Dynamic contrast-enhanced MRI in inflammatory arthritis, especially in conjunction with computer-aided analysis using appropriate dedicated software, seems to be a highly sensitive tool for monitoring the early inflammatory treatment response in patients with rheumatoid arthritis. This paper gives a review of the current knowledge of the emerging technique. The potential of the technique is demonstrated and discussed in the context of a case report following the early effect of an intra-articular steroid injection in a patient with rheumatoid arthritis flare in the knee.
Imaging modalities aiming to identify perfusion characteristics in inflammatory joint disease are receiving increasing attention after results from a recent publication have shown that measures of perfusion detected with ultrasound Doppler in the wrist joints of rheumatoid arthritis patients with low disease activity scores (DAS28) had the highest predictive value of future erosive outcome [
dynamic contrast-enhanced MRI (DCE-MRI) is such an imaging technique based on sequential acquisition of rapid MRI sequences before and during the infusion of a contrast agent. It can be used to evaluate synovial activity in patients with rheumatoid arthritis (RA) and has been shown to correlate closely to synovial vascularity and inflammation [
DCE-MRI has been tested on low-field [
Conventionally, DCE-MRI data is analysed using region of interest- (ROI-) based technique, where a small, few millimetre ROI is placed in the most enhancing part of the synovium, as perceived by an observer [
These issues have been addressed by application of a new technique for analysis of dynamic data developed by Kubassova et al. [
To use DCE-MRI data to monitor early changes in parameters of knee joint inflammation in a patient with a flair of RA following ultrasound-guided intra-articular injection of glucocorticoid (methylprednisolone acetate 40 mg/ml). The case will serve as an example of the technique and the changes seen will be discussed and explained in detail in order to give the reader a better understanding of the potential and pitfalls of using computer-aided analysis of DCE-MRI data. We hope that this paper could serve as an example of the potential of this methodology that can be further investigated in future larger studies.
This 52-year-old lady was affected by seropositive RA diagnosed 13 years before. The patient had side effects with several DMARDs, including methotrexate and was treated with prednisolone, 5 mg daily. Supplementary injections of methylprednisolone were given occasionally in joints with acute flares; the last intra-articular injection was performed in a wrist joint 10 months before the present treatment.
Clinical findings at baseline included a moderately swollen knee and slight-to-moderate joint pain with a 100 mm visual analogue scale of pain of 30 mm at rest and 50 mm on joint movement.
Joint aspiration yielded 25 cc of clouded synovial fluid, and after arthrocenthesis, 1.5 ml glucocorticoid methyprednisolone 40 mg/ml was injected in the lateral recess of the knee with almost complete resolution of symptoms within day 2 of injection and complete clinical remission at day 7. The effect lasted for 2 months. The patient had normal kidney function measured by serum creatinine and estimated glomerural filtration rate (e-GFR).
After informed and written consent, the patient had conventional static MRI as well as dynamic MRI performed on day 0, 1, 2, and 7 using a 0.2 T musculoskeletal extremity scanner (Esaote E-scan). The patient was examined in supine position with the knee positioned centrally in the receive-only cylindrical solenoid knee coil. The following pulse sequences were applied: gradient-echo scout, sagittal STIR (TR/TE/TI: 1310/24/85, fov/matrix: 200 × 170 mm/192 × 163, slice thickness 4 mm) and axial 3D T1 gradient echo (TR/TE: 38/16, fov/matrix: 180 × 180 × 100 mm/192 × 160 × 72, slice thickness 0.8 mm). After these images were acquired, an intravenous injection of 0.1 mmol/kg body weight Gadolinium-DTPA (Magnevist, Schering AG, Berlin, Germany) was administered over a period of 30 seconds. At the time of Gadolinium injection, 30 consecutive 5 mm axial gradient echo dynamic MRI (DCE-MRI) images (TR/TE 60/6, FOV/imaging matrix 160 × 160 mm/256 × 128) in three prepositioned planes were started and obtained every 10 second, covering the superior, medial, lateral, and posterior joint recesses in the knee. Image time was 300 seconds. Finally, the static axial 3D T1 gradient echo sequences were repeated. The acquisition time of each sequence ranged from 4 to 8 minutes, with one signal acquired. Total imaging time was 30 minutes.
The conventional static imaging data was displayed using an AGFA PACS system (Figures
STIR images showing an evident signal decrease from the joint cavity between baseline and day two, corresponding to a reduced joint effusion. The effusion in the images is marked with an asterix (*).
The dynamic enhancement pattern in the inflamed knee synovium was analyzed using the software Dynamika-RA (
Further, the data was analysed using the voxel-by-voxel-based approach, incorporated into the software, and the enhancement characteristics of each voxel was computationally mapped to one of 4 enhancement models [
Parametric maps derived from the same DCE-MRI dataset reflecting the Gadolinium behaviour and distribution over time (Gd), maximum enhancement (ME), initial rate of enhancement (IRE), and time onset of enhancement (T-onset). The maps are derived from the baseline images of the case study representing a flair of moderate arthritis activity of the knee. Note the pulsation artefacts of the popliteal artery that give false “hot” points in the image related to a horizontal line of the artery.
The vertical colour bars or the
The horizontal colour bar shows the number of voxels and their percentage of the total in ( ) for each statistic in the corresponding IMAP, for example, 1 (0.01%) or 248 (91%), and so forth. For more information, visit
After the movement correction and the fully automatic analysis of the knee joint was performed, several regions of interest (ROIs) were drawn (Figures
A rough ROI outlining the anterior part of the knee in the dynamic image guided by the ME maps and their corresponding dynamic curves from baseline through day 7 (a–d). Note the decrease of enhancing voxels in the parametric map over time as well as a simultaneous decrease of the slope in the corresponding dynamic enhancement curves. The
Parametric maps derived from DCE-MRI data maximum enhancement (ME) (a, b) and the corresponding static postcontrast 3D T1-w gradient echo images (c, d) from baseline and day 2. Arrows pointing at the enhancing synovial membrane in the post contrast images (c, d). Examples of the applied ROIs are shown in red in (a) representing the synovial ROI in front of the knee, the popliteal artery ROI behind the knee, and the muscle ROI (oval circle).
The conventional STIR images (Figure
ME and IRE statistics, extracted from the dynamic data of this case and generated for the whole joint, showed no significant changes in the days following the steroid injection (Table
Body part | Acq. Date | ROI name | Mean ME | Std. Dev. ME | Mean IRE | Std. Dev. IRE | ME normalized* | IRE normalized* | |||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
KNEE | Baseline | Fully automatic | 1.59 | 0.23 | 0.010 | 0.01 | 45 | 609 | 185 | 839 | 794 | 1262 | 7.94 |
KNEE | Day 1 | Fully automatic | 1.59 | 0.37 | 0.011 | 0.02 | 57 | 590 | 313 | 960 | 903 | 1436 | 9.93 |
KNEE | Day 2 | Fully automatic | 1.63 | 0.28 | 0.015 | 0.02 | 28 | 119 | 132 | 279 | 251 | 410 | 3.77 |
KNEE | Day 7 | Fully automatic | 1.72 | 0.40 | 0.024 | 0.03 | 48 | 118 | 89 | 255 | 207 | 356 | 4.97 |
KNEE | Baseline | Synovium | 1.59 | 0.23 | 0.008 | 0.01 | 29 | 531 | 85 | 645 | 616 | 979 | 4.93 |
KNEE | Day 1 | Synovium | 1.46 | 0.16 | 0.004 | 0.001 | 24 | 376 | 103 | 503 | 479 | 701 | 1.92 |
KNEE | Day 2 | Synovium | 1.43 | 0.09 | 0.002 | 0.001 | 16 | 39 | 23 | 78 | 62 | 88 | 0.12 |
KNEE | Day 7 | Synovium | 1.47 | 0.13 | 0.003 | 0.001 | 8 | 31 | 4 | 43 | 35 | 52 | 0.11 |
KNEE | Baseline | Vessel | 1.79 | 0.19 | 0.038 | 0.02 | 0 | 10 | 65 | 75 | 75 | 134 | 2.85 |
KNEE | Day 1 | Vessel | 2.06 | 0.39 | 0.041 | 0.02 | 0 | 10 | 56 | 66 | 66 | 136 | 2.71 |
KNEE | Day 2 | Vessel | 2.00 | 0.24 | 0.037 | 0.01 | 0 | 10 | 51 | 56 | 56 | 112 | 2.47 |
KNEE | Day 7 | Vessel | 2.14 | 0.36 | 0.054 | 0.03 | 0 | 9 | 61 | 70 | 70 | 150 | 3.78 |
*Normalized values: sum of total number of voxels with plateau and washout information multiplied by the ME and IRE, respectively.
We further outlined a rough ROI positioned to include the synovial membrane and to exclude the larger vessels especially behind the knee joint. There was no need to position ROI precisely, as the measurements were only done on the enhancing voxels inside the ROI (Figures
The IRE of the roughly outlined synovial ROI decreased from baseline values over the first two days by a factor of 4 and stayed in the low end at 1-week followup. The mean ME showed no significant reduction (Table
The dynamic curves and corresponding enhancement statistics from the vessel ROI including the popliteal artery remained relatively unchanged over time but showed a day to day variation (Table
In order to normalize the ROI data, we multiplied the sum of
When static post contrast T1 weighted MRI is used to monitor, the early inflammatory treatment response in patients with RA a change of up to 30% in enhancing volume is needed to imply a one step change in the inflammation score [
In contrast, DCE-MRI seems to be highly sensitive to the early treatment response, but even though the methodology of DCE-MRI has been known for several years, previous studies have reported problems with reproducibility of results due to large variations in the ROI analysis [
Fully automated data analysis of the whole joint revealed that the mean IRE and ME did not change significantly over time even though the number of enhancing voxels showed a dramatic decrease between day 1 and 2. The reason for this seems to be due to the confounding effect from the large vessels behind the knee, where the values of ME and IRE are the highest; thus, the enhancement changes in the synovial membrane are “shadowed” by the activity in the neighbouring vessels. On the other hand, making a rough ROI surrounding the synovial membrane, and thus removing confounding influence from the major blood vessels revealed a significant treatment response in the slope of the ROI curve that decreased by a factor of 4 between baseline and day 2 and remained in the same lower range at one-week follow-up. Based on these observations, we recommend to exclude the larger vessels from DCE-MRI analysis of the knee joint, which can be done by using a rough ROI and an appropriate software tool.
The interpretation of changes in DCE-MRI following intra-articular steroid administration is also potentially confounded by the heterogeneity of treatment response across the whole synovial tissue mass. Thus, as expected, the ME did not change following treatment, since the remaining voxels demonstrating enhancement in the follow-up examinations reached approximately the same level as observed in pretreatment images. However, it took a longer time to achieve the plateau because of a lower steepness of the slope (Figure
As the software uses a model-based enhancement classification, there is no need to apply a threshold, nor do we recommend to normalise the data to the enhancement characteristics of the vessels or the muscles, because there seem to be a relative large day to day variation in the ROI statistics.
We have examined the current patient 4 times within a week to measure the effect of the steroid injection. This approach cannot be recommended for routine clinical use for many reasons, including the use of i.v Gadolinium, expensive and time-consuming MRI examinations, availability of MRI scanners for RA patients. The case should be seen as an example of the potential of the technique, but based on our results which have to be confirmed in larger studies, we speculate that there could be a benefit of using the technique to get a more objective idea of the early treatment effect of the more expensive biologic treatments, that is, within 2–4 weeks of treatment, which could lead to better patient care by reducing the time spend on an ineffective treatment and in the long run money could be saved on the health economy budget.
This study has several limitations, as our findings are based on a case report, and we have used the knee joint to illustrate the potential of the technique, where ROI-based exclusion of the larger vessels is fairly easy. We cannot assume that our findings can be extrapolated to smaller and more complex joints like the wrist. The therapeutic intervention employed in this case was an intraarticular steroid injection known to have a potent anti-inflammatory activity, and we cannot assume that an observable treatment effect would be as pronounced and as rapid when using conventional DMARDS or even biologic therapies.
In conclusion, DCE-MRI in conjunction with analysis using appropriate software seems to be a highly sensitive tool for monitoring the early inflammatory treatment response in patients with RA, as demonstrated by the assessment of a knee joint inflammation following an intra-articular steroid injection. The decrease in IRE and ME at day two follow-up in this case example, especially seen in the normalized data, corresponded to improvement in the patient’s clinical symptoms. These findings have to be further tested in larger clinical trials on several joints to see whether the observed benefit in the current case using dynamic MRI may be used in general as a sensitive biomarker to track the early treatment response in patients starting potent anti-inflammatory treatments such as local/systemic steroid, and/or biologics.
This study was supported by the Oak Foundation and an unrestricted grant from Abbott Denmark.