CAN REGISTRATION-BASED LOCATION-INDEPENDENT MEASUREMENT INCREASE THE SENSITIVITY TO BETWEEN-GROUP DIFFERENCES IN LONGITUDINAL CHANGE OF LAMINAR CARTILAGE T2?

Abstract

INTRODUCTION

Location-independent measurements of cartilage thinning and thickening were shown to be more sensitive to differences in longitudinal change between groups than location-based measures [1,2]. They remove the link between the magnitude and direction of the change and its location, and hence are sensitive to local changes in the joint, independent of where they occur. Location-independent measures of T2 lengthening and shortening computed from 16 femorotibial subregions have been previously applied to a model of early OA. The model compared 3y T2 change in KLG 0 knees with contralateral (CL) joint space narrowing (JSN) vs that in KLG 0 knees with CL KLG 0 (controls) [3]. In this model, location-independent measures were found to provide similar discrimination between these two groups as location-based measures. However, location-independent measures obtained across all individual voxels in the joint (instead of subregions) have been previously suggested to provide more detailed insights into OA-related cartilage thickness changes [4], but no study previously evaluated the sensitivity of such voxel-based shortening and lengthening scores to differences in change of laminar T2.

OBJECTIVE

To compare the sensitivity of voxel-based location-independent lengthening and shortening T2 scores to between-group differences in longitudinal change vs. the previously established technique of subregion-based location-independent and location-based measures in the above early OA model.

METHODS

Multi-echo spin-echo (MESE) MRIs were acquired at year 1 and 4 in the OAI (3T Trio, Siemens). We studied 39 KLG 0 knees with CL JSN, and 39 matched controls (criteria: same sex pain frequency, similar age (±5y) and BMI (±5kg/m²)) with bilateral KLG 0 [2]. Segmentation of the 4 femorotibial cartilages (medial/lateral tibia: MT/LT and central medial/lateral femoral condyle: cMF/cLF) was performed manually by experienced readers. Laminar T2 was computed for each segmented cartilage voxel and classified as deep or superficial, based on the distance to the cartilage surfaces. Location-based and subregion-based location-independent measures were obtained as described previously [2]. Voxel-based location-independent changes in laminar T2 were derived, summarizing the negative/positive changes across all voxels, for each of the femorotibial cartilages using the voxel-based approach (Fig. 1) These were then summarized across the entire femorotibial joint (FTJ). Location-based, subregion-based location independent, and voxel-based location-independent laminar T2 change was compared between the CL JSN vs. control knees using Cohen's D as a measure of effect size with 95% confidence intervals obtained using boot-strapping.

RESULTS

In the deep layer, location-based longitudinal change in femorotibial T2 revealed a Cohen’s D between both groups of 0.37 [0.04, 0.69]), the subregion-based location independent analysis of 0.33 [0.00, 0.65]), and the voxel-based location-independent analysis of 0.36 [0.04, 0.68]) (Fig. 2). In the superficial layer, only the voxel-based absolute change score was sensitive to differences in longitudinal T2 change between both groups (Cohen’s D: 0.34 [0.02, 0.66]). Figure 3 shows the pattern of voxel-wise differences in superficial layer cMF and cLF T2 change between KLG 0 knees with CL JSN vs. control knees.

CONCLUSION

Effect sizes for the different location-based and location-independent T2 analyses (subregion and voxel-based) were similar for the deep cartilage. However, the new voxel-based method appeared to be also sensitive to between-group differences in T2 change in the superficial cartilage layer, where location-based and subregion-based location-independent measures failed to provide notable discrimination. In addition, the voxel-based technique allows to visualize patterns of differences in change between groups that can inform future analyses focusing on specific regions of interest.