|Home | About | Journals | Submit | Contact Us | Français|
This nonsystematic literature review provides an overview of magnetic resonance imaging (MRI) of subchondral bone marrow lesions (BMLs) in association with osteoarthritis (OA), with particular attention to the selection of MRI sequences and semiquantitative scoring systems, characteristic morphology, and differential diagnosis. Histologic basis, natural history, and clinical significance are also briefly discussed.
PubMed was searched for articles published up to 2011, using the keywords bone marrow lesion, osteoarthritis, magnetic resonance imaging, bone marrow edema, histology, pain, and subchondral.
BMLs in association with OA correspond to fibrosis, necrosis, edema, and bleeding of fatty marrow as well as abnormal trabeculae on histopathology. Lesions may fluctuate in size within a short time and are associated with the progression of articular cartilage loss and fluctuation of pain in knee OA. The characteristic subchondral edema-like signal intensity of BMLs should be assessed using T2-weighted, proton density-weighted, intermediate-weighted fat-suppressed fast spin echo or short tau inversion recovery. Several semiquantitative scoring systems are available to characterize and grade the severity of BMLs. Quantitative approaches have also been introduced. Differential diagnoses of degenerative BMLs include a variety of traumatic or nontraumatic pathologies that may appear similar to OA-related BMLs on MRI.
Subchondral BMLs are a common imaging feature of OA with clinical significance and typical signal alteration patterns, which can be assessed and graded by semiquantitative scoring systems using sensitive MRI sequences.
Although deterioration of the hyaline articular cartilage and osteophyte formation are considered the hallmarks of knee osteoarthritis (OA), it is now widely accepted that OA is a disease of the whole joint including the subchondral bone, synovium, menisci, and ligaments (1). Subchondral bone marrow edema-like lesions (BMLs) detected on magnetic resonance imaging (MRI) are a frequent finding of OA and have been shown to be associated with clinical manifestations such as pain and with structural progression (2–6). However, the signal alterations of subchondral BMLs are nonspecific, and a variety of traumatic and nontraumatic pathologies may exhibit similar imaging characteristics. Knowledge of the underlying pathophysiology, clinical significance, and natural history of subchondral BMLs is still limited but has been increasing rapidly in recent years. This review article covers the histologic basis, natural history, clinical significance, and MRI evaluation of BMLs in association with OA. We particularly focus on technical aspects of MRI acquisition and semiquantitative scoring systems, the characteristic morphology of BMLs, and on the differential diagnosis of subchondral BMLs.
A PubMed search for articles published through October 2011 was performed, using the keywords “bone marrow lesion”, “bone marrow edema/oedema”, “bone marrow edema/oedema pattern”. This search strategy yielded 1044 abstracts from all types of publications. The search was then narrowed by adding keywords “osteoarthritis”, “MR imaging”, “histology”, “pain”, and “subchondral”. Relevant to this nonsystematic literature review were original publications that focused on BMLs or BMLs were one of several structural features assessed. Of particular interest were articles on structural/clinical correlations, histopathology, and MRI methodologic studies, as well as articles focusing on the natural history and those assessing interventions. All joints were included. Articles focusing on primary inflammatory arthritides such as rheumatoid arthritis were not considered. The emphasis was on articles published within the last 10 years and written in English, but older publications were also included if considered essential for this review. Altogether, 114 articles (18 articles published between 1957 and 1999 and 96 articles published between 2000 and October 2011) were included. In addition, the reference lists of all articles cited in this review article were screened to complete the literature search. The initial literature search was performed by LX and DH, and screening for relevance was performed by all authors. Evaluation was based on the authors’ own clinical and research experience in the field.
Regional bone marrow signal intensity alteration on MRI was first described by Wilson et al., using the term “bone marrow edema” (BME) to describe MRI findings in painful joints lacking any specific radiographic abnormalities (7). The bone marrow of the affected joint showed ill-defined decreased signal intensity on T1-weighted (T1w) images and increased signal intensity on T2-weighted (T2w) images. A study showed that the areas of these MR signal changes in conjunction with OA corresponded to local scintigraphic uptake (8). The terms BME or BME pattern have been commonly used to describe these ill-defined MR signal changes of bone marrow. However, because studies have shown that various pathologic entities, not just edema, can exhibit BME pattern on MRI, the use of the term “BML” became common, particularly in the OA research community (4,9,10). However, in clinical practice, the terms BME and BME pattern are still more commonly used by radiologists.
According to Roemer et al., traumatic and nontraumatic BMLs should be differentiated, and furthermore, the associated pathology should be carefully described (11). Examples could be “traumatic BML without associated fracture,” “traumatic BML in conjunction with osteochondral fracture,” “idiopathic nontraumatic BML,” “OA-associated BML,” “chronic BML in conjunction with osteonecrosis,” and so on. For this review, we focused on the discussion of BMLs in association with OA (12). Although most studies of BMLs in conjunction with OA have focused on the knee joint, BMLs are not unique to the knee. The hip is also frequently affected (13–17), as well as the hand (18,19), ankle and foot (20,21), shoulder (22), and spine (23).
Elucidating the histopathology of BMLs in conjunction with OA contributes to better understanding of their natural history and relationship with symptoms. Surprisingly, there is only limited literature that focuses on the correlation between MRI-detected BMLs and histopathology in osteoarthritic joints. All available data are derived from patients who had total knee or hip replacement because of advanced OA (14,15,24,25).
Zanetti et al. reported that the “bone marrow edema pattern zone” mainly consists of normal tissue (53% fatty marrow, 16% intact trabeculae, and 2% blood vessels) and a smaller portion with several types of abnormalities, including bone marrow necrosis (11% of area), necrotic or remodeled trabeculae (8%), bone marrow fibrosis (4%), edema (4%), and bone marrow bleeding (2%) (24). Saadat et al. observed subchondral ingrowth of fibrovascular tissue and increased bone remodeling at the exact anatomical location of the BME pattern on MRI in 3 osteoarthritic knees that were analyzed but did not find edema (25). The authors suspected that increased perfusion to the area owing to the ingrowth of fibrovascular tissue contributed to the BME pattern on MRI.
Similar findings were reported by Taljanovic et al. (14) and Leydet-Quilici et al. (15) in regard to hip OA. Taljanovic et al. described microfractures in different stages of healing that were present in all patients and found that the number of microfractures correlated with the amount of BME visualized by MRI (14). Leydet-Quilici et al. divided bone marrow signal abnormalities into more edema-and necrosis-like patterns and found edema-like patterns to be associated with histologic BME and necrosis-like patterns to be associated with bone marrow fibrosis and necrosis on histology (15).
The natural history of BMLs includes development, progression, regression, and resolution. Unlike cartilage abnormalities, BMLs represent various histologic findings and can fluctuate in size within a relatively short time (3,26,27).
One longitudinal study assessed subjects with a clinical diagnosis of knee OA and reported that in 0.6% of knees BMLs shrank, in 73% BMLs were stable, and in 27% BMLs increased in size (4). However, the stability of the BMLs may well be because the study used only MRI sequences without fat suppression, rather than MRI sequences that are sensitive to BMLs (ie, fluid sensitive with fat suppression). In contrast, the MOST (Multi-center Osteoarthritis) study, a large-scale longitudinal study of subjects with or at high risk for radiographic knee OA (28), used fat-suppressed MRI to characterize BMLs and found that nearly one-third of knees without baseline BMLs developed new lesions at follow-up, 66% of the prevalent BMLs changed in size, and 50% of subregions with prevalent BMLs showed either regression or resolution (5). In addition, a significant increase in the prevalence of BMLs with an increase in the OA score was observed by Phan et al. (29) and Felson et al. (2).
Cross-sectional associations between BMLs and cartilage damage in the same location within the knee joint have been reported (30,31). One study involving 176 healthy, middle-aged women demonstrated a significant cross-sectional positive association between BMLs and cartilage damage after adjusting for the potential risk factors: age, height, weight, and cartilage volume (odds ratio (OR) 1.12 to 2.82) (32). The relationship between longitudinal cartilage loss and BMLs in the same location has also been studied by many investigators. Progression of cartilage damage was found to be an independent risk factor for compartment-specific incident BMLs (33). Also, a number of studies have demonstrated that the severity of BMLs at baseline, progression of BMLs, and newly developing BMLs are independently predictive of compartment-specific longitudinal cartilage loss in patients with or without OA or pain (4–6,32,34,35).
Roemer et al. demonstrated that, when compared to stable BMLs, subregions without BMLs at baseline are associated with a decreased risk of cartilage loss (OR 0.1 to 0.3), while progression of BMLs (OR 1.5 to 5.2) and the development of new BMLs (OR 2.1 to 5.9) are associated with an increased risk of cartilage loss in the same subregion (5). In addition, BML size at baseline was found to be associated with an increased risk of subsequent cartilage loss. A study performed by Hunter et al., including 271 older adults with clinical knee OA, showed that both medial and lateral knee compartments with a higher baseline BML score had greater cartilage loss, and an increase in BMLs was strongly associated with worsening of the cartilage score (4). However, the association of BML change with medial tibiofemoral cartilage loss was not significant after adjusting for alignment. Kothari et al. investigated 177 osteoarthritic knees and found that BMLs at baseline predicted cartilage loss in the same subregion 2 years later, after adjusting for subchondral bone cysts and bone attrition at baseline (OR 1.59 to 8.82) (35).
BMLs are often associated with subchondral cysts, which are characterized by well-defined rounded areas of fluid-like signal intensity on T2w fast spin echo fat-suppressed sequence MR images (24,36). Carrino et al. reported that 92% of incident subchondral cysts developed in regions with BMLs, and the size of BMLs always changed with cyst development (37). In the MOST study it was demonstrated recently that prevalent BMLs showed a strong and significant association with incident cysts in the same subregion (as defined by the Whole Organ Magnetic Resonance imaging Score and WORMS (38)), with an odds ratio of 12.9 after adjustment for full-thickness cartilage loss (36). These investigators all maintained that BMLs can be a precystic lesion (36,37), but that not all BMLs will become cysts (37).
Systemic factors that have been reported to be related to BMLs in the tibiofemoral joint (TFJ) include age (30,31), male gender (30,31,39), height (30,32), weight (32), and body mass index (31,33,39). For the latter 2, the relation may be present because the knees with BMLs are osteoarthritic. Structural factors other than cartilage and sub-chondral cysts that are commonly investigated in association with BMLs in the knee include malalignment of the TFJ (4), bone area of the lateral tibial plateau (30,32), and meniscal deterioration in the ipsilateral compartment (40–42). Those risk factors were reported to be associated with BMLs in the medial TFJ compartment, lateral TFJ compartment, or total TFJ compartment (Fig. 1). BMLs localized in areas not covered by articular cartilage, such as the interspinous region at the tibia and femoral notch, are known as insertional BMLs and are highly associated with anterior cruciate ligament (ACL) or posterior cruciate ligament damage and may be a consequence of tensile stress on these ligaments (43).
Pain in patients with OA may be multifactorial, and one of the major tissues responsible for pain is the subchondral bone. Fibrovascular replacement of adipose marrow tissue within the BMLs was demonstrated by Walsh et al. to be associated with growth factor expression (44). Growth factor may facilitate the growth of sensory nerves or sensitized nerves (45,46), and therefore, it was hypothesized that the growth of fibrovascular tissue within the BMLs may be a source of pain in patients with OA (44). Impaired venous drainage from the BMLs as well as calcitonin gene-related peptide, substance P, and GAP-43/B-50 protein have also been suggested as potential mechanisms of pain generation in patients with OA (47,48).
Felson et al. reported that BMLs are cross-sectionally associated with the presence of pain (OR 3.31, 95% confidence interval 1.54 to 7.41) after adjustment for severity of radiographic OA, effusion, age, and sex, and that large lesions appeared exclusively in the patients with pain (2). Ip et al. reported a significant association between BMLs and pain on climbing stairs but not pain from walking (49). According to Sowers et al., BMLs <1 cm were found more frequently (OR = 5.0; 95% confidence interval = 1.4, 10.5) in the painful knee OA group than the painless knee OA group. However, compared to the group with knee pain but without OA, BMLs were 4 times more likely to occur in the painless knee OA group (50).
Some longitudinal studies have elucidated the relationship between progression of BMLs and pain. Using data from the MOST study, Felson et al. showed that incidence or progression of BMLs was higher at follow-up in subjects with pain than in control subjects without pain (51). Zhang et al. reported that changes in the severity of BMLs were associated with fluctuations in the severity of frequent pain in knees with and without baseline radiographic OA (3).
Although some studies reported no association between pain and BMLs (29,52,53), a recent systematic review including 5 longitudinal and 17 cross-sectional studies demonstrated, with moderate levels of evidence, that BMLs are associated with knee pain (54). Thus, overall evidence suggests that BMLs can be a source of pain in knee OA.
Selecting sensitive MRI sequences, using appropriate evaluation methods, and thorough knowledge of the characteristic imaging manifestations of degenerative BMLs on MRI are all indispensable for accurate assessment of BMLs. The general consensus in the OA research and radiological community is that fluid sensitive fat suppressed (FS) sequences, i.e. T2w (long repetition time (TR) and long echo time (TE), e.g. TR/TE =3500/120 ms), proton density-weighted (PDw) (long TR and short TE, e.g. TR/TE = 3500/20) or intermediate-weighted (Iw) (PDw with a TE of about 40 ms, e.g. TR/TE =3500/40) fast spin echo (FSE) sequences, or a short tau inversion recovery (STIR) sequence should be used to assess BMLs (11,52,55–59). Hayashi et al demonstrated, in a direct comparison, that the IW FS sequence depicts more subchondral BMLs and that the FSE sequence depicts the lesions as larger when compared to a gradient echo type sequence (DESS) (60). The aforementioned fluid-sensitive FSE sequences should thus be used for determination of lesion extent whenever the size of a subchondral BML is the focus of attention (Fig. 2). However, one should also note that T2w FS images may exhibit areas of inhomogeneous fat saturation, which can mimic BMLs (57).
Gradient recalled echo (GRE)-type sequences such as spoiled gradient recalled acquisition at steady state (SPGR), fast low angle shot (FLASH), double echo steady state (DESS) and others have been shown to be very sensitive in delineating subchondral cysts (Fig. 2), but these sequences are insensitive to marrow abnormalities due to trabecular magnetic susceptibility or T2* effects and will not show the true extent of the lesion (57). GRE-type sequences are helpful to distinguish cystic from ill-defined edema-like BMLs (60). On contrast-enhanced MRI, BMLs show avid enhancement and depict the volume of BMLs in an identical fashion to PDw FS or STIR sequences (61,62). For reasons of cost and for patient comfort, optimizing MRI sequence protocols is essential in the routine clinical setting and in epidemiological and clinical studies. While there is no evidence to guide the number of planes in which pulse sequences should be acquired, acquisition of fluid-sensitive fat-suppressed sequences in 2 or 3 orthogonal planes is common to avoid misinterpretation of partial volume effects.
Several semiquantitative scoring systems are available that allow cross-sectional and longitudinal evaluation of BMLs. Analyses based on semiquantitative scoring have added deeply to our understanding of the pathophysiology and natural history of OA, as well as the clinical implications of the structural changes that are assessed (3,4,36,41,51). To date, 4 semiquantitative scoring systems for whole organ assessment of knee OA have been published (Table 1).
The WORMS (38) has been widely used in epidemiologic studies and clinical trials to assess several OA features of the knee (4,63–65). WORMS uses a subregional approach to the scoring of BMLs because recording the exact number of individual lesions is time-consuming and sometimes difficult, as lesions may be directly adjacent to each other or will merge or split in longitudinal assessments (66,67). The Knee Osteoarthritis Scoring System covers MRI-detected OA features much like WORMS (68), except that subchondral BMLs are scored individually for each subregion and each score is differentiated by the size of the lesion. The Boston Leeds Osteoarthritis Knee Score (BLOKS) system uses a subregional division similar to the Knee Osteoarthritis Scoring System.
All 3 of these scoring systems have been published with comparable excellent reliability data (Table 1). Deciding which scoring system should be used for a given study can be tricky. In a recent comparison, MRIs were assessed separately using either the WORMS or the BLOKS systems (66,69). WORMS was found to be superior to BLOKS for assessing BMLs. Two reasons were given for this superior performance: (1) scoring individual lesions over time is not advisable for BMLs, which can split or merge in follow-up scans, making it difficult to characterize change in a single lesion; and (2) BLOKS uses regions that are much larger than WORMS and therefore much less sensitive to changes in lesion volume.
A fourth system, the Magnetic Resonance Imaging Osteoarthritis Knee Score (MOAKS) system (73), was developed in an attempt to rectify these deficiencies in BLOKS and uses an approach to BML scoring that incorporates the WORMS approach of scoring BMLs in a subregional fashion (67). MOAKS scores the number of lesions per subregion, summed lesion size as a percentage of subregional bone volume, and the ratio of noncystic to cystic lesion in the subregion. An important issue to note is that the same lesion may be assigned different grades in different scoring systems (Fig. 2).
Quantitative measurements of BMLs using MRI provide a tool for evaluating and monitoring lesions that is less observer-dependent than subjective scoring. However, the feathery, ill-defined margin of the lesions makes it difficult to quantify their volume accurately (57). When choosing a sequence for quantification, it is important to consider that the different sequences have several factors that affect the measured T2 relaxation time (70). Static field inhomogeneities and the error propagation introduced by off-resonance effects also affect quantification accuracy (71).
In quantitative MRI, BMLs have to be manually or semiautomatically segmented using a grayscale threshold (20,72). Compared to manual quantification, threshold-based segmentation of BMLs shows higher intra- and interobserver reliability (73,74). Three-dimensional MR spectroscopic imaging may also provide a tool to evaluate BMLs quantitatively in OA, based on the fact that significantly elevated water and unsaturated lipids have been observed in BMLs. The volume of elevated water calculated by 3-dimensional MR spectroscopic imaging correlated significantly with the volume of BMLs based on segmentation on MRI in knees with OA (73).
The hip is particularly vulnerable to OA as a weight-bearing joint, and OA is the most common disease of the hip joint in adults (75). BMLs associated with OA in the hip appear as an extensive subchondral edema-like pattern within the femoral head and neck and the acetabulum (Fig. 3) (13,76), which corresponds to a combination of edema, fibrosis, and necrosis at histopathology (15). The hip OA MRI scoring system, recently developed by Roemer and colleagues, allows the joint to be evaluated as a whole organ (17). The intra- and interobserver reliabilities of BML scoring with hip OA MRI scoring are 0.72 and 0.85 (weighted kappa), and the intra- and interobserver agreements for all BML scores are 86.7% and 84.9% (17). In the same study, large lesions were found to be significantly related to the Kellgren Lawrence grade of the hip radiograph (P = 0.002).
Hand OA is one of the most prevalent forms of OA (77). Tan and colleagues evaluated early hand OA by combining high-resolution MRI with histologic examination and found that the small-joint collateral ligaments and tendons have a central role in the early stages of hand OA (78). The investigators detected BMLs at the collateral ligament origins adjacent to the focal subchondral bone edema. This pattern of BML was found extending to the subchondral regions even where the articular cartilage was well preserved. The pattern was thought to be explained by the “enthesis organ concept” and differed in pathologic content from the subchondral BMLs associated with cartilage degeneration (78,79). A semiquantitative MRI scoring system, the Oslo Hand OA MRI score (18), was recently proposed specifically for hand OA. In this scoring system, subchondral BMLs are scored as the proportion of bone with a lesion in the distal and proximal part of the proximal interphalangeal and distal interphalangeal joints separately; moreover, BMLs at collateral ligament insertion sites are scored separately in the radial and ulnar part of the proximal interphalangeal and distal interphalangeal joints (18). The intra- and interobserver reliabilities were reported as 0.89 and 0.83 for subchondral BML scoring and as 0.81 and 0.42 for BMLs at collateral ligament insertion sites (18).
In OA of the ankle and foot, the metatarsophalangeal joints are commonly affected (80) (Fig. 4A). Unlike the knee and hip, primary OA of the ankle is much less frequent than posttraumatic OA (21). Besides OA, a wide range of inflammatory diseases such as rheumatoid arthritis and seronegative spondyloarthropathies can produce periarticular BMLs in multiple bones. Associated MRI findings such as osteophytes and diffuse articular cartilage loss in OA (Fig. 4B), and periarticular soft-tissue edema, synovitis, and marginal erosions in rheumatoid arthritis, can aid in the diagnosis (81).
Vertebral endplate signal changes are common findings among patients with degenerative disk diseases of the lumbar spine. Modic et al. classified them into 3 types: type 1 changes represent bone marrow edema pattern; type 2 changes are fatty conversion from normal red bone marrow; type 3 changes correspond to subchondral bone sclerosis (82,83). Modic changes are most common at the L4-L5 and L5-S1 levels, usually adjacent to degenerated or herniated intervertebral disks (82–84). Modic type 1 changes are unstable lesions and seem to be strongly associated with low back symptoms and segmental instability (23).
Last, in regard to OA of the shoulder joint (Fig. 5), there is a paucity of publications specifically regarding BMLs in association with glenohumeral OA.
Imaging evaluation of BMLs may also potentially be performed using dual-energy computed tomography (85). This feasibility study showed that the “virtual noncalcium technique” enabled investigators to subtract calcium from cancellous bone, allowing bone marrow assessment in the posttraumatic knee with computed tomography. Whether this technique can be applied to BML assessment in OA is not known at present. Also, nuclear medicine-based imaging techniques reflecting bone metabolism will show increased tracer intake (86,87) or glucose accumulation (88) in areas of BMLs but they are nonspecific and will not add to the differential diagnosis. Conventional radiography, grayscale ultrasound, or Doppler techniques do not allow visualization of BMLs (11). Overall, MRI is the current standard for BML assessment.
Trauma-induced BMLs can be differentiated into lesions associated with acute trauma, such as bone contusions, and subacute lesions as a result of chronic overload, such as insufficiency fractures and repetitive microtrauma in conjunction with physical activity. Furthermore, the integrity of the overlying cartilage has to be considered when purely subchondral traumatic BMLs need to be differentiated from concomitant osteochondral and chondral injury.
Bone contusions demonstrate characteristic MRI findings as poorly defined, reticulated, heterogeneous alterations and show a distinct location pattern according to the mechanism of injury (89). Subchondral impaction injuries are a result of severe impaction forces with larger BML volumes and variable degrees of depression of the articular cortical surface, compared to contusions (90,91).
Traumatic BMLs may appear following stress fractures, which include insufficiency fractures and fatigue fractures (92). Insufficiency fractures occur in pathologic bone unable to withstand the stresses of normal activity and can be seen in a variety of diseases such as osteoporosis, osteomalacia, and Paget disease. Subchondral insufficiency fractures mainly involve the femoral head (93–95), although they can involve other joints (96). In contrast, fatigue fractures are the result of excessive strain on normal bone that exceeds the bone’s capacity to maintain itself. Athletes, military recruits, and distance runners are particularly prone to fatigue fractures (97–99). The irregular, discontinuous, low-intensity band seen on T1-weighted images surrounded by BML is the characteristic MRI appearance of subchondral stress fractures and corresponds histologically to the fracture line (93,97).
Stress-related BMLs can appear without coexisting degenerative or traumatic changes. Subchondral BMLs related to exercise may appear in a golfer’s hand or in the ankles and feet of asymptomatic physically active individuals, with no other associated characteristic MRI findings (100,101). Another form of BML occurs at the insertions of ligaments, especially the ACL and posterior cruciate ligament, and is considered to be due to the traction from ligaments (43). BMLs related to ACL injuries mostly persist after 1 year of follow-up according to 1 report (102).
Nontraumatic BMLs occur in a group of disorders in which an underlying disease or a prior surgical procedure dominates the history, clinical findings, prognosis, and course of the disease. Avascular necrosis (AVN) is depicted as a lesion of high signal intensity surrounded by a reactive rim of low signal intensity on T1-weighted images, and a mixed high and low signal intensity on T2-weighted images (“the double line sign”) is pathognomonic for AVN. This characteristic rim may present postcontrast enhancement on T1w images (103). Progression to larger epiphyseal necrosis and ultimately osteochondral collapse is associated with the appearance of localized BMLs around the infarct. After the femoral head, the knee is the second most frequent site for osteonecrosis: in order of frequency, in the lateral femoral condyle, lateral tibial plateau, and medial femoral condyle (104). Spontaneous osteonecrosis of the knee (SONK) is thought to be caused by an insufficiency fracture, usually in the weight-bearing region of the medial femoral condyle, most often occurring in elderly women (105–107). After the fracture, regions of osteonecrosis develop in the bone superficial to the fracture. The focal subchondral area of low signal intensity adjacent to the subchondral bone plate is a specific MRI finding for SONK, and no postcontrast enhancement of this area is detected on T1w images. A lack of peripheral low signal intensity rim as seen in AVN and nonspecific BML patterns are characteristic of SONK (108).
Bone marrow edema syndrome (BMES) refers to transient clinical conditions with an unknown pathogenic mechanism, such as transient osteoporosis of the hip (which typically affects women during the third trimester of pregnancy, but can also affect middle-aged men and nonpregnant women ), regional migratory osteoporosis, complex regional pain syndrome, and reflex sympathetic dystrophy. BMES most commonly affects the proximal femur, sometimes bones around the knee, and infrequently bones in the ankle or foot (110). Reversible large diffuse BMLs without persistent abnormalities suggest a diagnosis of BMES (111). Reactive inflammatory BMLs are most commonly observed in chronic polyarthritis, reactive arthritis, rheumatoid arthritis, and bacterial arthritis. The predominant synovial and periarticular soft tissue involvement and bone erosions are MRI characteristics of the inflammatory arthropathies (112). Disuse osteoporosis is a pathologic condition of bone characterized by focal demineralization that occurs as a result of immobilization. The etiology of this entity appears to be related to loss of mechanical stress with subsequent metabolic changes in the osteoclast and osteoblast activity (113). The fluid-sensitive MR sequences show diffuse multiple dotted hyperintensity throughout the bones involved (114).
Diffuse infiltration of the bone marrow is a common finding of a variety of oncological and hematologic diseases, such as lymphoma, myeloma, and leukemia, as well as some benign tumors, such as Langerhans cell histiocytosis, chondroblastoma, osteoid osteoma, and osteoblastoma. Tumor infiltration presents as well-defined signal alterations with enhancement on postcontrast T1-weighted images, corresponding to the complete replacement of fatty marrow (104). The tumor-associated BML is an ill-defined edema-like signal alteration which surrounds the tumor.
The MRI-detected subchondral BML, comprised of fibrosis, necrosis, edema, and bleeding into fatty marrow in different proportions as well as abnormal trabeculae, is a common finding in patients with OA. Recent studies have shown that BMLs are involved in the progression of articular cartilage loss and fluctuation of pain. Fluid-sensitive T2w fast spin echo fat-suppressed sequence or STIR sequences should be used to assess semiquantitatively the maximum size of BMLs in OA. GRE-type sequences do not allow accurate lesion size estimation but are helpful to distinguish subchondral cysts from BMLs. Several semi-quantitative scoring systems with proven high reader reliability are available to assess and grade the severity of BMLs. A variety of diseases other than OA may show BML-like signal changes that should be distinguished from degenerative BMLs.
David T. Felson’s work is supported by NIH Grant AR47785.
Ali Guermazi is the president of Boston Imaging Core Lab, LLC (BICL), Boston, MA, a company providing radiological image assessment services. He is a consultant to MerckSerono, Novartis, AstraZeneca, Genzyme and Stryker. Frank Roemer is a shareholder of BICL.
None of the other authors have declared any possible conflict of interest.