Advancements in Subchondral Bone Biomechanics: Insights from Computed Tomography and Micro-Computed Tomography Imaging in Equine Models

Stress fractures are a common injury affecting the integrity of the SCB. In athletic horses, these fractures typically occur due to excessive repetition of normal loads of galloping, on either well adapted bone, or in some instances poorly adapted bone where the SCB has not been conditioned to the applied loading [13, 27]. Particularly, in racehorses, palmar/plantar osteochondral disease (POD) and third metacarpal/tarsal condylar fractures are considered fatigue injuries of SCB and calcified cartilage [28]. The excessive repetition of load can lead to micro-damage that accumulates over time, progressing to stress fractures or focal SCB injuries if the rate of accumulation of damage exceeds the rate of damage repair by the bone remodelling process [22, 29]. Many equine stress fractures arise from articular surfaces and pre-existing pathology is often observed in the SCB at sites where the fractures have propagated from. Early and accurate detection of these lesions in the joint will help to understand the mechanisms of injury as well as to assess injury risk.

Another joint disease that is directly associated with SCB integrity is OA, a degenerative joint disease characterized by the breakdown of cartilage and changes in the SCB, leading to pain and loss of mobility. SCB alterations play a critical role in both the initiation and progression of OA [30, 31]. Changes in the SCB, such as increased turnover, altered mineralization, and microarchitectural deterioration, often precede and predict articular cartilage degradation. Similarly, in early stages of POD in racehorses, SCB abnormalities occur whilst the articular cartilage, remains grossly intact (Turlo et al., 2022). These early changes in the SCB structure and microstructure, highlight the vital role of identifying and measuring these changes both in vivo and in vitro for clinical and research purposes.

CT is an imaging technique that captures cross-sectional images using X-ray photon attenuation. As a transformative tool in SCB research and veterinary diagnostics, CT imaging enables the differentiation of bone and soft tissue in three-dimensions, as well as identifying subtle SCB structural changes and BMD variations, including lesions and complex fractures not visible on radiographs [32, 33, 34, 35]. For example, POD lesions in the fetlock joints of racehorses, manifest as areas of reduced density on radiographs and CT scans, surrounded by trabecular bone with increased density [36, 37, 38]. Additionally, CT imaging offers a thorough assessment of fracture characteristics and associated SCB pathology, aiding in preoperative evaluation and prognostication [36]. Table 1 provides a comparison of the different CT methods used in the study of SCB biomechanics. It outlines typical use-cases, advantages, limitations, and relevant references for each method.

One of the most significant advancements in CT technology for equine limb imaging is standing CT. This innovation has transformed the practical application of CT in both clinical and research settings by allowing large animals to be scanned without the need for general anaesthesia [39, 40, 41]. Using a helical fan beam CT (FBCT) system, standing CT systems proved effective for diagnosing lameness in horses due to their ease of operation and quick diagnostic capabilities [39]. The system was able to safely scan sedated standing horses from the carpal or tarsal region distally, enabling it to be incorporating into lameness investigations, achieving clinical diagnoses for distal limb lameness.

Most recently, the effectiveness of standing cone-beam CT (CBCT) systems in equine SCB diagnosis and research has been explored. In conventional FBCT, the X-ray source and a curved detector rotate 360 degrees around the subject, taking overlapping images as either the subject moves through the beam or the beam moves over the subject. These images are combined to create a volumetric view. CBCT, however, involves a cone-shaped beam and a flat panel sensor that rotate less than 240 degrees around the subject. FBCT generally produces images with higher quality, including less artifact, lower noise levels, and better contrast detection than CBCT [35, 42, 43, 44]. An important advantage of FBCT is its speed in scanning each cross-section, which is crucial when imaging standing horses, while a slight movement of the horse in a standing CBCT can compromise the quality of the entire scan. Conversely, CBCT is fast in acquisition of volumetric data, e.g. < 1 min of total scan time for the distal limb or caudal neck of a standing, sedated horse [43]. The study found “excellent” satisfaction scores as determined qualitatively by radiologists for bone evaluation in both modalities. Compared to histological measurements, Lin et al. (2023) found that CBCT had slightly higher diagnostic sensitivity (88.5% vs. 84.1%) but lower specificity (61.3% vs. 72.3%) compared to FBCT [45]. Both studies were limited by their focus on cadavers, which are unaffected by motion artifacts which are more problematic in CBCT. Imaging standing Thoroughbred racehorses, a robotic CBCT was used for longitudinal monitoring of SCB adaptation to exercise [46]. Over the first year of training, SCB changes in all four limbs of racehorses at multiple intervals (0, 6 and 12 months) were tracked. Although the exact details of the training regimens were not quantitatively described, the study demonstrated that SCB sclerosis significantly increased over time, and the presence of SCB pathology also escalated in various parts of the metacarpal/metatarsal bone, particularly within the first six months​​.

Given these insights, the integration of advanced CT technologies like standing FBCT and CBCT into diagnostic protocols for equine OA and other SCB pathologies promises to enhance early detection and management. While FBCT offers detailed imaging that may remain superior for certain diagnostic needs, the rapid imaging capabilities, and adequate resolution of CBCT may be suitable for routine screening and monitoring of known conditions. Future research should focus on enhancing the specificity and sensitivity of these imaging modalities, exploring their application in live diagnostic settings, and expanding their use in longitudinal studies to better understand the progression of SCB conditions over time. Despite these advancements, further research is required to enable identification of horses at high risk of SCB injury [35].

Given that the primary aetiology involves mechanical loading, integrating CT imaging with assessments of mechanical strain in SCB is essential. For example, it has been postulated that two-year-old Thoroughbred racehorses experience high strain in the diaphysis and metaphysis of metacarpal/tarsal when training commences due to lower BVF and increased trabecular space [47]. However, there is little evidence on the actual strains within the bone in these cases. A recent study integrated mechanical testing, digital image correlation (a surface strain measurement technique) and standing CT imaging to investigate SCB fatigue injury in the third metacarpal bones of Thoroughbred racehorses. The study revealed increased strain under experimental loading in the parasagittal grooves of bones with SCB lesions, as detected by CT, compared to those without lesions [48].

In human medicine, the application of computational finite element (FE) models based on CT imaging has revolutionized the prediction of hip and vertebral fractures [49, 50, 51]. A similar approach could enhance the prediction of SCB failure and stress fractures in horses. Early work by Harrison et al. [52] developed a subject-specific FE model of the equine metacarpophalangeal joint that successfully incorporated deformable cartilage and elastic ligaments, but modelled bone as rigid. The model was validated against experimental data, demonstrating the importance of subject-specific geometry, in accurately predicting joint mechanics. Frazer et al. (2017, 2019) developed a CT-based FE model of the equine stifle joint [53, 54]. Using the same model with the same bone density distribution, they explored the effects of SCB cysts, with idealised geometries, on local bone stresses. Their studies revealed that the presence of SBCs or voids increased local stress levels, potentially contributing to further bone damage and persistence of these lesions. It is important to recognize that fatigue damage, including stress fractures or other lesions involve complex interactions influenced by bone repair and adaptation at the microstructural level. A combined strategy incorporating longitudinal CT scans, biomechanical testing, and computational modelling incorporating adaptation and repair components could facilitate more accurate predictions of injury and deepen our understanding of disease initiation and progression. One critical factor in developing such image-based pipelines that can accurately relate macroscale joint behaviour to its intrinsic microstructure and function is the precise evaluation and examination of SCB structure-functional relationships, which can then inform FE models of the joint. A recent study by Moshage et al. (2022) employed a comprehensive approach to establish the relationship between CT images, micro-CT data, and the elastic modulus of bone cores from juvenile horses [55]. This work highlights the importance of considering age-specific density-modulus relationships when developing FE models, ultimately leading to more accurate predictions of mechanical behaviour in equine bones. Micro-CT offers high resolution, submicron, three-dimensional images enabling the visualization of bone microarchitecture with exquisite detail.

When scanning SCB, the high resolution of micro-CT allows for detailed visualization of the porous structure of the SCB plate, the microstructure and surface shape of the tidemark, and the trabecular structure of SCB [2, 5, 16, 56]. Recent studies have shown that the superficial SCB within the first 10 mm, which is most susceptible to injury in racehorses, is less stiff than the deeper SCB. This difference persists under various loading conditions, including single impact and cyclic loading [2, 5, 57]. Additionally, superficial SCB microdamage, most commonly observed in the equine fetlock joint of racehorses, has been found to have the greatest influence on the bone’s mechanical response. While visualizing microcracks at the tidemark is usually a limitation, most micro-CT scanners can image microdamage and microfractures wider than a few micrometres or those containing highly mineralized material [5, 58].

One way to tackle the limitation of micro-CT in measuring fine microcracks has been the use of contrast-enhanced CT with lead-uranyl acetate. This heavy metal label accumulates in microdamage areas, enhancing electron density and improving visualization in micro-CT. Recent studies have developed protocols for quantifying such microdamage in SCB [2, 59]. Luedke et al. (2022) utilized contrast-enhanced micro-CT alongside mechanical testing to examine the interplay between fracture toughness and microdamage in the proximal sesamoid bones (PSBs) of racehorses [60]. They observed microdamage in the SCB independent of fracture events, although their study focused on the palmar aspect of the sesamoids where microdamage was minimal.

The loading conditions that SCB in racehorses experience is high-rate and repetitive. Understanding how this type of dynamic loading influences the microarchitecture and biomechanical responses of the SCB is crucial. One of the challenges in accurately measuring the response of SCB to mechanical loading is that SCB is integrated with its adjacent cartilage, complicating isolated testing without disrupting the most superficial layer of the SCB and the calcified cartilage, where most in vivo microdamage occurs. To address this challenge, micro-CT imaging and mechanical testing have been combined with digital image correlation (DIC). This approach allows testing of the cartilage-bone unit as a whole while extracting their individual mechanical properties [61]. Using a similar approach, Malekipour et al. (2022) demonstrated that microdamage induced in the SCB of racehorses by repetitive high rate loading in vivo reduced its mechanical stiffness under a single impact load, leading to considerably higher localized strains in damaged areas [5]. Pearce et al. (2022) utilized micro-CT alongside mechanical testing and DIC to investigate correlations between microstructural attributes to the impact response of SCB in Thoroughbred racehorses from various sites within the MC3 joint [2]. They found that bone density and mechanical properties varied across different joint locations. Particularly, in regions subjected to higher stress, i.e. the palmar aspect of the metacarpal condyles and proximal sesamoid SCB, exhibited higher stiffness and adaptive bone structure [2]. Shaktivesh et al. (2024) exclusively tested SCB and found high localized strains within SCB in relation to microdamage, under a spectrum of high-rate cyclic compression, similar to that applied to the bone in vivo [16]. Additionally, their results suggest that structural and mechanical properties of SCB exhibit variability throughout its depth from the joint surface [2, 5, 62].

While these methods provide valuable insights into the overall behaviour of the tissue, information about the internal strains and stresses within the SCB has been limited. However, with the increased use of micro-CT imaging and advancements in computational power, a combined approach integrating finite element (FE) modelling, micro-CT, and experimental data holds promise for identifying predictive parameters to predict fatigue injury in SCB, such as the volume of bone under high stresses [63]. Additionally, it is important to acknowledge that micro-CT may have limitations in accurately predicting SCB mechanical response [64], potentially due to other influencing factors such as material heterogeneity, anisotropy, and the complex geometry of bone. Therefore, a comprehensive understanding of SCB behaviour requires an integrative approach that combines high-resolution imaging, robust computational models, and experimental validation to improve the predictive accuracy for fatigue injuries and inform better clinical decisions.

Recently, the integration of Artificial intelligence (AI) and machine learning (ML) with human medical imaging has helped overcome some of the limitations of traditional methodologies in fracture detection, classification [65, 66, 67] as well as for segmentation procedures required for developing FE models. Manual segmentation, which is both time-consuming and prone to error, along with the subjective nature of diagnostic methods, are among the limitations of current methods [68]. Additionally, SCB biomechanics is a complex research area which requires analysing the interactions between mechanical loads and the responses of heterogenous and nonlinear biological tissues, especially in fatigue injury and stress fracture research. While AI, especially deep learning methods, holds promise in this area [69], these advances are in the initial stages of application within equine and other large animal models [70, 71]. AI has been used in canine models to identify fracture in long bones using radiography [72]. The slower progression in this area, compared to AI’s application in human studies, can be attributed to the smaller datasets available for these animals as compared to human datasets, which poses significant challenges for developing effective AI models [71]. Nonetheless, the adoption of standing CT in the equine domain is contributing to the creation of larger, more detailed datasets, which may improve AI application in veterinary settings. Additionally, recent improvements in algorithms, particularly those that employ data augmentation, are enhancing AI capability to process medical images of equine and large animals [65]. Rytky et al. (2021) applied deep learning segmentation to micro-CT images for assessing the three-dimensional morphology of calcified cartilage (CC) in rabbits [73]. They found strong correlations with traditional histological methods, with Dice scores of 0.891 for histology and 0.807 for micro-CT segmentation. It provided detailed insights into CC thickness variations across different anatomical regions, suggesting micro-CT as a superior method for analysing dynamic changes in cartilage mineralization. These advancements aid in predicting and diagnosing abnormalities and perform complex tasks such as segmenting different bone structures for FE models [74]. More importantly, techniques like transfer learning and data synthesis are further broadening AI applications, enabling the adaptation of models originally trained on specific datasets to new tasks, providing potential application of AI in both diagnostic accuracy and operational efficiency. For example, Amodeo et al. (2021) developed a maxillofacial fracture detection system using convolutional neural networks pre-trained on non-medical images, which was then re-trained and fine-tuned using CT scans to classify future CTs as either “fracture” or “no Fracture” [75]. The system achieved an 80% accuracy in classifying fractures, categorizing patients as fractured if two consecutive slices had a fracture probability higher than 0.99. This approach demonstrated the potential to assist radiologists by reducing diagnostic errors and delays, minimizing the risk of human error, and decreasing unnecessary hospitalizations. These studies highlight the potential of integrating advanced imaging with deep learning to improve diagnostic accuracy and clinical outcomes. Additionally, ethical considerations and the need for standardization in AI applications within veterinary medicine need to be investigated.