane thickness was accurate and reproducible for the as-sessment of knee OA synovitis using a noninvasive MRI tech-nology. This system will be very useful to evaluate and follow theresponse of arthritis patients to treatment with anti-inflammatoryand disease modifying OA (DMOAD) drugs.

329

RELIABILITY OF LOWER LIMB FRONTAL PLANEALIGNMENT MEASURES OBTAINED WITH THE USE OFA COMPUTER PROGRAM AND ELECTRONIC TOOLS

E.A. Sled1, D. Cooke1, L. Sheehy1, M. Lam1, P. Costigan1,M. Nevitt2, J.C. Torner3, C.E. Lewis4, L. Sharma5, D. Felson6

1Queen’s University, Kingston, ON, Canada; 2UCSF, SanFrancisco, CA; 3The University of Iowa College of Public Health,Iowa City, IA; 4University of Alabama, Birmingham, AL;5Northwestern University, Chicago, IL; 6Boston University,Boston, MA

Purpose: This study determined the inter-reader and intra-reader reliability of lower limb frontal plane alignment measuresobtained from digital radiographs using a computer softwareprogram.Methods: Measurements of lower limb frontal plane alignmentwere obtained from over 3000 full limb digital radiographs ofboth limbs of persons with or at risk for developing knee os-teoarthritis (OA), as part of the Multicenter Osteoarthritis Study(MOST). Three trained clinicians used a computer software pro-gram (Horizon Image Viewer, version 1.5, OAISYS Medical Inc.)to locate bone landmarks on the femur and tibia from whichstandard measures of alignment [e.g. the Hip-Knee-Ankle (HKA)angle] could be computed.To assess the reliability of these alignment measurements, 100randomly chosen digital radiographs, representing 200 limbs,were selected from the complete data set for a repeated analysiscarried out two or more weeks after completion of the firstmeasurements.Random effects 2-way analysis of variance (ANOVA) modelswere applied to estimate the interclass and intraclass correlationcoefficients (ICC), which correspondingly evaluated inter-reader

and intra-reader reliability for each of the angles and bonelengths.Results: High reliability measures were obtained for the HKAangle [inter-reader reliability: ICC=0.995 (95% CI, 0.994-1); intra-reader reliability: ICC= 0.998 (95% CI, 0.998-1)]. Reliability foradditional angles between the femur and tibia ranged from 0.839to 0.993 (inter-reader reliability) and 0.908 to 0.998 (intra-readerreliability). High reliability measures were also obtained for bonelengths (inter-reader reliability: ICC from 0.993 to 0.995; intra-reader reliability: ICC from 0.994 to 0.995).Conclusions: Each of the lower limb alignment and bone lengthmeasurements were highly reliable. The outcome supports theuse of computer software programs and software tools for anal-ysis of lower limb frontal plane alignment.

330

MRI OF ARTICULAR CARTILAGE ABNORMALITIES OVERFAR POSTERIOR FEMORAL CONDYLE OF THE KNEE

S. Ogino, T. Huang, P. Lang, H. YoshiokaBrigham and Women’s Hospital, Boston, MA

Purpose: To review MR imaging showing articular cartilage ab-normalities along the far posterior portion of the femoral condylesof the knee joint and investigate their prevalence and associatedbone morrow edema (BME) and knee derangements.Methods: 883 consecutive knee MR examinations from Januaryto April 2007 in our institution were reviewed retrospectively.We excluded studies with the following characteristics: differentprotocol images, prior knee arthroplasty, rheumatoid arthritis,and pigmented villonodular synovitis. 654 patients were finallyincluded in this study. All MR images were obtained with a 1.5T (GE Healthcare). Sagittal fast spin-echo (FSE) proton density-weighted images (PDWI) (mean repetition time [TR] msec/echotime [TE] msec, 2690/35) and sagittal FSE fat-suppressed PDWI(mean TR msec/TE msec, 2970/24) were acquired with an imagematrix of 512 x 512, a 3.5-mm slice thickness, and a 14- to 16-cmfield of view. MR images were evaluated by two readers by con-sensus. The following factors were assessed; (1) the prevalenceof cartilage abnormalities, (2) laterality (medial or lateral), (3)the type of cartilage abnormalities (focal or diffuse), (4) cartilageabnormality grading (signal heterogeneity, fraying, fissuring, lessthan 50% cartilage thinning, more than 50% cartilage thinning,and full thickness cartilage defect), and (5) associated BMEand complications such as meniscal injury and cruciate ligamentinjury.Results: In 654 cases, 172 cases (26.3%) (65 male and 107female; 87 right and 85 left knee; age range from 17 to 94 years,mean age 52.1 years) showed articular cartilage abnormalities atthe far posterior femoral condyle. Cartilage abnormalities alongthe lateral femoral condyle (LFC) were observed in 153 cases(23.4%) and those along the medial femoral condyle (MFC) in 61cases (9.3%). BME was demonstrated in 52 cases (7.9%) in theLFC and in 26 cases (3.9%) in the MFC. Along the LFC, focaland diffuse cartilage abnormalities were identified in 128 cases(19.6%) and 25 cases (3.8%), respectively. On the other hand,along the MFC, focal and diffuse cartilage abnormalities wereidentified in 24 cases (3.7%) and 37 cases (5.7%), respectively.Meniscus injuries were seen in 125 cases (medial meniscusinjury: 61 cases, lateral meniscus injury: 37 cases, both medialand lateral meniscal injury: 27 cases); anterior cruciate ligamentinjuries were seen in 20 cases.Conclusions: The present study shows that far posterior ar-ticular cartilage abnormalities over the LFC were 2.5 times asfrequent as those over the MFC. The cartilage abnormalities overthe LFC tend to be more focal, while those over the MFC tendto be more diffuse. Accompanying medial or lateral meniscalinjuries were seen at high rates. The high prevalence of focal

Osteoarthritis and Cartilage Vol. 15, Supplement C C185

cartilage abnormalities over the LFC may be due to the shallowand small lateral tibial plateau. As the knee is deeply flexed,the shallow posterior superior surface of the lateral tibial plateaucould impinge on the far posterior portion of the femoral condyleearlier on the lateral side than on the medial side. Far posteriorarticular cartilage abnormalities may be correlated with differentknee movement and morphology between the medial and lateraltibiofemoral compartments.

331

FULLY AUTOMATIC QUANTIFICATION OF KNEEOSTEOARTHRITIS SEVERITY ON STANDARDRADIOGRAPHS BY A NOVEL COMPUTER ASSISTEDSYSTEM

H. Oka, N. Yoshimura, S. Muraki, A. Mabuchi, K. Nakamura,H. Kawaguchi22nd Century Medical Center, and Sensory & Motor SystemMedicine, University of Tokyo, Tokyo, Japan

Purpose: Although radiographs are still considered the goldenstandard to evaluate the structural OA severity in clinical practice,the conventional categorical grading by the Kellgren/Lawrence(K/L) system or the OARSI radiographic atlas remains limited inreproducibility and reliability. Here we report a novel computerprogram called KOACAD (knee OA computer-aided diagnosis)which for the first time realized a fully automatic measurement ofthe six major parameters of knee OA on standard radiographs:medial & lateral minimum JSW (mJSW), medial & lateral jointspace area (JSA), osteophyte area (OPA), and tibiofemoral angle(TFA). This study examined the reproducibility and reliabilityof KOACAD, and further investigated the association of theparameters with knee pain using a large-scale database of theROAD (Research on Osteoarthritis Against Disability) study.Methods: The KOACAD was programmed to perform the follow-ing operations automatically on AP knee radiographs obtainedas a DICOM file. The medial and lateral JSAs were determinedas the area between the outlines of femoral condyle and tibialplateau. The minimum axial distances in the areas were mea-sured as medial and lateral JSWs. OPA and TFA were alsocalculated from the outlines of femur and tibia. The reproducibil-ity of the KOACAD parameters with various knee flexion angles(0-30 degrees) and X-ray beam angulations (0-15 degrees) wasevaluated by an intraclass coefficient of correlation (ICC) on ra-diographs of 20 individuals taken with a 2-week interval. Inter-and intraobserver variabilities were determined by κ value or ICCon 50 randomly selected radiographs. Correlations of sex, age,and the KOACAD parameters with knee pain were examined on1,979 radiographs by a stepwise logistic regression analysis.Results: Since the reproducibility of all KOACAD parameterswas highly maintained with 0 degree knee flexion and 0-5 de-gree X-ray beam angulation (ICC=0.88-0.99), it was decided totake radiographs with the knee extended and a horizontal X-raybeam. Although inter- and intraobserver variabilities were highby K/L grading (κ=0.76 and 0.84), OARSI grading (κ<0.72 and<0.75 in all parameters), and semi-automatic measurements onradiographs (ICC<0.65 and <0.75 in all parameters), the KOA-CAD system automatically measured the six parameters in lessthan one second without any variability (all ICC=1.0). Amongthe KOACAD parameters, medial mJSW was most stronglycorrelated with K/L grading (r=-0.41) and OARSI grading (r=-0.45). In the KOACAD parameters, OPA was associated withnone of the joint space parameters, suggesting different etio-logic mechanisms between osteophyte formation and joint de-struction. Logistic regression analysis after adjustment for agerevealed that low medial mJSW (OR=1.46, 95%CI=1.16-1.90in men; OR=1.41, 95%CI=1.22-1.63 in women) and high TFA(OR=1.07, 95%CI=1.01-1.13 in men; OR=1.07, 95%CI=1.03-

1.10 in women) were positively associated with the presence ofpain (n=594/1,979), while neither lateral mJSW nor OPA was.Conclusions: We have established a fully automatic computer-assisted program KOACAD to quantify the major features ofknee OA on standard radiographs with high reproducibility andreliability. Using this system which can evaluate distinct OAparameters, medial joint space narrowing and varus deformity ofknee joints were shown to be risk factors for symptomatic kneepain. This system will not only be useful for objective evaluationof the disease severity in daily clinical practice, but also act asa proper surrogate measure for the development of disease-modifying drugs for OA, just as bone mineral density does inosteoporosis.

332

STUDY OF LOCATION SPECIFIC RADIOGRAPHIC JOINTSPACE WIDTH FOR KNEE OSTEOARTHRITISPROGRESSION: ANALYSIS OF LONGITUDINAL DATAFROM THE OSTEOARTHRITIS INITIATIVE (OAI)

J. Duryea1, D.J. Hunter2, L.B. Chibnik1, C. Dabrowski3,M-P. Hellio Le Graverand-Gastineau4, M. Luchi5, C.R. Beals6

1Brigham and Women’s Hospital, Boston, MA; 2New EnglandBaptist Hospital, Boston, MA; 3GlaxoSmithKline PharmaceuticalInc., Collegeville, PA; 4Pfizer Inc., Ann Arbor, MI; 5Novartis Inc.,East Hanover, NJ; 6Merck Inc., Rahway, NJ

Purpose: To study the performance of location-specific radio-graphic joint space width (JSW). To facilitate this we madecomparisons to minimum JSW (mJSW) and examined the de-pendence of JSW on the Kellgren and Lawrence (K & L) gradeand varus-valgus alignment anatomical angle.Methods: Baseline and Year 1 knee radiographs of 160 subjectsfrom the Progression Cohort of the Osteoarthritis Initiative (OAI)were analyzed using a software technique that measured theradiographic joint space width (JSW). Bilateral knee radiographswere acquired using a fixed flexion protocol designed to maintainconsistent knee flexion and beam angle for all visits. The varus-valgus anatomical alignment angle and the K & L grades weremeasured for a single indexed knee at baseline for each subject.Measurements of medial compartment mJSW and JSW at fixedlocations were facilitated by the use of automated software thatdelineated the femoral and tibial margins of the joint. mJSWwas defined as the minimum distance between the delineatedfemur and tibia margins. Measures of JSW were defined asthe distance from the tibial margin to the femur margin at fixedlocations on the coordinate system shown in Figure 1. JSW wasmeasured at four fixed locations (x = 0.2, x = 0.225, x = 0.25,and x = 0.275) (Figure 2)

Figure 1 Figure 2

A human reader operated custom software to verify and correctthe software-drawn margins where necessary. Paired imageswere displayed with the reader blinded to the time point. As ametric to quantify performance we used the accuracy of eachmethod to correctly order the time point of the films. This wasbased on the reasonable assumption that structural changes inOA should not improve over time.