Dynamic contrast-enhanced magnetic resonance imaging of the wrist in children with juvenile idiopathic arthritis

Between March 2012 and July 2013, 32 consecutive children with JIA were prospectively included. Patients were recruited from one of three tertiary outpatient clinics. Institutional Review Board approval was obtained and informed consent was acquired from all parents, as well as from patients older than 12 years of age.

Three sets of criteria were used to classify the patients: diagnosis of the JIA subtype was made based on the International League of Associations for Rheumatology criteria (e.g., oligoarticular persistent, enthesitis-related arthritis), binary classification of disease activity was made based on the Wallace criteria (i.e. clinically active and clinically inactive) and a quantification on a continuous scale of disease activity was performed by means of the Juvenile Arthritis Disease Activity Score [26‐28].

All patients either had current wrist involvement (clinically active) or previous wrist involvement (clinically inactive). The indication to perform an MRI was always clinically driven. For example, in the case of children with clinically active JIA, the MRI was made to confirm synovial inflammation in the wrist. An example of MRI indication in a clinically inactive patient was to evaluate whether treatment could be downscaled in case of absent synovial inflammation on MRI.

Exclusion criteria were a history of intra-articular corticosteroid injection in the wrist within the last 6 months, the need for anesthesia during the MRI examination and general contraindications for MRI.

Clinical assessment was performed by one of the experienced pediatric rheumatologists (JMvdB 12 years, DS 5 years, KD 15 years, MvR >20 years) and consisted of a 67 active joint count defining presence of swelling, pain on motion/tenderness and limited range of motion. Visual analogue scales (100 mm) were used for the physician’s global assessment of overall disease activity, patient’s global assessment of overall well-being and a patient’s pain assessment. The Childhood Health Assessment Questionnaire was used to evaluate patients’ functional abilities [29]. Laboratory tests included erythrocyte sedimentation rate and C-reactive protein level. Clinical disease activity was reflected as a continuous variable by the Juvenile Arthritis Disease Activity Score, which ranges from 0 to 40 and is calculated by the sum of the following clinical variables: the physician and parent/patient global assessment of disease activity (both ranging from 0 to 10), reduced 10-active joint count and a normalized value of erythrocyte sedimentation rate to a 0-10 scale [28].

MRI was performed using an open-bore 1.0-Tesla MRI scanner (Panorama HFO; Philips Medical Systems, Best, The Netherlands) and a Sense wrist coil (Philips Medical Systems, Best, The Netherlands, 3 channels) with the child in a supine position with the arm along the side of the body [30]. No sedation was used. The most clinically affected wrist (present for the active patients or previous for the inactive patients) was imaged. The standard MRI protocol for the JIA wrist included the following sequences: before injection of intravenous contrast, a coronal short-tau inversion recovery sequence; coronal T1-weighted sequence, axial T1-weighted sequence and axial T2-weighted sequence with fat saturation. After the dynamic contrast-enhanced MRI (performed during injection of intravenous contrast agent), a coronal T1-weighted sequence and axial T1-weighted sequence with fat saturation were performed [31]. The dynamic contrast-enhanced MRI sequence was a 3-D T1-weighted fast field echo dynamic sequence, consisting of 28 dynamic scans and 31 slices per scan, with a temporal resolution of 15.5 s. Furthermore, the dynamic contrast-enhanced MRI sequence had an in-plane field of view of 80 × 100 mm and a slab thickness of 46.5 mm, with voxel size 1.0 × 1.0 mm × 1.5 mm, repetition time of 9.9 ms and an echo time of 6.9 ms. After 3 dynamic scans, intravenous contrast agent (0.1 mmol/kg bodyweight of Gadovist; Bayer Schering Pharma, Berlin, Germany) followed by a chase of 15 ml saline solution was injected through a 22-gauge needle with an injection speed of 3 ml/s by an automated injection device (Medrad, Warrendale, PA).

In-house written codes developed in Matlab (Mathworks, Natick, MA) were used for movement registration, drawing of region of interests and extraction of the dynamic contrast-enhanced MRI parameters. We performed non-rigid registration of all dynamic volumes to correct for possible small, residual movements of the wrist. For this movement registration, we adopted a Discrete Cosine Transformation model [32] with two subsequent cut-off bases of 50 and 25 mm to allow for global and local convergence.

A large anatomy-based region of interest was manually drawn on the axial maximal enhancement maps by a radiology trainee (3 years of experience with imaging in JIA), while using the axial post-contrast T1-weighted dynamic contrast-enhanced MR image as a reference. The region of interest was drawn to exclude visible vessels, tendons and the epiphysis of the radius and ulna (Fig. 1) because no synovium was expected. The three most proximal and three most distal slices of the dynamic contrast-enhanced MRI suffered from fold-over artifacts and were therefore excluded from the analysis. To maintain a consistent field of view in every patient, the 20 slices distal to the last slice on which the ulna appeared were considered suitable for analysis of dynamic contrast-enhanced MRI.

The dynamic contrast-enhanced MR images were analyzed with in-house developed software, running on Matlab [33]. Outcome measures were divided in time-intensity-curve shapes and conventional descriptive measures. Time-intensity-curve shapes represent the change in signal intensity per voxel over time. Although time-intensity-curve shapes do not provide quantitative measures, indicate tissue characteristics such as degree of vascularization, tissue viability, perfusion and volume of the interstitial space. Every voxel within the region of interest is classified independently in one of the following seven time-intensity-curve shapes with unique colors: non-enhancing (1, gray), slow enhancing (2, green), fast enhancing followed by either a plateau phase (3, blue), washout phase (4, purple) or gradual increase (5, yellow), arterial patterns (6, red) and all remaining patterns (7, white) (Fig. 2). The classification algorithm is described in detail in [21]. The time-intensity-curve shapes were then analyzed based on their relative frequency in the drawn region of interest: the number of every time-intensity-curve shape was calculated by dividing the absolute number of pixels presenting a specific time-intensity-curve shape type by the absolute total number of time-intensity-curve shapes 2-7 (i.e. all the enhancing pixels). Only time-intensity-curve shapes 2-5 were used for statistical analysis, because time-intensity-curve shapes 1, 6 and 7 are not relevant for disease activity in JIA.

While bones of the wrist joint were included in the region of interest, they do not contribute to the final result as the bone does not enhance on dynamic contrast-enhanced MRI and all the results were calculated only on enhancing voxels (types 2 to 7).

Conventional descriptive measures included the median value of maximum enhancement (maximum enhancement of all enhancing voxels within the region of interest with time-intensity-curve shapes 2-7), median maximum initial slope (maximum slope of increase of all enhancing voxels within the region of interest with time-intensity-curve shapes 2-7), median initial area under the curve (area under the enhancement curve between start of injection and 90 s after injection) and the enhancing volume (total volume in millimeters of all enhancing voxels within the region of interest with time-intensity-curve shapes 2-7).

Descriptive statistics were reported in terms of percentages means, medians, standard deviations and interquartile ranges. Normality plots were used to assess the normal distribution of the data. The Mann-Whitney U test was used to assess differences between clinical subgroups. A P-value <0.05 was considered statistically significant. A Spearman’s ρ correlation coefficient was used to assess the association between Juvenile Arthritis Disease Activity Score and dynamic contrast-enhanced MRI measures. Spearman’s ρ was interpreted as follows: <0.19 very weak, 0.20-0.39 weak, 0.40-0.59 moderate, 0.60-0.79 strong and >0.80 very strong. All statistical analyses were performed using SPSS version 20.0 (IBM SPSS, Chicago, IL).