Introduction

Knee osteoarthritis (OA) is a complex and progressive disease. It may evolve over several decades and, if left untreated, ultimately leads to knee replacement surgery. The burden of knee OA on healthcare services is expected to increase with improvements in global life expectancy [1,2,3,4].

Non-pharmacological interventions such as weight loss, physiotherapy, and joint replacement have been the main treatment options in routine clinical practice for knee OA management. Currently, no disease-modifying medication (DMOAD) for OA has been approved by regulatory authorities. DMOADs are groups of pharmacological therapies investigated for their protective or restoring effects on articular cartilage, reducing inflammation, and impacting subchondral bone changes to stop or delay structural progression [5, 6].

Several articular and periarticular tissues, such as articular cartilage, synovium, periarticular muscles, menisci, and subchondral bone, contribute to disease incidence and progression. The subchondral bone comprises the subchondral bone plate, the underlying trabecular bone, and the bone marrow space. In addition, a zone of calcified cartilage separates the subchondral bone plate (composed of cortical bone) from the articular cartilage [7,8,9,10] (Fig. 1).

Fig. 1
figure 1

Schematic illustration of normal and osteoarthritis knee joint and subchondral bone. Subchondral bone comprises the subchondral bone plate, the underlying trabecular bone, and the bone marrow. A zone of calcified cartilage separates the subchondral bone plate from the articular cartilage

Full size image

Biomechanical derangements such as malalignment result in altered load transfer on cartilage and the subchondral bone and will accelerate the progression of knee OA [8]. One study has shown that homeostasis and integrity of articular cartilage depend on both its biochemical and biomechanical interactions with the subchondral bone [11]. Biomechanically, early subchondral bone alterations may result in reduced shock absorption capability and, therefore, predispose the knee joint to accelerated cartilage loss. This may explain why subchondral bone trabecular morphometry measurements and subchondral bone mineral density (i.e., local periarticular BMD) changes begin early in the cascade of events, even before articular cartilage loss [12,13,14,15].

Furthermore, bone marrow lesions (BMLs) play a vital role in the detection and progression of OA, which is detected in both pre-clinical [16] and clinical knee OA [17, 18] using magnetic resonance imaging (MRI) examinations. MRI also allows for the noninvasive detection and characterization of longitudinal subchondral bone changes at an early stage in the development of knee OA [19]. These changes include disruption of trabecular connectivity, trabecular microfracture, increased porosity of the subchondral bone plate, decreased trabecular thickness, decreased subchondral bone volume to total volume ratio, and resorptive changes in the subchondral bone [20,21,22]. Therefore, there has been tremendous recent interest in detecting OA-related subchondral bone changes at early (i.e., preclinical) stages before cartilage loss is apparent to timely implement secondary preventive measures.

The focus of this narrative review is to give an overview of available and emerging imaging techniques to detect and characterize knee OA-related changes in the subchondral bone (Table 1), with a special focus on subchondral bone biomarkers that have been investigated for the assessment of DMOAD effects of various medications in clinical trials and observational studies.

Table 1 Summarizing different imaging techniques discussed and their potential uses for predicting or evaluating OA with special focus on the subchondral bone
Full size table

Subchondral bone in knee OA pathogenesis: brief overview of experimental evidence

There is still a lack of knowledge regarding the complex interactions between cartilage and subchondral bone during knee OA pathogenesis. The small molecules, growth factors, and cytokine pathways are involved locally in OA development and progression that could be potential treatment targets [23]. Results from animal models have revealed that there are distinct subchondral bone phenotypes in post-traumatic OA (PTOA) versus metabolic syndrome-associated OA (MetS-OA). In PTOA, subchondral bone changes have shown a biphasic nature in disease progression. In the early stages of OA, subchondral bone osteoclast bone remodeling and loss of osseous trabeculations are dominant. In later stages, subchondral bone sclerosis and osteophyte growth become prominent (Fig. 2A) [22, 24,25,26]. On the other hand, there is rapid subchondral bone trabecular growth and, consequently, thickening of the subchondral bone plate and trabecular bone in MetS-OA mouse models [27]. These subchondral bone alterations appear much earlier than the occurrence of cartilage degeneration. Moreover, unlike the accumulated senescent cells in cartilage and synovium in PTOA, increased senescent cells are primarily found in the subchondral bone in MetS-OA models. They secrete cytokines and proinflammatory factors to influence the subchondral bone microenvironment in MetS-OA (Fig. 2B) [28, 29].

Fig. 2
figure 2

Schematic illustration of osteoarthritis-related subchondral bone alterations: A Post-traumatic OA (PTOA). In early-stage PTOA, increased platelet-derived growth factor-BB (PDGF-BB), and active transforming growth factor beta (TGFβ) in marrow microenvironment initiate aberrant angiogenesis coupled with osteogenesis. In late-stage PTOA, aberrant bone formation and neo-vessel invasion into cartilage lead to articular cartilage ossification and joint degeneration. B Metabolic syndrome-associated osteoarthritis (MetS-OA). Under metabolic stress, preosteoclasts in the subchondral bone marrow undergo cellular senescence and secrete senescence-associated secretory phenotype (SASP) factors, which promote osteoblast differentiation and inhibit osteoclast differentiation for a rapid thickening of the subchondral plate and trabecular bone

Full size image

Recent evidence has suggested an upregulated signaling and increased concentration of transforming growth factor-β1 (TGF-β1) in the subchondral bone of animal models and human knee OA (Fig. 2A). The TGF-β1 antibody stabilizes subchondral bone architecture based on micro-CT (µCT) and micro-MRI (µMRI) results in animal models and thus could serve as a potential treatment target of OA [22, 30]. Canonical Wnt pathway activation, transferring signals between the cartilage and bone, can occur in subchondral bone and osteophytes after OA incidence and in parallel with subchondral bone growth according to µCT results in mice, and inhibition of this pathway ameliorates OA severity [31, 32]. Moreover, it has been shown that preosteoclasts in osteoarthritic subchondral bone express high levels of platelet-derived growth factor-BB (PDGF-BB), which triggers neovascularization of the subchondral bone, and that joint degeneration is attenuated by PDGF-BB knockout in animal models (Fig. 2A) [25].

These observed changes in the biochemical microenvironment of subchondral bone during early OA suggest that alterations of imaging biomarkers of subchondral bone could play a critical role in detecting knee OA at an early stage in the clinical setting as well as in the mitigation of its progression [18]. In other words, there is a gradual shift from a cartilage-centric paradigm to a broader perspective focusing on the subchondral bone regarding imaging and therapeutic approaches.

Imaging biomarkers for subchondral bone assessment relevant to OA outcomes

Plain radiography

Fractal signature analysis

Trabecular bone texture (TBT) became a focus of investigation for OA-related subchondral bone architectural changes, such as differences in texture and integrity. The use of TBT calculated by the Fractal signature analysis (FSA) is a technique that can analyze the complexity of subchondral bone structure using a two-dimensional radiographic image (as a projection of the three-dimensional architecture) at a variety of scales spanning the typical size range of trabeculae (100–300 µm) and trabecular spaces (200–2000 µm) (Fig. 3). The subchondral bone architecture is determined by the number, spacing, and cross-connectivity of bone trabeculae [33,34,35]. The predictive value of FSA of the subchondral bone used for predicting radiographic OA progression was tested by Kraus and colleagues. They demonstrated that FSA of the medial tibial plateau at baseline, derived from analysis of knee X-rays utilizing the Knee Analyzer tool (Optasia Medical, Manchester, UK), was predictive of medial joint space narrowing (mJSN) progression in patients with KL grades 1–3 and thus could be considered as a potential prognostic imaging biomarker of OA at its earlier stages [33]. Kraus et al. further looked at subchondral bone trabecular integrity on plain radiographs. They found that FSA of subchondral trabecular bone integrity could predict MRI measures of OA progression over 12- and 24-month periods [36].

Fig. 3
figure 3

Various regions of interest (ROI) can be defined to conduct fractal signature analysis. The gray dashed box seen in the lateral tibial compartment defines an ROI of height 6 mm aligned with the lateral tibial plateau and spanning 3/4ths of the length from the lateral tibial plateau to the lateral tibial spine (defined by Kraus et al.) [33]. In the medial tibial compartment, an ROI of height 6 mm and width 14 mm is defined and placed immediately below the cartilage-bone interface in the middle of the distance between the tibial plateau and spine (orange dashed box). A second ROI (transparent box) of height 14 mm and width 14 mm id placed below the dense subchondral bone area. These two medial compartment ROIs have been previously defined by Hirvasniemi et al. [37] (Images and ROIs to scale)

Full size image

Janvier et al. determined semi-automatic X-ray-derived TBT to be a good predictor of knee OA progression and JSN [38]. The findings of this pilot study indicate that joint unloading implants may alter knee OA progression, defined as changes in JSN, KL grades, and tibial osteophytes, by improving subchondral bone trabecular integrity. The study showed a decrease in fractal signatures of the vertically oriented trabeculae in the medial compartment of the treated knee despite an increase in the untreated knee [39]. Another randomized controlled trial on the structural integrity of the subchondral bone in knee OA showed that patients with marked JSN receiving risedronate (15 mg/day) retained vertical trabecular structure, and those receiving risedronate (50 mg/week) increased vertical trabecular number, both quantified using FSA [40]. Therefore, FSA has been a useful tool for complex measurements of OA-related subchondral bone changes, such as trabecular texture and integrity, through widely used radiographic imaging and demonstrates the potential for providing structural insights for future DMOAD clinical trials.

Computed tomography (CT)

Cone-beam CT (CBCT)

CT studies provide measures of subchondral BMD in knee OA patients, and the use of quantitative CT (QCT) has been shown to provide 1–2% precision in measuring subchondral BMD [41, 42]. As conventional CT scanners have limited spatial resolution compared to that demanded by the small size of trabeculae, trabecular morphometry measurements pose a challenging task for QCT using conventional multi-detector CT (MDCT) [43]. The emerging cone-beam CT (CBCT) systems for musculoskeletal (MSK) imaging provide a promising alternative CT technique [44,45,46,47]. Their spatial resolution is generally better than in MDCT because CBCT uses flat-panel detectors with smaller pixel sizes (< 200 µm) compared to MDCT detectors. Moreover, the spatial resolution of CBCT is isotropic, i.e., there is no difference between the in-plane and slice directions. The greater resolution of MSK CBCT was shown in a study by Demehri et al., in which the modality was superior to MDCT in image quality of knee examinations for bone visualization tasks, including the trabecular pattern of medullary bone, cortical bone, and fractures detection. [48]. Furthermore, most MSK CBCT scanners are capable of weight-bearing imaging of the lower extremity, thus potentially enabling simultaneous assessment of subchondral bone measures and JSN in OA [48, 49]. Therefore, ongoing efforts aim to establish the feasibility of new CBCT imaging techniques for quantitative assessment of subchondral bone changes in knee OA.

In terms of subchondral bone densitometry, Turunen and colleagues used contrast enhanced-CBCT (CE-CBCT) to obtain volumetric BMD (vBMD) of cortical bone, trabecular bone, subchondral trabecular bone, and subchondral plate in cadaveric specimens and in vivo human volunteers [50]. The vBMD estimates from CE-CBCT were shown to correlate strongly with conventional QCT. In a follow-up study of patients with suspected knee injuries, Myller et al. demonstrated that vBMD and delayed cartilage contrast enhancement by CE-CBCT are related to the severity of cartilage lesions [51].

In terms of macroscale morphological subchondral bone alterations, Segal et al. demonstrated better sensitivity and accuracy of MSK CBCT compared to radiography for the detection of subchondral cysts [52] (Fig. 4). Microscopic quantification of subchondral trabecular bone using MSK CBCT remains primarily in the phase of technical performance assessment. In the study of Guha et al. using distal radius bone samples, an MSK CBCT system (Planmed Verity, Helsinki, Finland) achieved correlations with gold standard µCT of r = 0.60 for measurements of trabecular thickness and r = 0.66 for trabecular spacing [53]. A new generation of CBCT using the recently introduced fast, low-noise, ultra-high-resolution complementary metal oxide semiconductor and indium gallium zinc oxide flat-panel detectors will likely provide improved trabecular imaging capabilities. For example, Subramanian et al. showed correlations of r = 0.96 for trabecular thickness and r = 0.85 for trabecular spacing in trabecular subchondral bone cores from the tibia using a prototype complementary metal oxide semiconductor–based MSK CBCT (Fig. 5) [54]. However, the magnitudes of microarchitectural metrics measured in MSK CBCT remain biased compared to µCT because of insufficient spatial resolution, even for the newest detector hardware [53, 54]. Clinical studies are needed to establish whether the strong correlations between CBCT and µCT are sufficient to detect subchondral bone changes associated with early OA.

Fig. 4
figure 4

Images in a 68-year-old male 9 months after open reduction and internal fixation (ORIF) for a patellar fracture. Cone beam computed tomography (CBCT) shows a small subchondral cyst in the coronal (A), sagittal (B), and axial (C) planes (arrowhead). D Skyline view radiograph of the same knee demonstrates the hardware used for ORIF. No discrete subchondral cyst is identifiable in the plain radiograph

Full size image
Fig. 5
figure 5

Comparison of conventional multi-detector computed tomography (MDCT) and musculoskeletal cone-beam computed tomography (CBCT) with a complementary metal oxide semiconductor (CMOS) flat panel detector in visualization of trabecular detail of subchondral bone: A Cross-section of a human tibial plateau from a cadaver sample. B Cross-sections of a trabecular bone core (~ 8 mm diameter) harvested from underneath tibial plateau. CMOS CBCT image in B is adapted from [54]

Full size image

Micro-CT has been employed in some experimental studies to investigate the therapeutic efficacy of DMOADs on the subchondral bone. The µCT results from an experimental study by Permuy and colleagues suggested that diacerein (cytokine modulator) and hyaluronic acid (cartilage matrix precursor) significantly modified the orientation of the trabecular lattice [5]. Farnaghi et al. have shown that atorvastatin and mitochondria-specific antioxidants can rescue unfavorable subchondral bone changes in animal models of MetS-OA according to µCT results [55]. Furthermore, in a study by Sun and colleagues, the administration of teriparatide to OA models has been found to slow the disruption of subchondral trabecular bone connectivity as measured by µCT examinations [26]. Additionally, according to experimental studies by Hayami et al. and Lindström et al., it has been demonstrated that cathepsin K inhibitor treatment restores the total thickness of the subchondral bone in animal models of OA based on µCT results [56, 57]. In this context, there is a hope that CBCT with a high resolution approaching µCT could be a potential target for future clinical studies evaluating the DMOAD effect of various drugs on the subchondral bone.

MRI

MOAKS semiquantitative MRI

The MRI osteoarthritis knee score (MOAKS) is an established and validated semi-quantitative scoring system that provides comprehensive and reliable measures of subchondral BMLs, cartilage, and meniscal morphology (Figs. 6, 7 and Table 2) [19]. As an evolution of the Whole Organ Magnetic Resonance Imaging Score and Boston Leeds Osteoarthritis Knee Score, it is one of the most popular and established MRI scoring systems used in large OA observational studies and clinical trials of knee OA and may be used to assess the structural progression of OA longitudinally years after initial diagnosis (Fig. 8) [58].

Fig. 6
figure 6

Schematic illustration showing the 14 segments of evaluation for MRI Osteoarthritis Knee Score (MOAKS). Boundaries of interest have been defined, such as the inclusion the patellar crista in the medial patellar compartment and inclusion of the femoral notch in the medial femoral compartment, respectively. TROCH trochlear, ANT anterior, CENT central, POST posterior, LAT lateral, MED medial

Full size image
Fig. 7
figure 7

Bone marrow lesion (BML) incidence: A Sagittal intermediate-weighted fat-suppressed magnetic resonance imaging (MRI) scan exhibits a small MRI Osteoarthritis Knee Score (MOAKS)/Whole-Organ MRI Score (WORMS) grade 1 BML in the anterior subregion of the lateral femur (arrows). B A follow-up image obtained one year after Fig. 7A. shows a new BML in the same subregion (large arrow) with the initial BML showing a slight increase in size (small arrow). Scoring would reflect a within-grade change of a grade 1 lesion using WORMS and MOAKS. Using MOAKS, the new BML would also be coded as the increase in the number of lesions from 1 to 2. Both lesions are ill-defined (edema-like) and no cystic portions are seen

Full size image
Table 2 Semi-quantitative MRI scoring osteoarthritis knee score (MOAKS)
Full size table
Fig. 8
figure 8

Bone marrow lesion (BML) assessment over several time points: A Baseline sagittal intermediate-weighted fat-suppressed magnetic resonance imaging (MRI) scan exhibits two distinct ill-defined BMLS in the central subregion of the medial femur (arrows). Overall lesion size in subregion qualifies as a grade 1 MRI Osteoarthritis Knee Score (MOAKS)/grade 2 Whole-Organ MRI Score (WORMS) lesion. B One year later within-grade increase in overall subregional lesion size in the same central medial femoral subregion is observed. In contrast to Fig. 8A, now there are three distinct lesions that can be differentiated (arrows). The anterior BML has split into two separate lesions with a decrease in lesion size, while the previous posterior lesion exhibits an increase in lesion size. Note, diffuse femoral cartilage loss in addition. C Follow-up MRI 2 years after baseline shows a decrease in overall central femoral BML size with now a total subregional score of 1 using both WORMS and MOAKS. There are two distinct lesions now with the most anterior lesion seen in Fig. 8B showing complete regression (small arrows). No cystic component of lesions is observed. There is also a large (grade 3 WORMS and MOAKS) incident lesion in the posterior medial tibia subregion (arrowheads). Without the clinical information, the posterior lesion cannot be further characterized. Traumatic bone contusions, e.g., due to an anterior cruciate ligament tear may exhibit comparable image morphology (reticular pattern with very diffuse borders atypical for OA-associated BMLs). Also, the articular cartilage in the posterior medial tibia is normal, which may further suggest a possible traumatic origin as OA-related BMLs are commonly associated with cartilage damage

Full size image

In 2017, Kogan and colleagues developed a hybrid 3.0 T PET-MRI system, in which they used the MOAKS scoring system to measure biomarkers of OA progression such as BMLs and osteophytes and found that their hybrid system successfully and simultaneously identified morphological (using MRI) and metabolic (using PET) biomarkers of OA progression in early knee OA patients [59]. In a 2020 study, Roemer and colleagues used MOAKS scoring on 3 T MRIs from the Foundation for National Institutes of Health Osteoarthritis Biomarkers Consortium (FNIH, a nested case–control study within the Osteoarthritis Initiative (OAI)) to demonstrate that the subchondral bone phenotype was associated with increased risk of composite (i.e., structural and clinical) progression[58]. Therefore, MOAKS-based structural stratification of knees can be used in patient selection for clinical trials.

MOAKS has also been a valuable tool in assessing the effects of DMOADs on imaging biomarkers of osteoarthritic knees. In a 2020 study, Kon and colleagues used the MOAKS system to evaluate the safety and benefits of using autologous protein solution injections in knee OA patients at baseline and after 24 months. Their results showed that in patients with mild to moderate knee OA and better cartilage status, autologous protein solution injections did not significantly change their OA severity, were safe, and lead to significant pain improvement 3 years after a single injection [60]. Genechten and colleagues aimed to measure the short-term clinical effect, therapeutic response rate, and safety of another injection treatment, namely autologous micro-fragmented adipose tissue injections, and used the MOAKS system to classify mild-to-severe osteoarthritic knees at baseline and up to 12 months [61]. They found that autologous µ-fragmented adipose tissue injections provided early clinical improvement, moderate therapeutic response rate after 12 months, and demonstrated a negative association between MRI-measured BML and therapeutic response rate during this time period [61]. These findings pointed to the role of increases in therapeutic response rate with autologous µ-fragmented adipose tissue injections in lowering patients’ MOAKS-assessed BML associated with OA progression. A randomized clinical trial on 74 subjects with knee OA showed that an intramuscular therapeutic dose of clodronate followed by a maintenance dose is effective for the improvement of BMLs and bone marrow edema, paired without treatment in (Fig. 9), according to the Whole Organ Magnetic Resonance Imaging Score over 12 months (200 mg daily for 15 days and then once weekly for the next 11.5 months) [62].

Fig. 9
figure 9

MRI sequence protocol considerations for bone marrow lesion (BML) assessment: A Sagittal intermediate-weighted fat-suppressed magnetic resonance imaging (MRI) scan depicts ill-defined BML in the anterior lateral femur (arrows), reflecting a grade 3 lesion in both systems: MRI Osteoarthritis Knee Score (MOAKS) and Whole-Organ MRI Score (WORMS). In addition, there is a large BML at the lateral patella (arrowhead). B Corresponding dual-echo at steady-state (DESS) MRI does not allow differentiation of BML from unaffected marrow due to magnetic susceptibility, despite DESS being a sequence with T2-weighted image characteristics. This low sensitivity for BML detection, especially for edema-like lesions, is the characteristic of all gradient echo sequences, including DESS

Full size image

A study on the longitudinal, multi-center OAI database demonstrated that statin treatment may be protective against the MRI derived OA-related subchondral bone damage in a subset of patients who had no/minimal baseline BML, were under statin therapy due to cardiovascular disease, and demonstrated generalized OA phenotype [63]. The effects of zoledronic acid administration on the subchondral bone are controversial in the literature. In a recent 2-year clinical trial, twice-yearly administration of 5 mg of zoledronic acid at baseline and within 12 months showed no significant improvement for MRI-detected BML size in a group of 223 patients [64]. This result contradicted the results of a small pilot trial including 59 individuals, which found that a single dose of zoledronic acid (5 mg) reduced MRI-detected BML size over 6 months [65]. Another 2020 study by Ballal et al. investigated oral treatments for OA using sagittal turbo spin echo fat–suppressed intermediate-weighted MRI of knee OAs in women on an oral bisphosphonate treatment and showed that after comparing BML volume at baseline and after 12 months, the treatment provided no benefit, in terms of changing BML volume, for knee OA patients[66]. In a new, 2022 study by Huang et al., they aimed to use intelligent algorithms, including gray projection algorithms, adaptive binarization algorithms, and active shape models, as a basis for treatment of knee OA with platelet-rich plasma [67]. In their study, they used the MOAKS score and measurements of the inferior angle of the femur, superior angle of the tibia, and tibiofemoral angle by an automatic magnetic resonance diagnostic model to validate the diagnostic findings of the intelligent algorithms and measure the treatment effects of platelet-rich plasma in osteoarthritic knees and found the MOAKS scores of OA patients to be lower after treatment.

In vitro histologic and experimental studies have shown promising results for the DMOAD role of strontium ranelate by inhibition of subchondral bone remodeling [68, 69]. Strontium ranelate has also been demonstrated to improve the joint structure and symptoms of knee OA, however, its impact on subchondral BMLs in clinical practice is not well understood [70, 71]. In a large population of knee OA patients, it was shown that MRI-measured bone marrow abnormalities were positively interrelated with urinary C-telopeptide of type C collagen (CTX-II) as a biomarker of type II collagen degeneration and progression of OA [72]. An early decrease in the level of CTX-II was seen in risedronate-treated patients, but the mechanism is unclear. This may represent a primary effect on subchondral bone turnover leading to improvement in cartilage stability [73].

Despite vitamin D’s crucial role in calcium homeostasis, bone metabolism, and subchondral remodeling, to date, original studies and meta-analyses on knee OA MRI have failed to prove the potential role of vitamin D supplementation on subchondral bone structure assessed as improvements in the tibia cartilage volume, JSW, and BMLs [74, 75].

High-resolution three-dimensional (3D) fast MR imaging with free steady-state precession [FISP])

Bolbos et al. used 3 T MRIs of TBT and showed that cartilage degeneration was associated with loss of mineralized subchondral bone [76]. There was an association between the decrease in the MRI-derived subchondral bone quantitative measures and abnormal measures of compositional MRI parameters for early cartilage degeneration such as increased T1rho and T2 relaxation times [76].

MacKay and colleagues further pursued subchondral bone texture analysis using MRI. They used fast imaging with gradient echo (GRE) sequence such as steady-state precession (FISP) to assess the subchondral bone texture of knee joints from the participants of the OAI Bone Ancillary Study at the 30- or 36-month and the 48-month visits and defined OA progression as radiographic medial tibiofemoral joint space width loss between the patients’ 36–72-month visits [77]. This study demonstrated the value of MRI-assessed subchondral bone texture as an imaging biomarker for OA progression in clinical trials as the 12–18-month change in subchondral bone architecture was significantly associated with radiographic OA progression at 36 months [77]. The periarticular bone was also quantitatively assessed as a biomarker for early detection of OA progression in a study by Lo and colleagues [78]. Their longitudinal study used 12–18 months of OAI MRI measurements to evaluate OA patients’ trabecular bone morphometry, defined as apparent bone volume fraction, trabecular number, thickness, and spacing. The researchers found an association between high progression of JSN with higher MRI-detected baseline measures of the bone volume fraction, trabecular number, and thickness as well as lowered baseline and decreasing trend in trabecular spacing [78]. Therefore, periarticular bone assessments provided an avenue for early detection of OA progression and later prevention of progress.

Pishgar and colleagues attempted to develop a framework for the computer-assisted measurement of trabecular biomarkers and extract subchondral trabecular biomarkers derived from conventional knee MRIs (intermediate-weighted sequences) of patients enrolled in OAI [79]. They extracted the subchondral bone quantitative biomarkers from OAI’s Bone Ancillary Study, such as trabecular thickness, spacing, connectivity density, and bone-to-total volume ratio obtained from conventional intermediate-weighted MRI sequences. Out of the evaluated biomarkers, trabecular thickness and bone-to-total volume ratio demonstrated high reliability compared to high-resolution FISP. There was also a modest association with OA progression up to 48 months [79]. Subsequently, they focused on the trabecular thickness and bone-to-total volume ratio measures from conventional MRI scans of OAI’s FNIH OA biomarkers consortium. They demonstrated an association between the evaluated biomarkers and increased medial tibial cartilage volume loss at both the 12-month and 24-month follow-ups [18]. Over the course of these studies, their team suggested the potential value of measuring MRI subchondral trabecular biomarkers from conventional intermediate-weighted sequences to detect structural OA progression as early as 1 year after baseline. Although some high-resolution GRE sequences (e.g., FISP) can depict subchondral trabecular morphometry, other GRE sequences, such as dual-echo at steady-state (DESS), are not suitable for BML detection due to low sensitivity caused by trabecular magnetic susceptibility (Figs. 9, 10) [80].

Fig. 10
figure 10

A Radiograph showing a normal knee joint with Kellgren-Lawrence grade 0. B Fast imaging with steady state precession (FISP) sequence MRI acquisition shows normal subchondral trabecular morphometry. C Radiograph of a knee with marked narrowing of the lateral joint space (Kellgren-Lawrence grade 4) D FISP sequence MRI acquisition of the same diseased joint shows altered subchondral trabecular morphometry in the lateral tibial plateau

Full size image

Other imaging modalities

Bone density scan (DEXA scan)

In addition to “local” subchondral bone analysis in knee OA patients (which is the primary focus of this narrative review), it has been also suggested in the literature that OA and “systemic” osteoporosis are negatively associated as they uncommonly occur simultaneously. Despite decades of research, the exact association pattern and underlying mediatory factors between systemic/local BMD and the incidence/progression of OA remain unclear. A prospective cohort of 86 subjects with symptomatic knee OA showed that BMD of the hip was negatively associated with cartilage damage and subchondral BMLs [81], and another prospective cohort of 153 subjects without clinical OA demonstrated that higher systemic BMD was associated with increased early knee cartilage damage [82]. In a prospective cohort of 158 subjects, DEXA-derived systemic BMD was positively associated with increased cartilage thickness in patients with radiographic OA using both cross-sectional and longitudinal analysis [83]. Additionally, high medial tibial subchondral BMD predicted an increase in medial tibial cartilage thickness longitudinally over the period of 2.7 years [83]. Further studies are warranted to reveal the mechanism of subchondral BMD changes using DEXA in the setting of knee OA, which may prove helpful in the development of DMOADs.

PET

PET imaging can offer a shift of focus on subchondral bone imaging biomarkers from quantitative structural biomarkers to those reflecting changes in subchondral bone metabolism related to knee OA pathogenesis. Using a 3 T hybrid PET/MRI system, Watkins et al. found that bone perfusion, tracer extraction fraction, and total tracer uptake into bone significantly differed in knee OAs [84]. These findings suggest that subchondral bone metabolism is altered in knee OA and provide a tool to further explore the subchondral bone during the initiation and progression of OA [84]. Subsequently, Jena et al. continued this examination of [18F] NaF uptake using hybrid PET/MRI systems in a series of case studies of patients with varying OA severity (KL grades 1–4) and also suggested the distinct value of PET for subchondral bone analysis in future DMOAD clinical trials [85].

Artificial intelligence for the assessment of subchondral bone

Exponential increases in the use of artificial intelligence (AI) in evaluation of knee OA can provide a venue for automated and rapid prediction of disease outcomes and progression in the research setting. A recent study by Almhdie-Imjabbar et al. builds on imaging studies that had identified TBT as an imaging biomarker for OA progression and introduces using AI on conventional radiographs to provide prognostic data for knee OA outcomes [38, 86]. They used data from the OAI and Multicenter Osteoarthritis Study cohorts and automatically segmented subregions of interest from the plain radiographs of patients with KL grades 2–3. In both cohorts, OA progression was defined as mJSN increases during the follow-up radiographic examinations. Akin to FSA, subchondral bone texture analysis using TBT-convolutional neural networks (TBT-CNN) predicted mJSN in OA patients over a period of 4–6 years [86]. Such algorithm-based subchondral texture analysis has also been recently used to automatically quantify structural changes induced by DMOADs. A 2022 study by Jang and colleagues used a fully automated AI algorithm model to analyze the changes in the upper tibial subchondral bone structure measures in patients undergoing polynucleotide filler injections (HP cell Vitaran J) once a week for 5 weeks [87]. They found that the patients’ Visual Analogue Scale and WOMAC decreased for 3 months after polynucleotide treatment and that AI-detected changes in their subchondral bone texture analysis measures [87]. Overall, these findings demonstrate that AI can be a valuable tool for automated detection of microstructural changes in the subchondral bone and may be used to asses therapeutic strength of DMOADs in future clinical trials.

Conclusions

The subchondral bone has been shown to be a possible target for secondary preventive measures of knee OA including future potential DMOAD developments and assessments. Fractal signature analysis with radiography has helped model and quantitatively measure subchondral bone microarchitecture while serving as a tool to predict trabecular texture and integrity as predictive measures of OA progression. New developments in CT techniques, such as CBCT and CE-CBCT with an improved spatial resolution for bone imaging, can tremendously impact OA diagnosis and outcome prediction. MRI has been tremendously implemented for subchondral bone analysis in knee OA patients, including BML (using the semiquantitative MOAKS and quantitative assessment), subchondral trabecular bone morphometry measurements, and TBT (using TrueFISP). DEXA scans provide BMD measurements which can be used to predict changes in subchondral BMLs, cartilage degeneration, and OA progression. PET scans can also provide quantitative measurements related to subchondral bone physiology and aid in the understanding of OA initiation and progression. AI is an exponentially expanding research focus in subchondral bone imaging research, which aims at automating TBT analysis and OA prediction. Despite the lack of pharmacological treatment for OA, familiarity with various available and advanced imaging modalities will certainly help in optimizing the imaging protocols for assessment of potential DMOADs and their protective effect on knee OA in future clinical investigations.