3D contour registration reveals that location-independent cartilage thickness scores and regional patterns differ between early and late follow-up periods after anterior cruciate ligament (ACL) rupture

Purpose: Anterior cruciate ligament (ACL) rupture is associated with acute joint trauma, chronically altered joint mechanics; and subsequently with an increased risk of incident knee osteoarthritis (OA). It has been reported that the mean femorotibial cartilage thickness increases in young adults after ACL rupture, with the annual rate during the first 2 years (BL→Y2) being similar to that during the subsequent 3 years (Y2→Y5). Previous work has shown that this mean cartilage thickness change does not reflect simultaneous thinning and thickening that occurred in different subregions of the joint, and that simultaneous thinning and thickening actually differed in magnitude between the first and later observation period, independent of treatment. Such cartilage thickness perturbation has been characterized by the location-independent cartilage change score, based on 16 femorotibial subregions across the femorotibial joint. Objectives of the current study were to expand on this by 1) using a 3D registration method for intra-person alignment of baseline and follow-up MRIs based on subchondral bone contours; 2) determining annualized location-independent thinning, thickening and change scores using thickness difference for each voxel after alignment rather than for subregions; 3) to compare these scores between the early (BL→Y2) vs. later (Y2→Y5) follow-up (FU) after acute ACL rupture in each of the 4 femorotibial cartilage plates, and 4) to use inter-person contour registration to determine regional patterns of cartilage thickness change averaged across the cohort and compare the early vs. later follow-up observation period.

Methods: We used gradient-echo MRIs from 121 acutely ACL ruptured knees from a published RCT with existing manual femorotibial cartilage segmentation. To compare thickness changes for each voxel (0.29 x 0.29 x 1.5 mm³), BL and FU images were unimodally registered based on segmented subchondral bone (tAB) contours of the femorotibial cartilages. The segmented BL and FU image datasets were used as input for a self-developed software. First, cartilage thickness values were calculated at each tAB voxel. Secondly, the software extracted the tAB contours of all four femorotibial cartilage plates, then combined the femoral (cMF and cLF) and tibial (MT and LT) cartilage plates, and then converted them into binary image stacks. Next, each femoral and tibial BL image stack was aligned to its corresponding FU image stack, using unimodal rigid registration. Thickness values, and BL→FU differences, were computed for each tAB voxel for MT, cMF, LT, an cLF. The software was developed in C++ using ITK for the registration algorithm. The summed positive changes per voxel (thickening score), the negative ones (thinning score), and the total thickness changes independent of their direction (total change score), were annualized and statistically compared between both observation periods using paired t-tests. Effect sizes between later vs. early follow-up were expressed by the Cohen’s D. To visualize the regional cartilage thickness changes for both observation periods, a combined approach of rigid and deformable registration was used for the alignment of binary BL→FU image stacks. Femoral and tibial registration were performed independently. Left knees were mirrored to be aligned with right knees. The corresponding thickness change values were then averaged for each follow-up tAB voxel.

Results: Statistically significant greater annualized thickening, thinning, and total change scores were observed for the early vs. late FU period in all femorotibial cartilage plates (Table 1). Effect sizes for the thickening score ranged from 0.50 (LT) to 0.85 (FTJ), those for the thinning score from -0.74 for cLF to -1.55 for FTJ, and those for the total change score from 1.33 (cMF) to 1.82 (FTJ). During early FU, the thickening/ thinning ratios were 1.11 for MT, 1.20 for cMF, 1.11 for cLF, but only 0.71 for LT. During the later FU, the ratios amounted to 1.38 for MT, 1.51 for cMF, 1.07 for cLF and 1.14 for LT.

The graphical representation of the regional thickness changes across the cohort displayed cartilage thinning to occur mainly in the posterior aspect of LT within the early observation period (Figure 1), whereas this was not seen during the late observation period (Figure 2).

Conclusions: We developed a novel method for intra- and inter-person alignment of baseline and follow-up MRIs using 3D contour registration, which permits to compare the cartilage thickness change within and between subjects at each voxel of the bone-cartilage interface. The application of this technology to ACL patients suggests greater cartilage perturbation (i.e. simultaneous thinning and thickening) during the first two years after acute injury as opposed to the latter 3 years, and this was observed in each of the femorotibial cartilage plates. However, over the early observation period, cartilage thinning was most prominent in the lateral tibia, whereas cartilage thinning and thickening in the LT were more balanced during the later follow-up period. Interestingly, the posterior LT is the site of greatest mechanical impact during ACL tear as well as the site of post traumatic bone marrow lesions (BMLs), indicating that the cartilage loss occurring in this particular region immediately after the ACL injury may be related to that mechanical impact and the associated BMLs.