Benefits of diffusion-weighted imaging in pediatric acute osteoarticular infections

Acute pediatric osteoarticular infections bear significant morbidity if diagnosis and treatment are delayed, with potentially irreversible damage to the growing skeleton [1‐3]. The need for prompt imaging and surgery is well documented and magnetic resonance imaging (MRI) holds a central role in current clinical algorithms [4, 5]. When acute osteoarticular infection is suspected in a child, emergency MRI aims to identify bone marrow signal abnormality, joint effusion and collections that can be targeted by puncture and aspiration to achieve a definite diagnosis [6, 7]. Alternatively, imaging findings will suggest a differential diagnosis and redirect management [8].

Contrast-enhanced MRI is the standard of imaging in children with suspected acute osteoarticular infection [9, 10]. However, safety concerns regarding the adverse effects of gadolinium retention in children raise the question of gadolinium-based contrast injection in pediatric MRI [11‐14] . An unenhanced protocol that could confirm musculoskeletal infection and assist in the decision for surgical management is desirable.

Diffusion-weighted imaging (DWI) relates to tissue and fluid composition, such as high cellularity and viscosity of pus [15]. It has proven helpful in musculoskeletal imaging of adults for the diagnosis of soft-tissue abscess [16], osteomyelitis of the diabetic foot [17] and spine infections [18]. As for pediatric musculoskeletal imaging, previous studies have investigated the role of DWI in juvenile idiopathic arthritis [19, 20], chronic recurrent multifocal osteomyelitis [21] and bone tumors [22‐24] but not for acute ostearticular infections.

The objective of our study was to assess whether DWI can replace contrast injection for the diagnosis of osteomyelitis, septic arthritis and abscess in children.

This retrospective study was conducted in a single pediatric radiology center from January 2015 to March 2020 and was approved by our hospital’s local ethics committee (CE 14-102R).

We reviewed all consecutive MRIs performed before treatment of children admitted for suspected acute peripheral osteoarticular infection. Children with known hematological or oncological disease were excluded. Electronic medical records were searched for clinical and biological data.

MRI were performed on a 1.5-T unit (Avanto; Siemens, Erlangen, Germany) using the following protocol: T1-weighted (T1-W) turbo spin-echo (one longitudinal plane); two orthogonal planes with T2-weighting plus fat suppression (T2-W FS), short tau inversion recovery (STIR) (longitudinal plane) and water-only data set of fast spin-echo T2-weighted Dixon sequences (axial plane); DWI (axial plane) using single shot echo planar imaging (EPI), acquired using two b values, 0 and 800 s/mm², with automatic monoexponential calculation of apparent diffusion coefficient (ADC) maps; and post-contrast injection T1-weighted spin echo with frequency-selective fat saturation (two orthogonal planes). Post-contrast sequences were obtained after injection of 0.2 ml/kg of gadoteric acid (Dotarem, Guerbet, France).

In young children, it may be difficult to localize the origin of infection because of ill-defined and nonspecific clinical symptoms. Therefore, MRI consistently started with a 3-D extended field of view (FOV) STIR to identify the anatomical site of signal abnormality. Subsequent sequences were acquired with a smaller FOV and with a dedicated peripheral coil targeted at the region of interest.

The approximate duration of MRI was 30 min, of which 15 to 20 min were dedicated to T1-W and T2-W FS sequences, 2 to 4 min were dedicated to DWI acquisition and 15 min to contrast injection, i.e. monitoring of venous catheter, manual injection and acquisition in two planes (acquisition time range: 7–11 min). General anesthesia was performed for children younger than 6 years old. Online Supplementary Material 1 includes a table with acquisition parameters depending upon the FOV dimension chosen to fit the anatomical location.

Magnetic resonance imaging is the modality of choice for investigating acute osteoarticular infections in children, decision-making about antibiotic therapy and surgical management [4‐6].

Previous pediatric studies showed that T1-W and fluid-sensitive (STIR and T2-W with fat suppression) sequences are sufficient to confirm or exclude osteomyelitis and arthritis, but gadolinium injection remains the reference standard to identify complications of osteomyelitis such as abcess, and growth cartilage and chondroepiphyseal injury [9, 10, 13, 14, 28]. Averill et al. [13] retrospectively studied the MRIs of 78 children with suspected osteoarticular infection and demonstrated equal sensitivity and specificity between pre-contrast and post-contrast images for the diagnosis of osteomyelitis. However, reader confidence for abscess was increased with contrast injection. Similarly, a more recent analysis of 90 pediatric MRIs by Markhardt et al. [14] showed that fluid-sensitive sequences were as sensitive as contrast-enhanced sequences with a 100% negative predictive value for osteoarticular infection, although the diagnosis of concomitant abscess was missed in three cases without contrast injection. Other authors reviewed MRI studies of 14 epiphyseal osteomyelitis cases in children in correlation with surgery [28]. Among the 10/14 abnormally enhanced epiphyses, 4 corresponded to either chondroepiphyseal abscess or injury without abscess, highlighting that an enhancement defect may indicate chondroepiphyseal involvement. In line with these studies, current MRI guidelines recommend witholding gadolinium injection when pre-contrast sequences are normal, and reserving contrast medium for suspected abscess and/or epiphyseal involvement [9, 10].

To avoid gadolinium injection, we evaluated the diagnostic performance of an unenhanced protocol with DWI in comparison with a standard contrast-enhanced MRI. To our knowledge, DWI is not currently part of routine pediatric MRI protocols for osteoarticular infection in most centers, although recently, Chaturvedi [29] reported the usefulness of DWI to assess for subperiosteal abscess in children based on unpublished data. So far, only adult studies have emphasized that ADC values can differentiate osteomyelitis from noninfectious bone marrow edema [17, 30, 31].

In our study, all cases of osteomyelitis and arthritis were correctly identified by unenhanced sequences. The addition of DWI increased the sensitivity of unenhanced sequences for the detection of abscess, by demonstrating bright signal on high b value images with corresponding low intensity on ADC map. Interestingly, we report one 11-year old boy with ankle osteomyelitis complicated by bone necrosis for which DWI showed restricted diffusion, but contrast-enhanced sequences did not confirm bone abscess. We considered this case a false-positive bone abscess on DWI. Importantly, in the absence of contrast medium, this misdiagnosis would not have altered the patient’s management. In either case, bone abscess and bone necrosis, surgical debridement is necessary to improve antibiotic therapy efficiency [13].

Another important issue shown by our study is that DWI is less sensitive in detecting femoral head chondroepiphysial involvement in children younger than the age of 30 months. This age group has an increased propensity for chondroepiphyseal infection [9, 26, 30]. The femoral head frequently does not enhance after contrast injection in hip infection. The reason for this peculiar behavior on MRI remains controversial and may be related to direct infection or to ischemic phenomena concomitant with joint infection [32‐34]. In our cohort, four diseased femoral heads were missed on non-enhanced and DWI images. There is an age-related increase in fat marrow content and parallel decrease of enhancement of the epiphyses [35]. This may partly explain why non-contrast images are less sensitive to abnormal epiphyseal signal in the younger patient, whose epiphyses contain less fat. Moreover, the small size of growing epiphyseal ossification centers may negatively affect spatial resolution and render difficult the detection of any abnormal signal. Similarly, we presumed that DWI could not identify altered signal of the femoral head chondroepiphyses because of their fatty, marrow-rich and low proton density content, which makes them appear normally hypointense on DWI and ADC map [15, 36]. In our series, the other seven epiphyses showed abnormal signal on pre-contrast sequences and therefore contrast-enhanced sequences had no added value for the diagnosis of chondritis and osteomyelitis.

Our study is limited by its retrospective nature and its small sample size. However, it is notable for the very young age of our cohort (72% were younger than 5 years old) for whom MRI, without contrast injection, could be of increased interest. One reason for the limited number of patients is that we already regularly dispense with contrast medium in MRI studies when pre-contrast and DWI sequences are informative enough to enable the decision for surgical management and to guide the procedure. The retrospective nature of our work, along with the absence of a control group, could have been the source of bias by positively influencing the interpretation in cases of uncertainty. Regarding DWI, we only analysed images on a qualitative basis, as either increased or decreased signal with visual correlation to the ADC map, without measuring ADC values. Although our inter-reader agreement was high, the visual analysis of DWI may lead to subjective interpretation in ill-trained radiologists. It may be even more difficult when small bones of the extremities are examined, where cumulative dephasing from increased K space filling to increase resolution can be responsible for image distorsion. In our study, we chose the axial plane for DWI to minimize geometric distortion.

Another limitation of our study relates to its reproducibility in other institutions. In our pediatric hospital, we are able to perform an emergent MRI any time septic arthritis or osteomyelitis is suspected, but we are aware that MRI and general anesthesia are not always available during night shifts and out-of-hours in every hospital. Finally, we emphasize that our study was designed to test DWI as part of an MRI protocol for invstigating osteoarticular infection, and not to differentiate infection from alternative diagnoses, such as Langerhans cell histiocytosis.

Our study acts as a reminder that all MRI protocols should be tailored to each individual patient. The decision to inject contrast medium should be undertaken while the unenhanced images are acquired. To that end, we encourage the routine addition of DWI to standard unenhanced sequences for the work-up of osteoarticular infections in children and suggest the following recommendations:

In addition, contrast-enhanced sequences should be maintained for suspected proximal femoral head involvement especially in young children, irrespective of the suspected underlying mechanism (direct infection, transient ischemia) since it impacts decision-making regarding joint aspiration, duration of antibiotic therapy and clinical follow-up. It would be useful to perform multicenter studies in a larger cohort to obtain data that would support our recommandations and our belief that contrast can be avoided in many children with this common acute pathology.

Emergency unenhanced MRI with DWI can establish the diagnosis of acute osteoarticular infections in children and negates the need for gadolinium. Such a protocol is sufficient for prompt medical and surgical management. However, gadolinium-based contrast injection remains necessary to evaluate the femoral head chondroepiphysis in younger children, in whom diffusion imaging is not reliable.