Article Text

Download PDFPDF

Does low-field dedicated extremity MRI (E-MRI) reliably detect bone erosions in rheumatoid arthritis? A comparison of two different E-MRI units and conventional radiography with high-resolution CT scanning
  1. A Duer-Jensen1,
  2. B Ejbjerg2,
  3. E Albrecht-Beste3,
  4. A Vestergaard3,
  5. U Møller Døhn1,
  6. M Lund Hetland1,
  7. M Østergaard4
  1. 1
    Department of Rheumatology, Copenhagen University Hospital at Hvidovre, Denmark
  2. 2
    Department of Rheumatology, Copenhagen University Hospital at Frederiksberg, Denmark
  3. 3
    Department of Radiology, Copenhagen University Hospital at Hvidovre, Denmark
  4. 4
    Departments of Rheumatology, Copenhagen University Hospitals at Hvidovre and Herlev, Denmark
  1. Dr A Duer-Jensen, Department of Rheumatology, Copenhagen University Hospital, Kettegaards Allé 30, 2635 Hvidovre, Denmark; anne-duer{at}dadlnet.dk

Abstract

Objectives: To compare the ability of two different E-MRI units and conventional radiography (CR) to identify bone erosions in rheumatoid arthritis (RA) metacarpophalangeal (MCP) and wrist joints with CT scanning as the standard reference method.

Methods: 20 patients with RA and 5 controls underwent CR, CT and two E-MRI examinations (Esaote Biomedica Artoscan and MagneVu MV1000) of one hand during a 2-week period. In all modalities, each bone of the wrist and MCP joints was blindly evaluated for erosions. MagneVu images were also assessed for the proportion of each bone being visualised.

Results: 550 bones were examined. CT, Artoscan, MagneVu and CR detected 188, 116, 55 and 45 bones with erosions, respectively. The majority were located in the carpal bones. The sensitivity of the Artoscan for detecting erosions was higher than that of the MagneVu and CR (MCP joints: 0.68, 0.54 and 0.57, respectively; wrists: 0.50, 0.23 and 0.29). Corresponding specificities for detecting erosions were 0.94, 0.93 and 0.99, respectively, in the MCP joints and 0.92, 0.98 and 0.98 in the wrist. The MagneVu allowed visualisation of 1.5 cm of the ventral-dorsal diameter of the bone. In the wrist, 31.6% of bones were visualised entirely and 37.9% of bones were 67–99% visualised. In MCP joints, 84.2% of bones were visualised entirely and 15.8% of bones were 67–99% visualised.

Conclusion: With CT as the reference method for detecting erosions in RA hands, the Artoscan showed higher sensitivity than the MagneVu and CR. All imaging modalities had high specificities. The better performance of the Artoscan should be considered when selecting an imaging method in RA.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Magnetic resonance imaging (MRI) has been shown to be more sensitive than conventional radiography (CR) in visualising erosive damage in rheumatoid arthritis (RA).1 Furthermore, MRI has the ability to visualise synovitis, which is the primary pathological process in RA joint involvement,24 and bone oedema, which is a strong predictor of future erosive bone changes.58 Consequently, MRI has a large potential as a powerful tool for monitoring disease activity and for prognosticating erosive disease in RA.

Several studies have demonstrated low-field dedicated extremity MRI (E-MRI) to be highly sensitive in detecting erosions when whole body MRI is used as the standard reference.911 A recent study from our group12 compared the two units that currently are the most likely MRI competitors to CR for widespread use for RA joint imaging: the Esaote Artoscan 0.2 T and the MagneVu MV1000 0.2 T (MagneVu). Both units detected more erosions than CR. Artoscan detected more erosions than MagneVu, due to a higher average image quality and a smaller proportion of bone being visualised by MagneVu. Since then, the MagneVu software has improved by adding 0.5 cm to the slab thickness.

Computed tomography (CT) is a tomographic radiographic imaging method that visualises calcified tissue with high resolution, and CT can be considered a standard reference for detecting destructions of calcified tissue such as bone erosions in RA.13 14 By using multidetector CT with multiplanar reconstruction, three-dimensional visualisation of joints is possible, whereas radiography is a projection technique offering only a two-dimensional visualisation of the three-dimensional anatomy. However, in comparison with MRI, CT inadequately visualises soft tissue changes.

To date, all studies comparing MRI and CT in detecting erosions in RA have been done on whole body MRI units. In the wrist, Perry and colleagues demonstrated an 87% agreement between CT and MRI for detecting bone erosions,13 while Møller Døhn et al, in bones without radiographic erosions, found an agreement of 77%.15 In a study of RA metacarpophalangeal (MCP) joints, applying CT as the gold standard, MRI had a sensitivity, specificity and accuracy of 68%, 96% and 89%, respectively.16 No comparative studies of E-MRI, CT and radiography in evaluating erosive bone changes in RA of the wrists and MCP joints have been reported.

The objective of the present study was to compare the ability of two different E-MRI units (Artoscan and MagneVu) and CR to identify bone erosions in RA MCP and wrist joints, using CT as the standard reference method.

METHODS

Twenty patients (17 women), all of whom fulfilled the American College of Rheumatology (ACR) 1987 criteria for RA were included in the study, as well as 5 healthy controls (3 women) of median age 35 years (range 25–59).

The clinically and radiographically most involved hand was chosen and, in order to ensure a range of pathologies, patients with and without erosions on CR were examined. Nine patients had at least one radiographically visible erosion on CR examination performed before the study in the area examined by MRI and CT. Eleven patients had no radiographically visible erosion in the same area on prior CR examination.

Eighty percent of the patients were positive for rheumatoid factor (>27 IU/l) and 36.8% had a serum C-reactive protein level >8 mg/l. The median age of the patients was 55 years (range 32–74) and the median disease duration was 9 years (range 0–15). Median clinical and laboratory values for measures of disease activity were as follows: 28 swollen joint count: 3 (range 0–17), 28 tender joint count: 13 (range 0–22), DAS28-CRP 5.2 (range 1.8–6.1). None of the controls had signs or symptoms of joint disease.

All patients had been treated with disease-modifying antirheumatic drugs (DMARDs) including corticosteroids. A total of 80% of patients had been or were currently being treated with methotrexate (MTX) in combination with other DMARDs (including corticosteroids), and 15% had been or were currently being treated with DMARDs other than MTX. One patient (5%) was being treated with biological therapy in combination with MTX. No other patients had been or were currently being treated with biological agents.

All subjects were recruited from the Department of Rheumatology, Copenhagen University Hospital at Hvidovre and underwent CT and two MRI examinations of the wrist joints and the second to fifth MCP joints of one hand on the same day. CR of the hand was performed within 2 weeks.

MRI

The two E-MRI units being tested were (a) 0.2 T Artoscan MRI unit (Esaote Biomedica, Genoa, Italy) and (b) 0.2 T portable MV1000 MagneVu MRI system (Carlsbad, USA).

Image acquisition using the Artoscan E-MRI unit

The patients were seated in an adjustable chair with the hand positioned in the centre of a dual-phased array wrist coil in a neutral position. A T1-weighted three-dimensional gradient echo sequence (3D T1-GE) with subsequent multiplanar reconstruction (slice thickness (thk) 1 mm, repetition time (TR) 30 ms, echo time (TE) 12 ms, number of acquisitions (NEX) 1, matrix 192×160×80, field of view (FOV) 140×140 mm, voxel size 0.55×0.55×1 mm, scan time 6 min) covering the second to fifth MCP joint and metacarpal bases, all carpal bones and the distal radius and ulna was performed (fig 1). In case of a very large hand, wrist and MCP joint images were obtained separately.

Figure 1

Field of view of MagneVu (solid lines) and the Artoscan (dashed lines).

Image acquisition using the MagneVu E-MRI unit

The patients were seated in an armchair with the hand in the coil, palmar side facing up. Pads of variable thicknesses were used to place the region of interest in the FOV. A 3D T1-GE sequence (16 slices, voxel size 1×1×0.94 cm, 1 mm coronal in-plane resolution, TR/TE 100/27 ms, NEX 2, FOV 50×75 mm, slab thickness 1.5 cm, acquisition time 14.48 min) was performed of the following regions: (a) second and third MCP joints; (b) fourth and fifth MCP joints; and (c) the wrist (fig 1). In the MCP joints it was aimed to visualise the most dorsal part of the bone in the first coronal image. Furthermore, it was aimed to visualise at least 1 cm of the shaft of the metacarpal bone and proximal phalanges, respectively. In the wrist, visualisation of the scaphoid, lunate, capitate, distal radius, trapezium and trapezoid (fig 1) was particularly prioritised.

CT scanning

Patients were placed in a prone position with the arm stretched and the palm facing down. A Siemens Somatom Sensation 10 unit (120 kV, 100 mAs, slice width 2.0, collimation 0.6 mm, Kernel U90u, feed/rotation 1.2 mm, rotation time 0.75 s) was used. Axial images with a voxel size of 0.6×0.6×0.6 mm were obtained and software for multiplanar reconstruction created coronal (and sagittal) reconstructions with a slice thickness of 0.6 mm.

Conventional radiography (CR)

CR of the wrist and MCP joints was performed in posterior-anterior and Nørgaard projections on an Agfa CR-system (50 kV, 4 mAs, film focus distance 1 m).

Image evaluation

Each set of images (CR, CT, Artoscan MRI and MagneVu MRI) was separately and blindly evaluated for the presence or absence of erosions in each bone (fig 2); the radiographs were evaluated by a musculoskeletal radiologist (AV), the CT images by another musculoskeletal radiologist (EAB) and the two sets of MR images by a rheumatologist (BE) experienced in reading RA MR images. All readers were blinded to clinical results and other imaging findings (ie, the reader evaluating Artoscan images was blind to the results from the MagneVu, CR and CT).

Figure 2

Erosion in the ulnar aspect of the scaphoid visualised by CT scanning and the Artoscan but not by the MagneVu and conventional radiography. (A) Coronal and (B) axial CT images (the arrow marks the erosion). (C) Coronal and (D) axial Artoscan images (the arrow marks the erosion). (E–H) Four consecutive coronal MagneVu images (the circle marks the scaphoid; no erosion is seen). (I) Conventional radiograph in posterior-anterior projection (the circle marks the scaphoid; no erosion is seen).

Bone erosions on CT were defined as a sharply demarcated area of focal bone loss seen in two planes, with a cortical break (loss of cortex) seen in at least one plane. Definitions of MRI erosions were as suggested by Outcome Measures in Rheumatology (OMERACT) Rheumatoid Arthritis MRI Scoring System (RAMRIS)17 (ie, a sharply marginated bone lesion with correct juxta-articular localisation and typical signal characteristics which is visible in two planes with a cortical break seen in at least one plane).

The unit specific parameters (including slice thickness) were chosen in order to test and compare the best performances of the two E-MRI units and the CT unit rather than testing the same parameters on different units. Axial and coronal CT and MRI images were always evaluated. Furthermore, sagittal images were allowed to use in order to confirm an erosion in the second plane if the specific erosion was seen only in either coronal or axial images. In addition, on the MagneVu images the percentage of each bone being visualised was estimated.

Statistical analysis

The sensitivity, specificity and accuracy of CR, MagneVu and Artoscan E-MRI were calculated separately for each assessed bone using interpretations from the CT images as the reference method. The proportion of bones that were 0%, 1–33%, 34–66%, 67–99% and 100% visualised by the MagneVu was calculated separately for each bone.

RESULTS

MCP joint coverage

The Artoscan unit allowed visualisation of the entire bones of the second to fifth MCP joints in two planes. The MagneVu allowed only visualisation of 1.5 cm of the ventral-dorsal diameter (slab thickness 1.5 cm) of the bone (fig 3), with the result that 84.2% of bones were visualised entirely and 15.8% of bones were 67–99% visualised (table 1). In one case the MagneVu examinations of the wrist and second to fifth MCP joints were not possible to evaluate and in another it was not possible to evaluate the examination of the fourth and fifth MCP joints.

Figure 3

Proportion of bones visualised by the MagneVu in the recent software version (slab-thickness 1.5 cm). The field of view is illustrated on a lateral radiograph.

Table 1 Visualisation of bones and bone erosions in the MCP joints (total number of examinations  =  25)

MCP joint sensitivity, specificity and accuracy

A total of 200 MCP joint areas were examined. CT, Artoscan, MagneVu and CR detected 31, 29, 22 and 18 bones with one or several erosions, respectively. The number of erosions seen in the different bones and the sensitivity, specificity and accuracy of Artoscan, MagneVu and CR when the CT unit was used as the reference method are shown in table 1.

The sensitivity of Artoscan (0.68) for detecting erosions was higher than MagneVu (0.54) and CR (0.57). This was mainly due to a higher sensitivity in metacarpal heads (0.74, 0.53 and 0.47, respectively). In phalangeal bases the sensitivity of CR for detecting erosions was 0.78, while the corresponding values for Artoscan and MagneVu were lower (both 0.56). The specificities for Artoscan, MagneVu and CR were 0.94, 0.93 and 0.99, respectively. The corresponding values for accuracy were 0.90, 0.88 and 0.93.

Wrist bone coverage

Artoscan visualised the entire wrist joint area. The FOV of MagneVu resulted in incomplete visualisation of the majority of the examined joint areas (table 2). Of the 350 bones examined, 31.6% were visualised entirely, 37.9% were 67–99% visualised and 19.7% were not visualised at all.

Table 2 Visualisation of bones and bone erosions in the wrist (total number of examinations  =  25; for MagneVu, only 24 examinations were evaluated)

Wrist sensitivity, specificity and accuracy

A total of 350 joint areas were examined. CT, Artoscan, MagneVu and CR detected 157, 87, 33 and 27 bones, respectively, with erosions. The distribution of the registered erosions and the sensitivity, specificity and agreement of Artoscan, MagneVu and CR when CT was used as the reference are shown in table 2.

The sensitivity of Artoscan for detecting wrist bone erosions was 0.50, showing the highest values in the radius, ulna, lunate, trapezoid and fourth metacarpal base bone. The sensitivities of both MagneVu and CR for detecting wrist bone erosions were low, with CR presenting slightly higher values (0.29) than MagneVu (0.23). MagneVu showed the highest sensitivities in the hamate, triquetrum and trapezoid. MagneVu found no erosions in the pisiform and trapezium, as these bones were not visualised in any patient. The specificities for Artoscan, MagneVu and CR were 0.92, 0.98 and 0.98, respectively. The corresponding values for accuracy were 0.75, 0.70 and 0.71.

DISCUSSION

This is the first study to compare and validate two different extremity dedicated MRI units (the MagneVu and the Esaote) and CR for assessment of RA bone erosions using CT as the reference method.

A higher sensitivity for the Artoscan (0.68) compared with the MagneVu (0.54) and CR (0.57) for detecting erosions in MCP joints was demonstrated. This was also found in the wrist bones, in which the sensitivity was markedly lower, but still higher for the Artoscan (0.50) than for the MagneVu (0.23) and CR (0.29). Specificities for the Artoscan, MagneVu and CR were high (⩾0.92) in both MCP and wrist joints.

While CR and the Artoscan allowed evaluation of the entire MCP and wrist joint area, the MagneVu only allowed assessment of limited parts of the bone, particularly in the wrist (tables 1 and 2).

Although CT is not a thoroughly validated method in RA, previous studies have found that it is very sensitive for RA bone erosions,13 15 16 and we found that comparison with CT would provide important validation of MRI and CR findings. The CT unit used in the present study provided high-resolution CT images and was considered suitable as a standard reference.

In this study, E-MRI findings were in good agreement with findings from the applied high-resolution three-dimensional tomographic x ray modality (ie, CT findings), even though MRI overall registered fewer erosions than CT. It can be considered preferable that MRI underestimates rather than overestimates the number of erosions in order to minimise false positive findings.

Overall, the MagneVu had a lower sensitivity for bone erosions than the Artoscan, probably related mainly to the lower coverage of the joints (smaller FOV). The lower sensitivity is consistent with the findings of previous studies comparing the Artoscan and MagneVu. In a previous study12 we found that the MagneVu had lower mean image quality (due to its image quality in general but also the frequent artefacts caused by the significantly longer acquisition times (6 min for MCP joints 2–5 on the Artoscan vs 30 min on the MagneVu)). The smaller FOV and inability to move the FOVs are other disadvantages of the MagneVu.

In the previous study we used a low-field MR unit (Artoscan) as the gold standard. Using a similar machine (ie, an MRI unit) may favour the MagneVu compared with CR. In the present study, in contrast, we used a radiographic technique (CT) as the standard reference. We feel that CT is the better reference method and would therefore be more confident in the sensitivities reported here than in the previous study.

In agreement with the present study, the previous study found fewer erosions detected by the MagneVu unit than with the Artoscan.12 The bone coverage of the MagneVu has been increased in the recent software version (used in the present study), since the slab thickness has increased from 1 cm to 1.5 cm (fig 3). The present study has shown, however, that this improvement was not sufficient to provide a comparable ability to the Artoscan to detect erosions. The incomplete bone coverage appears to be the main contributor to the lower sensitivity of the MagneVu compared with the Artoscan. Other technical reasons for the difference between the two E-MRI units could be the generally poorer image quality and the relatively long examination time for the MagneVu compared with the Artoscan (45 min vs 6 min) which inevitably increases the risk for movement artefacts and difficulties with positioning the wrist optimally.

The low sensitivity of CR for detecting erosions in the wrist bones is in accordance with earlier comparisons of CR and MRI.1 18 19 The particular difficulties of CR in the wrist joints compared with the MCP joints can be explained by projectional superimposition of bones in this complex joint. CR was, however, found to be more sensitive than MagneVu in detecting erosions in the wrist. This probably can be explained by the low coverage of the wrist by the MagneVu.

The relatively poor performance in the phalangeal bases of MRI compared with CR is probably explained by the fact that axial MR images often cannot visualise the very frequent small radial and ulnar erosions at the most proximal corner of the phalangeal base unless they are quite large. To increase the sensitivity of MRI in this area, it would probably be necessary to omit the requirement of erosions to be visible in two planes.

To our knowledge, no published studies have compared CT, E-MRI and radiography in RA joints. In a study conducted by Perry and co-workers,13 a comparison between CT and whole body MRI in wrist joints of nine patients with RA showed an overall agreement between CT and MRI of 87% in detecting bone erosions. More erosions were found with CT than with MRI.

In a recent study of 17 patients with RA, Møller Døhn et al20 reported sensitivities of detecting erosions in the wrist by MRI and CR (with CT as the reference) of 66% and 14%, respectively. In another study by the same authors,16 MRI had a sensitivity of 65% for detecting radiographically invisible MCP joint erosions. Our results are in agreement with these studies and support previous studies which suggested that E-MRI using the Artoscan is reliable for detecting RA bone erosions.

In agreement with previous studies1 3 19 21 which reported that CR has a poor sensitivity in detecting bone erosions compared with MRI, we found that CR had a lower sensitivity (34%) in detecting bone erosions in RA MCP and wrist joints than the Artoscan (53%). The fact that the MagneVu showed a lower sensitivity than CR in detecting erosions clearly demonstrates that E-MRI units do not perform identically and that thorough testing of new E-MRI units is essential.

CONCLUSION

When applying CT as the reference method for detecting erosions in RA hand, the Artoscan had higher sensitivity than the MagneVu and CR. All imaging modalities revealed high specificities. The field of view of the MagneVu is too small to visualise all bones in the wrist in one examination. With the present technique and software version, the MagneVu unit had lower sensitivity in detecting erosions than CR, and MRI using this unit therefore did not show a substantial improvement over CR. The better performance of the Artoscan should be considered when selecting an imaging method in RA.

Acknowledgments

The authors acknowledge MagneVu, Carlsbad, USA for donating the MagneVu MV1000 E-MRI unit for research purposes.

REFERENCES

Footnotes

  • Competing interests: None.

  • Funding: Financial support was received from The Danish Rheumatism Association, Elisabeth and Karl Ejnar Nis-Hanssens Memorial Award and Departments for Research, Quality and Education, Southern Region.

  • Ethics approval: Local ethics committee approval was obtained before starting the studies and all participants gave informed consent.