Introduction

Chronic nonbacterial osteomyelitis (CNO) is an autoinflammatory disease affecting children, adolescents, and young adults, mostly characterized by remitting and relapsing bone lesions scattered in time and location across the body. After more than two decades of its increased recognition, CNO still remains an elusive condition of unclear etiology and a diagnosis of exclusion. Its pathogenesis involves dysregulation of immune system in the background of genetic and epigenetic alterations [1••]. Primarily a pediatric condition, CNO can persist into or start at adulthood, the latter scenario (i.e., adult-onset disease, seen in 6% all CNO patients [2••]) displaying somewhat different imaging—as well as clinical—features [3•]. Initial presentation of children with CNO is typically around the age of 10 years. Magnetic resonance imaging (MRI) plays a key role in the diagnosis, therapy monitoring, and follow-up of CNO, owing to the fact that MRI is remarkably more sensitive or specific in this disease than clinical examination, radiography, or scintigraphy [2••, 4,5,6]. Early diagnosis of CNO is especially important as early therapy can lead to a better outcome [7]. The absence of conclusive clinical features and widely accepted laboratory tests—including validated disease biomarkers—for CNO substantiates the widespread use of MRI at the initial identification and during the course of this chronic condition, which can be self-limited, chronically active or recurrent [1••]. MRI clearly stands out in the imaging diagnosis of CNO due to its excellent and unparalleled comprehensive depiction of bone marrow, soft tissues surrounding the bones, and joints. The absence of X rays (an inherent feature of radiographs and computed tomography [CT]) and gamma rays (the essence of bone scintigraphy and positron emission tomography) is an additional advantage of MRI, as CNO is mostly a disease of the pediatric age group, where avoiding unnecessary risks from ionizing radiation is an especially important consideration.

Epidemiological, pathogenetic, clinical, and therapeutic aspects of CNO have been extensively studied and repeatedly reviewed elsewhere [1••, 6,7,8]. Due to its key role in diagnosis, imaging has also been accorded significant coverage in many of such reviews on CNO. As musculoskeletal and pediatric radiologists in an academic tertiary health care institution with a prominent pediatric rheumatology department, we have had the opportunity to evaluate over a decade 60 cases that received the diagnosis of CNO. This review centers on the current knowledge and our experience about the use of MRI in pediatric CNO and highlights the techniques used, pitfalls encountered, suggestive features identified, and differential diagnostic possibilities considered during daily practice.

Targeted and Whole-Body MRI for CNO

Almost half of CNO patients are reportedly not seen by a pediatric rheumatologist until at least one year after the onset of their symptoms [9]. Pediatricians who are not rheumatologists and many clinicians from other branches of medicine are still not sufficiently aware of CNO. In such a setting, targeted MRI of the most symptomatic body part is usually requested initially. This background increases the importance of the awareness of CNO by radiologists, who are instrumental in the referral to pediatric rheumatologists of patients not already seen by the latter. What appears to be the preferential involvement of some body parts may help in leading the way to a correct diagnosis even in the setting of an initial targeted MRI.

Some cases with CNO have active sacroiliitis during their disease course [10, 11] and sometimes it is on a sacroiliac, hip, or pelvic MRI that characteristic periphyseal osteitis of CNO in the vicinity of acetabular Y-cartilage and proximal femurs is identified (Fig. 1). Of a small cohort of 15 cases with CNO at our institution, eight (53%) had sacroiliitis [11]. Moreover, our observations in a more comprehensive ongoing study of 60 cases seen by our pediatric rheumatologists over a longer period disclosed what appeared to be “active sacroiliitis” on MRI, which is described as bone marrow edema in the sacrum or ilium adjacent to the sacroiliac joints [12, 13], at least once during their disease course in 39 (65%). Moreover, about half of these patients (22 of 39) also had evidence of chronic sacroiliitis usually in the form of erosions, mostly at the iliac sides. In all children undergoing sacroiliac MRI, regardless of the presence of sacroiliitis, we routinely look for osteitis away from the sacroiliac joints and particularly around the bilateral Y-cartilages, which are usually within the field of view on semiaxial sequences, to alert us to the possibility of CNO. Nevertheless, some patients with CNO might also feature overlapping forms of arthritis (i.e., psoriatic and enthesitis-related) [10] and might display sacroiliitis along with periphyseal osteitis (one patient in our cohort had psoriatic arthritis and another enthesitis-related arthritis along with CNO). Arthritis is not an uncommon feature of CNO patients [2••, 14], reported in 29% of the largest CNO cohort [2••] and described to be contiguous to bone lesions in 70% of patients in one study [14]. However, although mentioned in some reports on CNO [5, 10, 15], sacroiliitis has not been a specifically designated type of arthritis in most CNO cohorts so far. Currently, there are no internationally recognized diagnostic criteria to distinguish psoriatic or enthesitis-related arthritis from CNO [2••]. Nevertheless, the hallmark of psoriatic and enthesitis-related arthritis is enthesitis, whereby the bone inflammation is centered at an enthesis, whereas osteitis in CNO is periphyseal (or, less commonly, diaphyseal). It is not currently clear whether sacroiliitis is actually a part of CNO or osteitis adjacent to sacroiliac joints in children needs to factor in CNO as a separate entity rather than representing “active sacroiliitis.”

Fig. 1
figure 1

Coronal fat-saturated T2W images from the initial targeted pelvic MRI in these 6-year-old boy (a) and 7-year-old girl (b) with CNO showed osteitis that involved in the boy the sacral (S1) and iliac metaphyseal-equivalents abutting both sacroiliac joints, the left obturator ring (arrow, a) near the metaphyseal-equivalent ischiopubic synchondrosis, and periphyseal regions at the proximal femurs, and in the girl the left iliac bone abutting the sacroiliac joint (arrow, b), the bilateral Y-cartilages and proximal femurs. Note soft tissue edema surrounding the lesions, more prominent in the boy (a)

Full size image

Pelvic region including the hip and/or sacroiliac joints is by far the most common location of an initial targeted MRI in our experience. Other such locales include the knee, calf, thigh, ankle, sternoclavicular region (where the involvement of the medial clavicle and/or sternum is characteristic), elbow and wrist. When we see scattered or isolated periphyseal osteitis on the initial targeted MRI, we raise the possibility of CNO and recommend a whole-body MRI (WB-MRI) if also clinically warranted. CNO may start and stay unifocal in a small percentage (7–11%) of children [16, 17], hence the preference of the term “CNO” over “chronic recurrent multifocal osteomyelitis” (CRMO).

WB-MRI, on the other hand, provides a comprehensive overview of CNO lesions that are characteristically distributed spatially (Fig. 2) and temporally. WB-MRI can be performed as a screening test either right at the start of the clinical evaluation, when CNO is a serious clinical consideration, or, much more commonly in our experience, following the targeted MRI of a body part that yielded findings in favor of CNO. Of the 60 CNO patients we have performed MRI on so far, all but two had an initial targeted MRI for a body part (spine, pelvis, or extremities) and 53 (88%) had at least one WB-MRI during their management. In the largest reported cohort of CNO patients (comprising 6% adult-onset patients), 66% had targeted MRI, while 34% had WB-MRI [2••].

Fig. 2
figure 2

Coronal (a, d–f) and sagittal (b, c) STIR sequences covering the entire body and the spine, respectively, are usually sufficient in WB-MRI for CNO (arrows point to CNO lesions). WB-MRI is an excellent tool to show the distribution of lesions in CNO and monitor therapy response as in this 6-year-old boy (same patient as in Fig. 1). Note the disappearance after etanercept treatment of widespread foci of active osteitis lesions on the fourteenth month follow-up MRI (c, g–i). Mild collapse at the superior end-plate of T5 vertebral body (c, arrowhead) is a sequela of the earlier lesion

Full size image

Practical Technical Considerations for and Pitfalls in WB-MRI for CNO

The usual age of onset (around 10 years) and the time of follow-up imaging (second decade of life) mostly obviate the need for sedation (that might be necessary in younger children) before or during WB-MRI. Nevertheless, longer MRI examination durations are known to decrease patient compliance, resulting mostly in motion artifacts. It is important to keep WB-MRI examinations as short as possible for the sake of patient comfort; yet it is also essential not to compromise the diagnostic yield. As a bare minimum, WB-MRI for CNO must feature coronal short tau inversion recovery (STIR) images of the entire axial and appendicular skeleton [18, 19]. Due to the better assessment of the spine (which is a common location for CNO lesions in children and reported more often between 20 and 35% of cases in the published series [20]) on the sagittal plane (Fig. 2b, c), it is advisable to add sagittal STIR images of the spine to the imaging protocol of WB-MRI for CNO [5, 16, 20,21,22]. Such an MRI examination usually entails four to six coronal plane stations (subsections) to cover the entire skeleton and two sagittal plane stations to display the entire spinal column in a child or adolescent. Sagittal images of the spine depict vertebral body height loss much more clearly than do coronals, allowing better quantification. Such depiction is especially important, given the possibility of structural damage in the form of vertebral body fractures (with the attendant risk of long-term sequelae such as scoliosis and kyphosis) in 11% of CNO patients [15], whereby the use of bisphosphonates may be warranted [1••].

In order to minimize stitch artifacts that can occur as signal loss where two adjacent stations merge on a composite WB-MRI series (Fig. 3), phase oversampling can be used [23]. Although composite WB-MRI provides a nice overall picture of the entire skeleton, every subsection (station) of such a stitched depiction needs to be individually assessed with zooming, panning, and adjusting the grayscale on STIR images; otherwise some lesions might be overlooked or mischaracterized. This is especially important for identifying the involvement of small or slender bones such as the mandible, clavicle, scapula, sternum, ribs, patella, fibula, tarsals/metatarsals, and carpals/metacarpals. One case in point is the difficulty posed in the mandible, which is a recognized location of CNO involvement [24, 25], by unerupted teeth (characterized by a high STIR intensity dental follicle) [26] and a normal thin band of high STIR signal surrounding the teeth. We recommend cross-checking the mandibular STIR signal alterations from the WB-MRI coronal images on the sagittal spine images, which cover the jaw (Fig. 4); nevertheless, the possibility of dental periapical inflammation unrelated to CNO might necessitate clinical correlation [27].

Fig. 3
figure 3

The stich artifact at the level of the upper thighs on coronal T1W (a) and STIR (b) images from WB-MRI in a 9-year-old girl with CNO appears as a thin horizontal line devoid of signal. Osteitis lesions in CNO usually have a lower signal intensity (a, measured intensity level: 135) on T1W MRI than the normal skeletal muscle (measured “242” here, a). The normal hematopoietic bone marrow has a higher signal intensity on T1W MRI than the normal skeletal muscle (measured “460” here, a). The patient had complete response to methotrexate treatment. Biopsy from the distal right femoral metaphysis showed chronic inflammatory cells, edema and fibrosis within the bone marrow space (c) and no evidence of malignancy

Full size image
Fig. 4
figure 4

Sagittal STIR spinal image set (a) from WB-MRI is used to verify a focus of mandibular osteitis (arrow) identified on the coronal STIR image set (b, arrow) in the same 9-year-old girl with CNO as in Fig. 3. Dental periapical inflammation is a differential diagnostic challenge in some cases with mandibular osteitis as in this patient and might necessitate pedodontic correlation. On WB-MRI, the sternum in its entirety, clavicles at their medial aspects, and parts of hands/wrists, when placed under buttocks, are also covered on the sagittal STIR spinal images, allowing cross-checking when suspicions arise on the coronal images

Full size image

Suppressing the fat signal from the bone marrow as well as the subcutaneous and deep soft tissues, STIR images are the mainstay of WB-MRI, as they conspicuously depict osteitis in CNO as bright foci. Such foci can then be correlated on T1-weighted (T1W) sequences, which show in osteitis low signal usually deeper than both the intermediate-signal hematopoietic bone marrow and the normal skeletal muscle (Fig. 3). However, in our experience, the contribution of routine T1W sequences to the WB-MRI for CNO with the resultant prolongation of the total examination time does not justify the increased burden on the patient comfort and strain on the system throughput. STIR is usually sufficient for WB-MRI in the work-up of CNO and most of the time a musculoskeletal radiologist would not need T1W sequences (which may be used in select circumstances on a targeted study) as part of the CNO protocol. In other words, a WB-MRI comprising only a coronal STIR stitched set covering the entire body and a sagittal STIR stitched set showing the entire spinal column (Fig. 2a, b) is sufficient [5, 17, 23, 28, 29•]. Cumulative scan time for a WB-MRI comprising coronal whole-body and sagittal spine STIR sequences is approximately 20 min (with six coronal and two sagittal stations) in our 1.5 T unit; patient entry and exit, coil preparation, shim adjustment and table movement between stations take an additional 5–7 min, resulting in a total examination time of slightly under 30 min.

In keeping with the literature [5, 22], we do not use intravenous contrast for WB-MRI. Diffusion-weighted imaging as part of WB-MRI, which we sometimes use for research purpose, is generally not necessary for the diagnosis of CNO [18, 20].

In centers where more than one MRI is available, it makes sense to perform all WB-MRI examinations at the same machine (usually a 1.5 T unit) in cases suspected to have CNO, thereby enabling the use of the same protocol and parameters in the follow-up and increasing the robustness of comparability. Due to inherent magnetic field inhomogeneity and geometry of coil elements that pick up signal, WB-MRI is prone to suffer from inadequate signal acquisition from the arms, forearms, wrists and hands in a neutral supine position. We recommend the placement of hands under buttocks to ensure sufficient signal acquisition from forearms, wrists and hands. Placement of hands above the head or upon the pelvis has also been recommended [20]; their positioning over the abdomen, however, might result in degrading of signal from forearms and hands with motion artifacts from respiration.

Considering the possibility of aborting an MRI examination during the acquisition of images (due mostly to patient-related tolerance issues such as claustrophobia or body movement), the order of imaging sequences need to be prioritized according to the most-to-least helpful. In this regard, it is important to acquire the STIR coronal set first, followed by the STIR sagittal spinal set, and then the T1W coronal whole-body sequence set (if the latter is employed at all). Sufficient information to rule in or out CNO can usually be collected with a complete set of STIR coronal whole-body images [18].

MRI Features of CNO and Its Treatment Response

Whether targeted or covering the entire body, CNO has some distinct features on MRI. The hallmark of CNO on MRI is periphyseal osteitis that appears as bright foci on STIR either—and more commonly—on the metaphysis or epi-/apophysis of a long bone. Diaphyseal involvement is less common. The physis itself might also be involved. In fact, the physeal involvement in CNO has been described more than 30 years ago [30], along with its inherent potential to cause premature epiphyseal fusion that might result in sequelae such as limb length discrepancy, valgus/varus and other posture deformities. In view of the growth plates at the cranial and caudal ends of the vertebral bodies (physeal-equivalents), CNO involvement of the vertebral bodies is also essentially a periphyseal osteitis, sometimes with actual physeal destruction (Fig. 2c) leading to collapse. The small “physeal tongues” that we sometimes see (Fig. 5a, b) at characteristic sites of metaphyseal involvement in CNO are likely the sequelae of endochondral ossification disruption at the time of “physitis” allowing chondrocytes to extend into the metaphysis [31, 32]. CNO may also give rise to focal early closure of the physeal plate, known as a “physeal bridge (or bar)”, which is another type of sequelae of physeal insult (Fig. 5c). Growth arrest/recovery lines seen at or around CNO lesions in some patients even before any treatment also point to some possible disturbance with endochondral ossification during the course of the disease (Fig. 5d).

Fig. 5
figure 5

“Physeal tongue” at the distal tibia (arrows, a, b) in this 10-year-old-boy and “physeal bridge” at the left proximal femur (arrow, c) in this 14-year-old boy are likely sequelae of earlier physitis (both patients were under treatment for CNO). Note the calcaneal active osteitis in the younger boy (arrowheads, a, b). A 3-year-old girl, who was eventually diagnosed to have CNO, already had growth arrest/recovery lines accompanying osteitis around the knee at the time of her initial presentation (d)

Full size image

CNO lesions involve the bone marrow with or without inflammation of the adjacent periosteum and surrounding soft tissue (such inflammation is also bright on STIR) (Fig. 1). Less commonly, CNO might present as a lytic lesion on radiographs and a cystic/abscess-like or aggressive lesion on MRI, which might raise the possibility of subacute infectious osteomyelitis (Brodie’s abscess) or Langerhans cell histiocytosis.

Pelvic bones including the proximal femurs, long bones around the knees, lower > upper extremities, and vertebral bodies are usually involved in CNO. Clavicle, sternum, ribs, scapula, and mandible are also among sites of involvement. Lesions commonly, but not invariably [30, 33••, 34], have a symmetric distribution. Characteristic lesion sites include “metaphyseal-equivalents” [35, 36] such as bones surrounding sacroiliac joints, Y-cartilage, ischiopubic synchondrosis, greater trochanteric physes [37], and “epiphyseal-equivalents” such as the patella [36], tarsal and carpal bones [37]—both types of “equivalents” underlining the “periphyseal” predilection of the disease. Metaphyseal-equivalents denote portions of flat or irregular bones abutting cartilage, with marrow signal intensity characteristics and vascularity resembling those of the metaphysis of a long bone [35]. Bone areas adjacent to the sacroiliac joints having been designated as metaphyseal-equivalents [35,36,37] and active sacroiliitis having been defined as bone marrow edema/osteitis surrounding sacroiliac joints [12, 13], the occurrence of active sacroiliitis in almost two-thirds of our cohort of 60 patients at least once during their disease course should not be surprising.

Although several patterns have been described for CNO on WB-MRI (such as the “tibioappendicular multifocal” and “claviculospinal paucifocal” patterns [17]), we suspect that, with cohorts rarely exceeding 100 patients [2••, 33••] so far, investigators (including us) have been prone to highlight lesion distribution features not necessarily applicable to wider populations.

A caveat on imaging concerns bisphosphonates, which are a common second-line medication in CNO, used in approximately 13% of the largest reported cohort of CNO patients [2••], with the highest complete response rate among second-line treatment options [2••]. Bilateral symmetric metaphyseal bands bright on STIR after the initiation of pamidronate treatment likely represent the increased conspicuity of the expanded zone of endochondral ossification early on during the administration of this bisphosphonate (Fig. 6) [20, 29•]. Such high signal bands can mimic disease relapse and they underscore the importance of providing the radiologist information during the requisition for MRI on the CNO therapy (drug and the time of its first administration) the patient has been receiving. A key characteristic of such bands is their abrupt demarcation with respect to the adjacent metadiaphysis (Fig. 6a), which is not an expected feature of osteitis in CNO. Cyclic bisphosphonate therapy for CNO inhibits osteoclastic activity, thereby increasing the bone mineral density and resulting in thin bands of sclerotic lines (low signal on MRI and white on radiography) [38] at long bones. Such lines resemble growth arrest/recovery lines and can, when alternating with normal bone marrow density/intensity on radiography/MRI, produce a “zebra stripes” appearance (Fig. 6) [20, 29•, 38]. These lines represent sclerosed nondecalcified cartilage at the metaphysis, move toward diaphysis as the child grows (Figs. 6, 7), sometimes with focal undertubulation (Fig. 6) [38], gradually fade and are eventually converted into normal bone [39]. Individual pamidronate lines representing each cycle do not appear distinctly after a while and what appears to be a single band, presumably representing the cumulative batch of infusions, can be seen (Fig. 7). Interestingly, these lines and bright on STIR bands were not mentioned in earlier studies on the WB-MRI validation of the efficacy of pamidronate therapy in children with CNO [40, 41].

Fig. 6
figure 6

Follow-up WB-MRI in a 6-year-old boy with CNO early on during the course of pamidronate therapy (a) shows bright metaphyseal bands with mild undertubulation (arrow) around the knee that are iatrogenic, representing the expanded zone of endochondral ossification. Osteitis in CNO does not appear as such bright bands with uniform thickness and sharp demarcation. Therefore, in the absence of clinical information on therapy, this appearance should not be mistaken for residual or relapsing CNO. Anteroposterior knee radiograph in another 8-year-old girl with CNO after receiving a 5-cycle course of pamidronate therapy (1 mg/kg/day for 2 consecutive days every 3 months) shows the iatrogenic “zebra stripes” appearance with mild undertubulation (b, arrow). Bright metaphyseal bands (c, arrows) on a follow-up WB-MRI from the same day again represent the expanded zone of endochondral ossification, a therapy effect. Note the “geologic strata” type layering of the bands (c), pointing to the timeline of therapy, which is in contrast to what would be the more “geographic” distribution of hematopoietic bone marrow. The spacing of the lines on the radiograph (b) perfectly reflects the fact that the patient skipped the second cycle of the course; note that the other (later) four lines are evenly spaced apart. Despite bisphosphonate treatment, this girl had a new-onset metatarsal osteitis (d–f, arrows) on the same WB-MRI

Full size image
Fig. 7
figure 7

Coronal T1W (upper row) and STIR (lower row) images from WB-MRIs show sequential migration from metaphyses towards diaphyses of sclerotic lines secondary to pamidronate treatment in a girl who was diagnosed to have CNO at 8 years of age. This girl received six cycles of monthly pamidronate treatment (1 mg/kg/day for 3 consecutive days every month) a few weeks after her baseline whole body MRI at far left. Subsequent WB-MRIs were 11, 25, 35, and 44 months after the start of the first—and only—cycle of pamidronate treatment. Interestingly, but not surprisingly, actual distance measurements between the positions of the lines (double-sided arrows) were directly and perfectly proportional to the time elapsed since the start of the pamidronate therapy

Full size image

MRI Differential Diagnosis of CNO

Although many CNO lesions can be safely considered straightforward in terms of recognition by non-radiologists, there are instances where pitfalls or mimics might cause confusion, misdiagnosis or overdiagnosis. Edema-like marrow signal intensity on MRI, which is the hallmark of osteitis, is a common finding and has various etiologic factors and temporal evolution characteristics [42•]. Pediatric radiologists with exposure to musculoskeletal diseases or dedicated musculoskeletal radiology subspecialists, whose expertise encompasses imaging in orthopedics, sports medicine, rheumatology, oncology, and infectious disease, are ideally endowed to offer the best radiology service to children and adolescents with CNO. Nowhere is this more pronounced than in the imaging differential diagnosis of CNO, which comprises several general categories of conditions and diseases, many with distinguishing clinical and laboratory features (those highlighted here are exclusively MRI mimics of CNO at musculoskeletal sites). It should be borne in mind that some of the differential diagnostic considerations may coexist with CNO.

Normal Growth or Variant Anatomy

Normal conversion of red to yellow marrow during the early decades of human life follows a predictable course from the distal aspects of the extremities to the proximal and occurring in the epiphysis and diaphysis earlier than in the metaphyses, eventually being confined by 25 years of age mostly to the axial skeleton and proximal femur and humerus [43]. Exceptions do exist, however, to this gradual and orderly conversion of marrow, such as small globular or patchy foci of residual red marrow at and around the ankles and wrists in children and adolescents [43,44,45]. Such foci should not be mistaken for osteitis, which is usually larger and more prominent on the STIR sequence. Nevertheless, there might be instances where targeted imaging of the wrists/hands or ankles/feet is necessary in CNO patients (or patients suspected to have CNO) to exclude such foci of residual red marrow on WB-MRI.

Focal periphyseal edema (FOPE) is a transient and spontaneously resolving condition that presents in some adolescents centered at the physes, characteristically extending into both the metaphysis and epiphysis (Fig. 8a) [46]. Considered to represent focal stress and microtrauma occurring secondary to decreased flexibility at the site of a physeal closure, FOPE can be mistaken for CNO foci due to its characteristic spatial and temporal distribution around the knee and puberty, respectively.

Fig. 8
figure 8

MRI differential diagnosis of CNO include focal periphyseal edema (FOPE) as in this 13-year-old girl (a), talocalcaneal impingement with accessory anterolateral talar facet (in a 16-year-old boy) (b, asterisk), S. aureus osteomyelitis (in a 12-year-old boy) (c, d), acute lymphoblastic leukemia (in a 6-year-old boy) (e), Hodgkin lymphoma (in a 15-year-old boy) (f), and Langerhans-cell histiocytosis (in a 10-year-old boy) (g). Bilateral femoral neck stress fractures (h, arrows) that occurred after playing several repetitive football games in another 9-year-old boy and coexisted with CNO resolved after activity modification

Full size image

Variant anatomy such as lumbosacral transitional vertebra, bi-/multipartite patella, tarsal coalition [47], type 2 accessory navicular bone [48] and accessory anterolateral talar facet [49] (Fig. 8b) can sometimes present with bone marrow edema-like changes (owing to stress related to biomechanical imbalance/instability) surrounding a synchondrosis or opposing bone surfaces that might be mistaken for osteitis in CNO.

Infectious Osteomyelitis

Metaphyseal predilection and multifocality in up to 10% of cases [50] with acute infectious hematogenous osteomyelitis (along with unifocality of CNO in about 29% of the largest reported cohort [2••]) might pose a problem in imaging-based differentiation from CNO. Although periostitis is commonly seen as a thin bright stripe on STIR enveloping the bone at the site of active CNO lesions, surrounding soft tissue inflammatory changes are not as common (and virtually non-existent in the case of surrounding soft tissue abscesses) in CNO and are, when present, characteristically less severe than in infectious osteomyelitis (Fig. 8c, d) [29•]. Infectious (bacterial or mycobacterial) osteomyelitis of the spine usually involves the intervertebral disk; diskal abnormality is not a common feature of CNO unless the growth plate itself (a physeal-equivalent) is involved [29•].

Neoplastic Diseases

Common childhood malignancies such as leukemia (Fig. 8e) and lymphoma (Fig. 8f), and bone metastases are among neoplastic mimickers of CNO. Such neoplastic processes cause marrow replacement—not marrow edema/inflammation as in CNO—and are not necessarily centered at periphyseal regions (Fig. 8e) as usually are CNO lesions. Metastastic lesions tend to have sharper borders on MRI than inflammatory lesions would have. Presence of widespread lymphadenopathy—not expected in CNO—also favors systemic lymphoma, which is the more frequent cause of osseous involvement in lymphoma [29•].

Systemic Diseases

Langerhans cell histiocytosis (LCH) (Fig. 8g) can present with multiple bone lesions that can mimic CNO. LCH commonly involves the skull, spine and pelvis; in the long bones the diaphysis is favored [2••]. Juvenile idiopathic arthritis (JIA), when not coexisting with CNO, can sometimes be confused with the latter. However, bone involvement in JIA is usually in the form of enthesitis (i.e., centering at entheses) (Fig. 9) or periarticular—and not periphyseal as in CNO.

Fig. 9
figure 9

Coronal fat-suppressed T2W and post-contrast fat-suppressed T1W MR images and the corresponding coronal reformatted CT show bone inflammation centered at the greater trochanteric enthesis (gluteus minimus tendon insertion) (ellipses, a, b) in this 17-year-old boy with enthesitis-related arthritis. These findings are in contradistinction to periphyseal osteitis in CNO. Note small erosions on both greater trochanters (arrowheads, a–c) and osteitis (arrows, a, b) of right acute on bilateral chronic sacroiliitis with iliac-sided erosions on both sides

Full size image

Other Conditions

Stress fractures and complex regional pain syndrome (CRPS) also need to be considered in the imaging differential diagnosis of CNO. Strenuous physical activity in an otherwise healthy child with characteristic sites for the former (Fig. 8h) and regional involvement for the latter with autonomic dysfunction findings and trophic changes usually help in arriving at the correct diagnosis [51]. In children, CRPS is more common among adolescent girls [51]. Conspicuous bone marrow and soft tissue edema that usually surrounds an osteoid osteoma can also mimic CNO; identification of a usually ovoid lesion with a central nidus may sometimes warrant CT in the suggestive clinical setting.

Future Directions

Increased use of WB-MRI in recent years will help further promote the awareness of this condition, resulting in earlier diagnosis in more cases [52]. Examination duration for WB-MRI will likely further decrease during this decade, rendering this technique more patient-friendly. Advanced acceleration techniques, such as synthetic MRI that reconstructs multiple sequences (e.g., T1W, fat-suppressed T2W, STIR) in a single acquisition, will have a key role in this regard [53•].

In some conditions involving bone marrow such as inflammation, infection, and traumatic contusion, complete disappearance of lesions on MRI may come weeks even months after clinical resolution [42•]. Our, as well as other groups’, experience suggests that this is also likely in CNO, whereby the occurrence of asymptomatic lesions on WB-MRI has been reported [4, 33••, 54]. Nevertheless, prospective studies in larger cohorts with clinical correlates of lesions identified on WB-MRI are needed to validate whether such a scenario (as encountered with other inflammation, infection and trauma) exists also for CNO. This might have ramifications in the design of a reasonable MRI follow-up regime, which is yet to be outlined in CNO. Within the scope of current knowledge base, we recommend the use of WB-MRI in an annual—if not more frequent—basis as long as patients are under treatment or until patients are symptom-free for at least one year without any treatment. By showing the extent of lesions, hence disease burden, this imaging follow-up scheme may guide tailoring effective treatment regimens and result in accumulating valuable information about the morphological course of an elusive chronic disease, hopefully streamlining better management practices.

Attempts are underway to formulate a scoring system [7, 22, 55] and a radiological activity index [56] for CNO. In the absence of reliable biomarkers, MRI will likely continue to take center stage in scoring. Such efforts will hopefully help standardize WB-MRI reporting for CNO and these formulations, after validation with further prospective studies, may be used as a versatile tool of disease activity assessment.

Another advancement in future may occur in the venue of artificial intelligence applications for the detection of inflammatory foci—and presumably the quantification of lesion load—on WB-MRI in CNO. Machine learning has already been shown to be capable of detecting improvement or worsening of CNO lesions on WB-MRI [57].

Conclusion

MRI is currently indispensable in the diagnosis and follow-up of CNO. Although targeted MRI may be initially used (and thereafter employed whenever confusion exists regarding the coexistence of another condition or clearance of differential diagnostic considerations), WB-MRI is the reference imaging standard for the comprehensive assessment of the baseline status in and follow-up of CNO. Coronal and sagittal STIR sequences covering the entire body and the spine, respectively, are mostly sufficient for WB-MRI, requiring an approximately 30-min examination duration. The hallmark of CNO is periphyseal (metaphyseal and/or epi-/apophyseal) osteitis, identified as bright foci on STIR, with or without inflammation of the adjacent periosteum and surrounding soft tissue. Response to bisphosphonate treatment for CNO has some unique MRI findings that should not be mistaken for residual or relapsing disease. We encourage pediatric rheumatologists to seek the expert opinion of musculoskeletal/pediatric subspecialty radiologists, who are well poised to handle the complexities of planning, lesion and pitfall identification, and differential diagnosis in MRI of patients suspected to have CNO.