Research Article: Automated assessment of bone changes in cross-sectional micro-CT studies of murine experimental osteoarthritis

Date Published: March 23, 2017

Publisher: Public Library of Science

Author(s): Patricia Das Neves Borges, Tonia L. Vincent, Massimo Marenzana, Alejandro A. Espinoza Orías.


The degradation of articular cartilage, which characterises osteoarthritis (OA), is usually paired with excessive bone remodelling, including subchondral bone sclerosis, cysts, and osteophyte formation. Experimental models of OA are widely used to investigate pathogenesis, yet few validated methodologies for assessing periarticular bone morphology exist and quantitative measurements are limited by manual segmentation of micro-CT scans. The aim of this work was to chart the temporal changes in periarticular bone in murine OA by novel, automated micro-CT methods.

OA was induced by destabilisation of the medial meniscus (DMM) in 10-week old male mice and disease assessed cross-sectionally from 1- to 20-weeks post-surgery. A novel approach was developed to automatically segment subchondral bone compartments into plate and trabecular bone in micro-CT scans of tibial epiphyses. Osteophyte volume, as assessed by shape differences using 3D image registration, and by measuring total epiphyseal volume was performed.

Significant linear and volumetric structural modifications in subchondral bone compartments and osteophytes were measured from 4-weeks post-surgery and showed progressive changes at all time points; by 20 weeks, medial subchondral bone plate thickness increased by 160±19.5 μm and the medial osteophyte grew by 0.124±0.028 μm3. Excellent agreement was found when automated measurements were compared with manual assessments.

Our automated methods for assessing bone changes in murine periarticular bone are rapid, quantitative, and highly accurate, and promise to be a useful tool in future preclinical studies of OA progression and treatment. The current approaches were developed specifically for cross-sectional micro-CT studies but could be applied to longitudinal studies.

Partial Text

Osteoarthritis (OA) is the most prevalent joint disease; an incurable and painful condition that is a leading cause of disability worldwide. Although cartilage degradation is a hallmark of disease, OA is regarded as an ‘organ failure’, involving the whole joint [1, 2]. Many changes occur in bone, including attrition, sclerosis, formation of osteophytes, cysts, and marrow lesions [1, 3]. Excessive bone remodelling has been linked to cartilage degeneration [4, 5] and pain [6] from early on in disease [7], but the nature of the relationship between both tissues and how lesions progress over time remains unclear [8, 9]. This is partly because cartilage loss frequently progresses prior to development of symptoms and partly because available tools are insensitive and do not permit early diagnosis [10]. In clinic, disease progression is mostly assessed by radiographic scoring using semi-quantitative systems [11–13] once bone changes are well-established, but this evaluation lacks the sensitivity to track temporal changes [14]. Furthermore, tissue is usually only available at the late stages of disease, keeping early events poorly understood. Therefore, experimental models hold a key role, not only to help understand pathogenesis and to chart temporal changes in bone and cartilage, but also to develop strategies for early detection and therapeutic targeting [10]. The mouse model, largely due to its being amenable to genetic modifications, is widely used in research. In the recent years, micro-computed tomography (micro-CT) has become the gold-standard imaging modality for bone assessment in this model, owing to excellent resolution, 3D capability, and utility in longitudinal studies [15]. However, micro-CT lacks validated methodologies for automated analysis of the epiphyseal subchondral bone, and mostly for structure segmentation, which is frequently based on manual contouring of regions-of-interest [16–18]. To increase segmentation throughput, automated approaches have been proposed; mostly for compartmentalisation of cortical and trabecular bone [19–22]. While useful, these methods often rely on thresholds, based on the premise that cortical and trabecular bone can be differentiated by their different grey level intensities. The choice of an appropriate threshold is, however, critical for accurate segmentation and minor changes may cause errors [22], leading to mis-estimation of structural parameters [15]. Accuracy on compartmentalisation can be improved by combining thresholding methods and microstructural criteria, such as thickness differences between cortical and trabecular bone [23, 24]. Additionally, despite osteophytes being a well-established feature of osteoarthritic joints, there seems to be a lack of validated methods for measuring these bony structures. Assessment in both clinical [12] and experimental models [25] has been often limited to semi-quantitative grading based on their size and maturity, but micro-CT scans have been shown to have the potential to provide volumetric measurements of osteophytes [26].

In this work, we present an automated epiphyseal analysis for characterization of bone changes in experimental OA. Although we confined joint analysis to tibiae, histology has shown that lesions in this model are prevalent in the medial tibial epiphysis [27] and consequently, assessment in tibiae should provide a representative measure of OA changes in the whole joint. On the tibial surface, we generated planar heat maps of subchondral bone thickness to locate the areas of major alterations. From these data, progressive sclerosis was exclusively observed in the medial compartment of operated joints. This supported previous evidence that lesions are time-dependent and confined to load-bearing areas or sites of trauma [38]. Nevertheless, it has been suggested that the cortical plate and trabecular bone have different involvement in OA [39] and should be evaluated separately [40]. Segmentation of cortical and trabecular bone in the tibial epiphysis has been previously performed by manual contouring [16–18], which is time-consuming and subjective. Automated approaches for compartmentalisation proposed in literature rely mostly on thresholds [19–23], a critical parameter in micro-CT segmentation, and are prone to error [15]. For increased robustness, segmentation methods should also consider microstructural criteria, such as thickness differences between cortical and trabecular bone [23, 24]. Our method relies on the degree of macro-porosity, used to classify bone structure at the macroscopic level [32], to distinguish between cortical and trabecular bone and therefore, has the strength of being threshold independent in the differentiation between compartments. We confined volumes-of-interest to the load bearing regions of the tibial plateau where consistent increases in subchondral bone thickness were observed by our planar heat maps. To improve the robustness of the measurements, we applied 3D registration to co-align pairs of DMM/contralateral tibiae prior to analysis. This methodology has demonstrated accurate longitudinal monitoring of bone resorption and apposition [41] and improved sensitivity and reproducibility of micro-architectural quantifications [42, 43]. In our study, it reduced the variability of the measurements, suggesting that the robustness of the method was also improved. Furthermore, we used the paired alignment to determine shape differences between DMM and contralateral tibiae. Shape comparisons have been previously proposed as a potential imaging biomarker of pathology progression in clinical studies [44–46], while in experimental models, and also using alignment by 3D registration, they have demonstrated the ability to track and quantify osteophytes [47]. In our method, we considered that outer shape and size differences between DMM and contralateral tibiae were caused by osteophytes, a valid assumption when comparing pairs of bones from the same animal. Not only did this method demonstrate high accuracy in detecting osteophytes (from 2-weeks post-surgery), but also demonstrated robustness by the strong agreement with manual segmentation and histology. We further proposed epiphyseal volume as a surrogate measurement of osteophyte growth, since osteophytes broaden the articular surface [48]. Early epiphyseal expansion (from 4-weeks post-surgery) was detected by this semi-automated method, thus demonstrating its potential for early osteophyte measurement.

The methods described in the current report provide a robust, high-throughput, platform for the quantitative assessment of epiphyseal bone changes in experimental OA. In our cross-sectional study, we demonstrated that progressive structural changes in subchondral bone and osteophyte formation are detectable as early as 4-weeks after DMM surgery, despite articular cartilage score did not show significant changes until 12 weeks post-DMM surgery. Our automated quantitative imaging methods will allow to sensitively and rapidly measure disease progression on periarticular bone in a range of experimental models of murine OA, with the possibility of extending its use to longitudinal studies using in vivo micro-CT scanners.




0 0 vote
Article Rating
Notify of
Inline Feedbacks
View all comments