Morphological study of lateral meniscus using statistical shape modelling: a study using data from the osteoarthritis initiative

Abstract

s / Osteoarthritis and Cartilage 22 (2014) S57–S489 S298 sizes, we adjusted the CDI by dividing the raw data by a participant’s body height. Change in height-adjusted CDI was calculated as follow-up minus baseline. To explore the construct validity of the new cartilage quantification method we tested for a linear trend of baseline CDI and change in CDI across higher baseline grades of JSN and KL. Results: The average measurement time was 14 minutes (SD 1⁄4 2) per pair of knees. Knees with greater JSN score (i.e. greater OA severity) had lower mean medial femur, tibia, and tibiofemoral CDIs (i.e. less cartilage), see Table 1. There was generally a lower CDI across increasing KL, with the exception that knees with KL grade 2 had a greater CDI compared to those with KL grade 1 (Table 2). Conclusions: While we only used 18 informative locations, our CDI showed similar results with other studies that used cartilage segmentation. For example, we found relationships between the CDI and the severity of radiographic OA. Our findings also support previous reports that knees with KL 1⁄4 2 may have a greater CDI (i.e. greater cartilage) compared with those with KL 1⁄4 1. Overall, the CDI quantification method is rapid, reliable, and valid in the medial tibiofemoral compartment. The CDI has utility in larger studies 521 MORPHOLOGICAL STUDY OF LATERAL MENISCUS USING STATISTICAL SHAPE MODELLING: A STUDY USING DATA FROM THE OSTEOARTHRITIS INITIATIVE K.Y. Zang, A.E. Kedgley, C.D. Donoghue, D. Rueckert, A.M. Bull. Imperial Coll. London, London, United Kingdom Purpose: The menisci bear 40%-70% of the load across the knee. In healthy subjects, the menisci adapt to cope with force transmission required during daily activities, whereas in pathology, the menisci change their shape and these changes are associated with osteoarthritis (OA). In a rat study (Anonymous,1986), it was shown that training for 12 weeks resulted in a change inmeniscus shape. The aim of this study was to perform 3D morphometric analyses of the lateral meniscus (LM) in order to reveal the significant shape variances, thereby gaining a better understanding of the difference in meniscal shape between control subjects and those with knee OA as a precursor to further diagnostic and prognostic structure-function studies. Methods: 100 sets of MRI images (sagittal 3D water excitation DESS) were randomly selected from the Osteoarthritis Initiative (OAI), 50 sets from each of the non-exposed control and OA subcohorts. These were segmentedmanually to create three dimensional (3D) surfacemodels. A statistical shape modelling technique was used to perform 3D morphometric analyses to reveal the significant shape variations. To compute the statistical shape model (SSM), one LM surface model in the dataset was randomly chosen as the reference segmentation. All other LM surface models were aligned to the coordinate system defined by the reference segmentation using the iterative closest point (ICP) algorithm (Anonymous, 1992). Pointto-point correspondences were then established for each subject to the reference using the multi-resolution free-form deformation (FFD) registration algorithm (Anonymous,1999). A mean model was computed from all the LM surface models and principal component analysis was performed to extract the principal morphological variations (PMVs) of linear combinations of point coordinates. Correlations between the PMVs and anthropometric parameters were tested. Anthropometric parameters that were selected included height, weight and body mass index (BMI). Results: The 1st PMV explains 48.2% of the total variation in the meniscal shape for the sample of OAI controls and OA subjects. It corresponds predominantly to the size of the meniscus (Fig. 1) and was found to be correlated with height (p 1⁄4 0.012, r 1⁄4 0.216). The control and OA subjects are statistically different on the 4th (p < 0.001) and 6th (p < 0.001) PMVs (Fig.2). The 4th PMV appears to correspond to the extrusion. Compared to the control subjects, the LMs of OA subjects tended to extrude laterally. This PMV was also found to be correlated with weight (p 1⁄4 0.016, r 1⁄4 0.239). When the weight of subject increases, the lateral meniscus extrudes laterally. The 6th PMV is likely to correspond to the width of the horn and curvature of the lateral peripheral horn. The LM horns of the OA subjects are widened and the medial aspect of lateral spherical horn is straight when compared to those of the control subjects. No statistically significant correlations were found between the other PMVs and the anthropometric parameters. Conclusion: The morphological variations extracted indicate that OA subjects have extruded LM compared with the control subjects. The correlation between the 4th PMV and weight suggests hat heavier subjects have an increased likelihood of an extruded meniscus and thus might be at increased risk of OA.