To compare the quantitative and qualitative information obtained by Doppler ultrasound (US) measurements of the wrist joints and the small joints of the hand with the information obtained by postcontrast magnetic resonance imaging (MRI) and to correlate the imaging results with clinical observations in patients with rheumatoid arthritis (RA).
Twenty-nine consecutive RA patients were studied; 196 joints (29 wrist and 167 finger joints) were examined by both US and MRI. Parameters of inflammation were the color fraction and the resistance index (RI) obtained with color Doppler US and the thickness of enhanced synovium (in mm) and the MRI score obtained with postcontrast MRI. Clinical examination and measurements of the erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) level were performed on the same day as the imaging studies.
There was a highly significant association between US indices of inflammation and postcontrast MRI scores. The mean values for both the color fraction and the RI were significantly different in the group without joint swelling compared with the other groups. The mean RI values were significantly different in the group without joint tenderness compared with the other groups. The mean thickness of enhanced synovium on postcontrast MRI was significantly different between the group without joint swelling and the other groups, but this difference was statistically significant only for the comparison of the group without joint tenderness versus the group with maximum tenderness. No association between the MRI or US estimates of inflammation and values on the visual analog scale for pain, Health Assessment Questionnaire, duration of morning stiffness, ESR, or CRP was found.
Estimates of synovial inflammatory activity by Doppler US and postcontrast MRI were comparable. Estimation of synovial inflammatory activity by the RI and color fraction parameters of US appears to be a promising method of detecting and monitoring inflammatory activity in patients with RA.
Optimum management of rheumatoid arthritis (RA) requires early diagnosis and timely introduction of agents that reduce the probability of irreversible joint damage (1). Studies have suggested that early aggressive treatment may alter the disease course (2–4), and close monitoring for signs of persistent synovitis and treatment failure is required. Synovial inflammation is accompanied by hyperemia, and during disease progression, angiogenesis in a hypervascularized pannus appears to be a prerequisite for damage to cartilage and bone (5, 6). Thickening of the inflamed synovium reflects the presence of vascular congestion, edema, cellular infiltration, and pannus (7).
Magnetic resonance imaging (MRI) has been proposed as an imaging modality for estimating inflammation in RA. International guidelines for the use of MRI in RA have been developed by the Outcome Measures in Rheumatology Clinical Trials (OMERACT) group (8). It has been shown that MRI of even the small joints after injection of a contrast agent yields a quantitative estimate of inflammatory changes in the synovium (9, 10) and is able to differentiate between inflamed synovium and inactive fibrous synovium. The enhanced synovial membrane seen on postcontrast MR images is believed to represent the highly vascularized inflamed synovial tissue, which can therefore be used as an estimate of synovial inflammation (11, 12).
The thickness of enhanced synovium (in mm) together with the degree of synovial inflammation on postcontrast MRI (by semiquantitative scoring) are the two standard ways of describing the degree of inflammation (8). Although MRI is a potentially powerful technique in the evaluation of inflammation, it is currently not a routine procedure due to limited availability. Ultrasound (US) technology, which may be available in outpatient clinics, has no contraindications, and poses no problems with regard to patient compliance, can also be used to obtain a valid estimate of inflammatory activity.
US is now used for the detection of synovial changes in RA and has been validated as a measure of hyperemia in RA inflammation. Previous studies have shown that Doppler US can detect synovitis in the small joints (13–15), and a quantitative estimate of the degree of inflammation can be obtained with the use of color or power Doppler by measuring the fraction of color pixels in the region of interest (ROI) and evaluating the changes over time (16, 17). Spectral Doppler US may also be used to obtain a quantitative estimate of the degree of inflammation by evaluating abnormal perfusion in the inflamed synovium, using the flow profile of the vessels visualized in the inflamed synovium by color or power Doppler.
The aim of the present study was to compare the quantitative and qualitative information obtained by Doppler US measurements in the wrist joint and small joints of the hand with static postcontrast MR images. We also correlated these imaging results with clinical observations.
PATIENTS AND METHODS
Twenty-nine consecutive RA patients (17 women and 12 men) were included in the study. Their mean age was 56 years (range 25–83 years), and their mean disease duration was 7 years (range 1–27 years). All patients fulfilled the American College of Rheumatology (formerly, the American Rheumatism Association) 1987 criteria for RA (18).
In all patients, one wrist joint was examined by both MRI and US. The finger joints of the same hand (metacarpophalangeal [MCP] and proximal interphalangeal [PIP]) were included in the study if they were evaluated by both techniques. A total of 196 joints were examined: 29 wrist, 91 MCP, 74 PIP, and 2 first carpometacarpal (CMC) joints.
The study was approved by the local ethics committee. Informed consent was obtained from each patient.
All joints were assessed clinically by the same investigator (LT). Each joint was scored on a scale of 0–3 for the degree of tenderness and swelling. Each patient completed a Health Assessment Questionnaire (HAQ), estimated his or her level of pain with the use of a 0–100-mm visual analog scale (VAS), and reported the duration of morning stiffness (in minutes). Erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) levels in each patient were obtained on the same day as the imaging studies.
The joints of the hands were examined with an Acuson Sequoia US instrument (Mountain View, CA) equipped with a 15-MHz linear array probe. The patient was examined while sitting upright, with the hand placed on a cushion and fully pronated. The dorsal side of the wrist was scanned from side to side in the longitudinal plane and from superior to inferior in the transverse plane. The finger joints were scanned in the longitudinal plane only; palmar aspects were not investigated. The first CMC, first MCP, and all PIP joints were scanned in a 180° arc, from the ulnar side to the radial side. The second through fifth MCP joints were scanned in the regions that were accessible from the dorsal side: in a 150° arc for the second and fifth MCP joints, and in a 120° arc for the third and fourth MCP joints.
The color Doppler settings were the same for all patients and all joints, with a gain setting just below the noise level, using our setup for low flow: Nyquist limit ±0.014 m/second and 7-MHz Doppler frequency. With this setup, all the color pixels in the image correspond to motion, i.e., blood flow. We used color Doppler rather than power Doppler because, at present, both modalities have the same sensitivity on the Sequoia US instrument.
Synovial vascularization in the joints was visualized by color Doppler, and the image with maximum color activity (if any) was selected for analysis. The US flow pattern for the synovium in the joints with activity on color Doppler was evaluated with quantitative spectral Doppler, with automatic calculation of the resistance index (RI). The RI is defined as follows: ([peak systolic velocity – end diastolic velocity]/peak systolic velocity). The RI was determined in 3 arteries within the synovial membrane, when possible, and the mean value served as an estimate of synovial inflammation. We used a maximum value of 1.00 for the RI, because the analysis was limited to one side of the Doppler baseline. The reason for this is that we sample small vessels and, most often, we sample the artery and its concomitant veins simultaneously. The negative part of the arterial signal will thus be drowned in the venous signal. When spectral Doppler measurements could not be measured due to lack of detectable vascularization in the examined wrist, the RI was recorded as 1.00, since the resistance in the synovial arteries was presumed to be the same as extrasynovial musculoskeletal flow. The examination time for the wrist and finger joints of each patient was ∼15 minutes for the color Doppler procedure and an additional 15 minutes for the spectral Doppler procedure, depending on the number of available vessels.
Evaluation of US images.
A quantitative estimate of the vascularization of the synovial membrane was performed using the color Doppler image. The maximum color activity was selected for analysis. The digitally stored color Doppler image in DICOM format was transferred to a processing program (PhotoPaint 7; Corel, Ottawa, Ontario, Canada). The synovium inside the color box was outlined to define the ROI (Figure 1). By using a color recognition function, the number of color pixels was then expressed in relation to the total number of pixels in the marked ROI (19). This value represents the color fraction.
A qualitative estimate of inflammation was also performed with color Doppler. Since no thresholds between normal and pathologic conditions have been established for US assessments, a rigorous definition of “normal” was used in order to avoid subjective and possibly fluctuating thresholds. Therefore, a joint was defined as inflamed if any color pixels were found in the ROI and as uninflamed when no color pixels were present in the ROI.
MRI of the wrists and finger joints was performed with a 1.5T MR system (Gyroscan ACS-NT; Philips, Best, The Netherlands). Patients were placed in a prone position with the hand above the head, and a 20-cm circular-surface receive-only coil was placed over the hand. The position was maintained and movement avoided with the aid of sand bags. Three-dimensional coronal T1-weighted fast field-echo images of the hand were obtained using the following technical parameters: repetition time 25–26 msec, echo time 4.6 msec, flip angle 50°, slice thickness 1 mm, gap 0 mm, matrix size 256 × 256, and field of view 220 mm. Examination time was 6–8 minutes. Voxel size was 0.86 × 0.86 × 1 mm.
While the patient remained motionless in the MR unit, 0.05 mmoles of gadodiamide (Omniscan; Nycomed Amersham, Oslo, Norway) per kilogram of body weight was injected into a vein in the contralateral arm via a cannula that had been inserted before the examination. The T1-weighted 3-dimensional fast field-echo sequence was then repeated within 30 seconds of the injection. The 3-dimensional volume scan was used to reconstruct axial images.
Evaluation of MR images.
Synovitis in a joint on MRI was defined as an area in the synovial compartment that showed enhancement on the postcontrast image. On the axial T1-weighted postcontrast scan, we selected the slice showing the most thickened area of synovium, and the maximum thickness of the enhanced synovium was then measured (in mm) (20).
The degree of synovitis in each joint examined was graded on a scale of 0–3 according to the OMERACT definitions (8). A joint was defined as inflamed (grade >0) if any detectable enhancement was present. While OMERACT recommends measurements of the wrists and the first through fifth MCP joints only, we also performed estimates of the PIP and first CMC joints using the same procedure. All images were evaluated under blinded conditions.
Statistical analysis was performed with the Statistical Analysis System program (SAS Institute, Cary, NC). The influence of joint tenderness and swelling on the color fraction, mean RI, and MRI grading of the different joints were tested by one-way analysis of variance as well as for trends (general linear model) since the groups were of unequal numbers. Student's t-test for unpaired data was used to compare the individual groups for the joint assessments. All test were 2-tailed, and P values less than 0.05 were considered significant. A maximum score for swelling was found in only 3 joints; therefore, joints scored 2 and 3 were pooled for analysis.
Agreement between the 2 imaging modalities was tested using the chi-square test and kappa statistic. Spearman's rank correlation coefficient was used to examine correlations between the 2 imaging modalities.
Findings of US evaluation of all 196 joints.
Doppler US revealed inflammatory activity in 52 of the 196 joints examined (Table 1). All 3 joint swelling groups differed significantly in both the color fraction and the mean RI values, whereas only the mean RI was significantly different in all 4 joint tenderness groups (Figures 2A and B). There was no correlation between the US estimates of inflammation and either the VAS for pain, the HAQ score, the duration of morning stiffness, the ESR, or the CRP value.
Table 1. Agreement between MRI and ultrasound findings in all joints as well as in the joint subgroups*
Postcontrast MRI revealed inflammatory activity in 79 of the 196 joints examined (Table 1). The postcontrast MRI estimate of inflammation of the synovial membrane (synovial thickening [in mm]) discriminated between the 3 joint swelling groups, but was able to discriminate between only 2 of the joint tenderness groups, those without tenderness and those with pronounced tenderness (Figure 3). There was no correlation between the MRI scores for inflammation (both the MRI score and synovial thickness in mm) and either the VAS score for pain, the HAQ score, the duration of morning stiffness, the ESR, or the CRP level.
The 2 MRI scores for inflammation correlated closely with each other (rs = 0.98, P < 0.001). Therefore, synovial thickness (in mm) was used throughout this study in statistical evaluations of the 2 imaging modalities.
Comparison of US versus MRI.
Inflammation in the joints was estimated according to the presence of enhancement on postcontrast MRI and the presence of color pixels on color Doppler US. There was total agreement between the 2 imaging modalities for 75% of the 196 joints examined (Table 1), with a kappa value of 0.45.
There were statistically significant correlations between the synovial thickness on postcontrast MRI and both the color fraction (rs = 0.59, P < 0.001) and the mean RI (rs = –0.54, P < 0.001) on color Doppler US. In the group of joints without enhancement on postcontrast MRI, the mean RI and the mean color fraction were significantly higher and significantly lower, respectively (P < 0.0001), compared with the group of joints with enhancement on postcontrast MRI (Table 2).
Table 2. Agreement between US parameters and postcontrast MRI parameters in all 196 joints*
Color Doppler US parameters
Resistance index, mean ± SD
Color fraction, mean ± SD
US = ultrasound; MRI = magnetic resonance imaging.
0.86 ± 0.19
0.14 ± 0.16
0.99 ± 0.07
0.01 ± 0.03
The mean RI values were significantly different in all joint tenderness and joint swelling groups compared with the respective grade 0 groups (no joint tenderness and no joint swelling) (Figure 2B). Both the color fraction and the postcontrast MRI score were significantly different in all joint swelling groups versus the group without swelling. With regard to joint tenderness, both the color fraction and the postcontrast MRI were significantly different only between the group with grade 3 tenderness and the group with grade 0 tenderness (Figures 2A and 3).
The comparisons shown in the tables represent only the differences in relation to the respective grade 0 groups (no joint tenderness and no joint swelling). However, for all 3 assessment parameters, there was a statistically significant difference for grade 1 versus grade 2 joint swelling (P < 0.0001 for each comparison). There was also a statistically significant difference for grade 1 versus grade 3 and for grade 2 versus grade 3 joint tenderness (P < 0.0001 and P < 0.01 for the mean RI, P < 0.0001 and P < 0.001 for the color fraction, and P < 0.001 and P < 0.01 for the postcontrast MRI). There was no statistically significant difference between grade 1 versus grade 2 for any of the parameters.
Comparison of the 2 imaging modalities with the clinical assessment.
Table 3 shows the associations between findings on the 2 imaging modalities and findings on the clinical assessment. Despite the absence of both tenderness and swelling in 123 joints on clinical assessment, inflammatory activity was identified by both imaging techniques in 10 joints. In the 43 clinically tender and swollen joints, both imaging techniques revealed an absence of inflammation in 5 joints.
Table 3. Agreement between the imaging modalities and the clinical assessment of joint tenderness and swelling*
No activity by ultrasound
Activity by ultrasound
Total no. of joints
MRI = magnetic resonance imaging.
Clinically nontender, nonswollen joints (n = 123)
No activity by MRI
Activity by MRI
Total no. of joints
Clinically tender and swollen joints (n = 43)
No activity by MRI
Activity by MRI
Total no. of joints
Findings of US and MRI evaluation of individual joint groups.
The agreement between the 2 imaging modalities for the individual joint groups was the same as for the pooled joints (79% for the wrist joints, 74% for the MCP joints, and 74% for the PIP joints). The kappa values for the wrists and MCP joint groups (κ = 0.45 and κ = 0.41, respectively) were similar to the kappa value for the pooled joints, but the kappa value was lower for the PIP joint group (κ = 0.17). The analysis of the mean values for the 3 imaging parameters in the different joint swelling and tenderness groups was estimated for the wrist, MCP, and PIP joint groups, and the same tendency as for the pooled joints was found for both the MCP and PIP groups, but not for the wrist group. This is possibly due to the low number of joints in the different joint tenderness and swelling groups (data not shown). Because of the larger number of joints and because the analyses of the wrist, MCP, and PIP joint groups showed the same tendency as the pooled joints, we focused on the pooled joints in this study.
Findings of joint evaluation in only the joints recommended by the OMERACT guidelines.
The OMERACT guidelines recommend evaluation of only the wrists and the first through fifth MCP joints. When we evaluated only these joints, we found similar associations, and the agreement between the 2 imaging modalities was the same (data not shown).
In the present study, both a qualitative estimate and a quantitative estimate of inflammatory changes in the synovium of the hands in 29 RA patients were performed by US and MRI. The imaging parameters assess different aspects of inflammation. Inflammation on MRI is seen as enhanced, thickened synovium. Enhancement is caused by an increased presence of gadolinium in the synovium, which is primarily caused by intravascular gadolinium (hyperemia) as well as by gadolinium leaking into the extravascular space (increased vessel permeability). We used the MRI methods recommended by OMERACT (8), in which, in a given joint, the slice with the thickest synovium on postcontrast images is selected for analysis. An MRI score (0–3 scale) is assigned, and the synovial thickness is measured (in mm). Accordingly, inflammation can only be diagnosed by MRI if the enhanced part of the synovium has a thickness greater than the thickness of the joint capsule.
With US, the focus is on the hyperemic part of the inflammatory process. Gray-scale US gives valuable information about accessible bone surfaces, tendons, and joints. This method can reveal synovial hypertrophy and synovial fluid (21), but inactive and active synovial hypertrophy cannot be distinguished. When color Doppler is added to the US technique, information about the vascularization of the synovium (as a sign of inflammation) can be obtained (13, 17, 19, 22), and it may help in depicting the areas of synovial inflammatory activity (23) (Figure 1). The presence of color was interpreted as representing the presence of pathologic changes in this study, and we did not have any problems with the interpretation of the images. In future studies, where we foresee working with a specific cutoff level of color fractions, interpretation problems may arise.
On US, artifacts caused by noise, motion, as well as highly reflective interfaces, such as bony surfaces and tendons, may occur. Artifacts may be distinguished from true flow by the stable location of the artifact, whereas arterial flow displays pulsation. The aliasing artifact is unimportant in this setting, since with color Doppler, we are only interested in the presence or absence of arterial flow. However, false-negative US findings can occur if the examiner presses too hard against the tissue with the transducer, thereby blocking the arterial flow.
We selected for analysis the US image that had the most color pixels inside the synovium. The presence of color was then quantitated according to the percentage of the entire synovial area that contained color pixels, with a higher percentage indicating more severe inflammation. The intra- and interobserver intraclass correlation coefficients for this method were determined in a previous study to be 0.82–0.97 and 0.82, respectively (19).
Our second quantitative US parameter, the resistance index, is only applicable when vessels are present on the color display and when at least 1 artery among the vessels presents a signal above the threshold for sampling with spectral Doppler (color Doppler is more sensitive than spectral Doppler). The RI focuses on the type of flow. The normal high-resistance flow in a resting extremity has zero flow or negative flow as an end diastolic value, in which case the RI is 1.00. The RI diminishes as the relative difference between systolic and diastolic flow diminishes (decreasing peripheral resistance). Thus, a lower RI in rheumatology is interpreted as the presence of more severe inflammation, as has been indicated in previous studies (19, 24–26); however, a definite cutoff level for abnormality remains to be established (25). A flow profile characterized by a diastolic component is normally present in renal, cerebral, and umbilical arteries, in which low peripheral resistance ensures a permanent and high perfusion. In normal resting musculoskeletal tissues, it is possible to use color or power Doppler to visualize flow in muscles, connective tissue, and, sporadically, in joints. If the flow type is assessed in these tissues, no diastolic flow will be found, and therefore the RI is 1.00.
Our color Doppler settings mostly exclude noise artifacts but include blooming artifacts. Our settings are adjusted for the detection of low flow, which means that we do not apply a filter, and we use a low repetition frequency. If we chose to lower the Doppler gain and thereby minimize the blooming artifacts, we would lose the weakest signals, and the smallest vessels would go undetected. We regard the blooming artifact as a systematic error and accept its presence, knowing that all color pixels in our image are generated by flow.
The finding of 75% agreement between the 2 imaging modalities for all joints in our study indicates a parallelism between MRI and US, representing the 2 most feasible imaging techniques for the determination of synovial inflammatory activity in arthritis. The kappa statistics showed a moderate value of 0.45 between the presence of synovial enhancement on postcontrast MRI (a measurable enhancement of synovial thickness [in mm]) and the presence of color pixels on US, which is consistent with the experience of other investigators performing musculoskeletal imaging (27, 28). The correlation coefficients between the MRI score and the US parameters were statistically significant, indicating that the 2 imaging modalities are both detecting inflammation, although they are describing different aspects of that inflammation.
More inflamed joints were detected by MRI than by US. As shown in Table 1, the disagreement was mainly for the PIP joints (the lowest kappa value). We have previously described the presence of synovitis-like changes in normal joints (29). This was also observed by other investigators in a recent study in which static MRI detected synovitis-like changes in the MCP joints (12 of 112 joints) and wrist joints (12 of 28 joints) of healthy control subjects (30). With the use of dynamic MRI (or the relative enhancement per second during the first 55 seconds postinjection ), no active synovitis was found in these subjects. Some of the disagreement between the 2 techniques may also be caused by false-positive US findings. For the present study, we defined inflammation as the presence of even minute flow in the synovium (Figure 4). We did this because of a lack of a defined cutoff level. Thus, inflammation was diagnosed based on US findings in cases in which it would not have been diagnosed based on clinical findings. On closer examination of the 11 cases in which there were positive US results but negative MRI results, we found that 6 cases were examples of the presence of minute flow. In the remaining 5 MRI-negative cases, however, US definitely showed hyperemia, and we cannot account for the discrepancy.
Our results were obtained with very high sensitivity Doppler techniques, and the findings approach the boundary between normal and abnormal flow, as demonstrated in the elbow joint (32). We believe that our study exemplifies the difficulties in standardizing US techniques. The best equipment will occasionally pick up normal intraarticular flow and will require a definition of thresholds to distinguish normal from abnormal findings. In contrast, with older US equipment, underdiagnosis is a problem.
With regard to the clinical results, we found associations between the inflammatory indices as defined by both US and postcontrast MRI (enhanced thickness of the synovium) and local clinical parameters (degree of tenderness and swelling in the joints). No correlation with global parameters of disease activity can be expected (33).
We believe that both MRI and US have the ability to diagnose subclinical cases of arthritis. However, our findings raise the question as to whether a clinically nontender, nonswollen joint is in fact inflamed when the MRI and/or US features indicate the presence of inflammation. In the present study, there were 123 clinically nontender, nonswollen joints; 31 of these joints were positive by MRI, 9 were positive by US, and 5 were positive by both MRI and US (Table 3). Thus, while current clinical standards categorized all 123 joints as uninflamed, the US and MRI findings indicate that at least some of them were inflamed.
The discrepancies between the US and MRI findings remain a problem. Again, whether inflammation is present or not cannot be answered in the absence of a gold standard for these 2 imaging techniques. Doppler ultrasound and postcontrast MRI estimations of synovial inflammatory activity yielded comparable results. However, synovial inflammatory activity estimated by the mean RI and the color fraction may offer a more precise measurement of disease activity and will allow for easier monitoring of changes in the degree of inflammation.
Both imaging techniques challenge clinical evaluations of both the degree and the presence/absence of inflammation and define cases that are very probably misclassified with present clinical standards. The agreement between the quantitative estimates by MRI and US was 75%; however, the kappa value was moderate. Further studies comparing high-end US with dynamic MRI are needed to establish the diagnostic capabilities of these techniques.
We thank Nycomed Amersham (Oslo, Norway) for supplying the contrast agent and Acuson (Mountain View, CA) for technical support.