Introduction	Choose Top of pageABSTRACTIntroduction <<Materials and MethodsResultsDiscussionReferencesCITING ARTICLES

The value of using intraarticular gadolinium-based contrast agents to enhance the depiction of the internal structure of joints has been well established, and the use of these agents is currently considered routine, particularly in the shoulder [1-6]. An accepted procedural approach for direct MR arthrography involves injecting a fixed concentration of a gadolinium-based contrast agent into the articular space using fluoroscopic guidance with iodinated contrast injection to confirm proper needle placement [7]. In an in vitro study designed to optimize a gadolinium-based contrast agent protocol for direct MR arthrography [8], the highest signal intensity in a series of saline and albumin dilutions imaged at 1.5 T was found to be in the range of 1.25-2.0 mmol/L. In several other recent studies, authors have reported using concentrations at or above this range (2.00-5.00 mmol/L) when imaging a variety of joints at 1.5 T [9-12].

Although MR arthrography at 1.5 T is considered to be the imaging technique with the highest diagnostic efficacy for assessing internal derangements of major joints, limitations have also been reported [5, 13]. MRI at 3 T, with its inherently higher signal-to-noise ratio (SNR), has the potential to improve image quality and to provide a higher spatial resolution [14-16], which would be particularly favorable in enhancing the diagnostic performance of MR arthrography. To our knowledge, however, no study has been performed to optimize direct MR arthrography procedures at 3 T, nor have suitable gadolinium-based contrast agent concentrations been examined in depth at 3 T. Optimal gadolinium-based contrast agent concentrations used at 3 T may differ from those used at 1.5 T because of predictions based on field strength and associated relaxation times.

Because 3-T scanners will be increasingly used clinically and also to perform direct arthrographic examinations of joints, the aims of our study were to determine the optimal concentration of gadolinium-based contrast agents at 3 T versus 1.5 T and to assess the additional effect of iodinated contrast agents on imaging at the higher field strength.

Results	Choose Top of pageABSTRACTIntroductionMaterials and MethodsResults <<DiscussionReferencesCITING ARTICLES

The T1 and T2 decay curves are shown in Figures 3 and 4. Mean and SDs of the percentage changes in T1 and T2 from 1.5 to 3 T were calculated for contrast concentrations from 0.078 to 5 mmol/L. There are no substantial differences in the T1 curves between 1.5 and 3 T for either saline or saline- and-albumin gadodiamide solutions (0.5% ±; 2.8% and -0.4% ±; 6.6%). For gadodiamide concentrations less than 1 mmol/L, the saline solutions showed a 10-15% increase in T1 over albumin solutions at the same concentration. The T2 values for these solutions were generally greater at 3 T than 1.5 T, with the difference increasing at lower concentrations of contrast material. For iodinated contrast material, we found a consistent increase in T1 (21% ±; 3.8%) and decrease in T2 (-20% ±; 4.8%) at 3 T compared with 1.5 T. Calculating NSIs versus gadodiamide concentrations from these data for T1-weighted sequences resulted in a small shift of the curves for saline and saline-plus-albumin to lower gadodiamide concentrations at 3 T compared with 1.5 T, and a significantly larger signal decrease due to iodinated contrast material at 3 T (Figs. 5A, and 5B). Calculated T2- and proton density-weighted sequences showed similar increased signal losses for iodinated contrast material at 3 T.

View larger version (14K)

Fig. 5A. —Graphs show calculated normalized signal intensity variation due to measured T1 and T2 versus gadolinium concentration for T1-weighted spin-echo sequence. Calculated graph for T1-weighted spin-echo sequence with TR/TE of 500/15. Note that at 3 T, curve is shifted slightly to lower concentrations for albumin solutions. Curves for normal saline solutions were similar and are not shown. Substantially lower signal intensities are found for solutions with gadodiamide and iohexol, with significantly greater decrease at 3 T than at 1.5 T.

View larger version (14K)

Fig. 5B. —Graphs show calculated normalized signal intensity variation due to measured T1 and T2 versus gadolinium concentration for T1-weighted spin-echo sequence. Calculated graph for more heavily T1-weighted spin-echo sequence with TR/TE of 500/8. Note that for gadodiamide concentrations below about 1 mmol/L, the addition of iodinated contrast material should increase signal-to-noise ratio (SNR) for this sequence, although peak SNR achievable with iodinated contrast material is still less than that achievable with gadodiamide alone.

View larger version (16K)

Fig. 6A. —Measured signal-to-noise ratio (SNR) versus concentration of gadodiamide in normal saline, and saline with iohexol (iodinated contrast material), for four sample sequences. SNR curves for other T1- and T2-weighted sequences were similar in form to those shown here. T1-weighted spin-echo sequence, TR/TE, 500/15 (1.5 and 3 T) (A); proton density-weighted spin-echo sequence, 2,550/30 (1.5 and 3 T) (B); T2-weighted spin-echo sequence, 4,000/45 (1.5 and 3 T) (C); and gradient-echo sequences, 33.3/13 (3 T) and 34/13 (1.5 T) (D). In A and B, observe some shift to lower concentrations for saline solutions at 3 T relative to 1.5 T. Note increase in SNR at 3 T for all sequences except for T2-weighted sequence with iodinated contrast material. Also note decrease in SNR with iodinated contrast material except for T1-weighted sequence at low gadolinium concentrations.

View larger version (14K)

Fig. 6B. —Measured signal-to-noise ratio (SNR) versus concentration of gadodiamide in normal saline, and saline with iohexol (iodinated contrast material), for four sample sequences. SNR curves for other T1- and T2-weighted sequences were similar in form to those shown here. T1-weighted spin-echo sequence, TR/TE, 500/15 (1.5 and 3 T) (A); proton density-weighted spin-echo sequence, 2,550/30 (1.5 and 3 T) (B); T2-weighted spin-echo sequence, 4,000/45 (1.5 and 3 T) (C); and gradient-echo sequences, 33.3/13 (3 T) and 34/13 (1.5 T) (D). In A and B, observe some shift to lower concentrations for saline solutions at 3 T relative to 1.5 T. Note increase in SNR at 3 T for all sequences except for T2-weighted sequence with iodinated contrast material. Also note decrease in SNR with iodinated contrast material except for T1-weighted sequence at low gadolinium concentrations.

The signal intensity versus gadodiamide concentration curves that resulted from directly measuring the signal intensities in the vials showed a similar shape at 3 and 1.5 T, as shown in Figures 6A, 6B, 6C, and 6D—for example, T1-weighted, proton density-weighted, T2-weighted, and gradient-echo sequences for normal saline and iodinated contrast material. The curves for saline plus albumin were similar to the normal saline curves and are omitted for clarity. For T1-weighted sequences, the shift of the signal intensity curve at 3 T to lower gadodiamide concentrations relative to the curve obtained at 1.5 T was more pronounced than that seen in the calculated NSI curves, as can be seen in Figure 7A, where the curves have been plotted normalized to the peak signal intensity at each field strength. For the T1-weighted images at a TR/TE of 700/14, the peak shifted from a concentration of 1.25 to 0.625 mmol/L (Figure 7A); whereas for the TR/TE of 500/15, the peak shifted from 2.50 mm/L to 1.25 mmol/L (Figure 6A). However, the peaks of the intensity versus concentration curves were relatively broad. For example, for the T1-weighted 700/14 sequence, the intensity for 3 T at 1.25 mmol/L was similarly high to that at the peak concentration of 0.625 mmol/L.

View larger version (13K)

Fig. 6C. —Measured signal-to-noise ratio (SNR) versus concentration of gadodiamide in normal saline, and saline with iohexol (iodinated contrast material), for four sample sequences. SNR curves for other T1- and T2-weighted sequences were similar in form to those shown here. T2-weighted spin-echo sequences, TR/TE, 4,000/45 (1.5 and 3 T).

View larger version (15K)

Fig. 6D. —Measured signal-to-noise ratio (SNR) versus concentration of gadodiamide in normal saline, and saline with iohexol (iodinated contrast material), for four sample sequences. SNR curves for other T1- and T2-weighted sequences were similar in form to those shown here. Gradient-echo sequences, 34/13 (1.5 T) and 33.3/13 (3 T).

View larger version (13K)

Fig. 7A. —Measured signal-to-noise ratio (SNR) versus concentration of gadodiamide in normal saline, with curves scaled to equal peak heights for 1.5 and 3 T. T1-weighted spin-echo sequence, TR/TE of 700/14 (A) and T2-weighted spin-echo sequence, 4,000/45 (B). Note shift of peak to lower concentrations at 3 T for T1-weighted sequence.

View larger version (12K)

Fig. 7B. —Measured signal-to-noise ratio (SNR) versus concentration of gadodiamide in normal saline, with curves scaled to equal peak heights for 1.5 and 3 T. T1-weighted spin-echo sequence, TR/TE of 700/14 (A) and T2-weighted spin-echo sequence, 4,000/45 (B). Note shift of peak to lower concentrations at 3 T for T1-weighted sequence.

For the T1-weighted sequences, the signal intensity versus gadodiamide concentration curves for the iodinated contrast solutions were similar at the two field strengths, each having a relatively flat peak region, between 0.156 and 0.625 mmol/L (Figure 6A). Similar findings were obtained in the calculated NSI versus concentration curves shown in Figures 5A, and 5B. Note that the calculated NSI takes into account only changes due to T1 and T2, whereas the experimental SNR curves of Figures 6A, 6B, 6C, and 6D are also reflective of other factors such as coil sensitivity and overall increase in signal strength at 3 T.

Similarly to the T1-weighted spin-echo sequences, the saline SNR curve for the proton density-weighted spin-echo sequence at TR/TE of 2,550/30 shifted to the left at 3 T, with the peak shifting from a concentration of 0.625 to 0.313 mmol/L (Figure 6B). Corresponding iodinated contrast signal intensities at 3 T appear to approach a maximum value as gadolinium concentrations decreased to the lowest concentration measured (0.078 mmol/L), whereas at 1.5 T the SNR reaches a flat peak at concentrations of 0.156 mmol/L and below.

Peak SNR for saline for the more heavily T2-weighted fast spin-echo TR/TE of 4,000/45 (Figure 6C), fast spin-echo TR of 3,000/45, and fast spin-echo TR of 2,000/45 sequences at both 3 and 1.5 T all occurred at a gadodiamide concentration of 0.313 mmol/L. However, at the optimal concentrations obtained from T1-weighted images—0.6-2.5 mmol/L—no major decrease in SNR was noted on these moderately T2-weighted scans because the peaks were quite broad. Iodinated contrast SNRs were similar at the two field strengths, increasing as gadodiamide concentration decreased. Any peak of SNR for these solutions appears to be at or below the 0.078 mmol/L gadodiamide concentration.

SNR for the gradient-echo sequences (33.3/13 at 3 T and 34/13 at 1.5 T, Figure 6D) peaked at an identical gadolinium concentration of 2.50 mmol/L for the saline and saline-with-albumin solutions. The iodinated contrast SNR curve also showed no shift between 1.5 and 3 T for these sequences, with the peak occurring at a concentration of 1.25 mmol/L.

Among the normal saline dilution series and the saline-with-albumin dilution series, no substantial differences in signal intensities were observed at either field strength for any sequence investigated. Adding iodine to gadodiamide solutions caused a decline in SNR for almost all gadodiamide concentrations for all investigated pulse sequences at both 1.5 and 3 T because of the large susceptibility-induced decrease in T2. Because this effect is offset to some extent by the iodine-induced increase in T1, T1-weighted sequences show the least SNR decrease. For low concentrations of gadodiamide (< 0.5 mmol/L) and heavy T1 weighting, an increase in SNR with the addition of iodine may even be observed, as shown in Figures 5A, 5B, and 6A. Figure 5B shows the predicted normalized signal variations using our measured T1 and T2 values for a more heavily T1-weighted sequence (500/8) than was performed in this study, where for any gadodiamide concentration below approximately 1 mmol/L we would expect an SNR increase with the addition of iodine.

On all sequences, the relative decline in signal intensity caused by adding iodine was more pronounced at 3 T than at 1.5 T, and the SNR at the optimum gadodiamide concentration was always reduced with iodination. These decreases at optimum gadolinium concentrations for saline and iodine gadolinium solutions on T1-weighted images (500/15) were 36% at 3 T and 24% at 1.5 T. For proton density-weighted TR/TE of 2,550/30, decreases were 52% at 3 T and 40% at 1.5 T; and for T2-weighted TR/TE of 4,000/45, they were 61% and 49%. From the predicted intensity curves for the most heavily T1-weighted image investigated (500/8), we find decreases of 27% at 3 T and 14% at 1.5 T.

Using identical sequence parameters and comparing peak SNR in the saline solutions, an increase from 1.5 to 3 T was found by factors of 1.50 for TR/TE of 500/15, of 1.51 for 700/14, of 1.62 for 2,550/30, of 1.38 for 4,000/45, of 1.35 for 3,000/45, of 1.31 for 2,000/45, and of 1.18 for the gradient echo sequences, respectively.

Discussion	Choose Top of pageABSTRACTIntroductionMaterials and MethodsResultsDiscussion <<ReferencesCITING ARTICLES

In this phantom study comparing signal intensities of gadolinium-based contrast medium dilutions at 3 and 1.5 T, we found that the signal intensity peaks tended to shift to lower gadolinium concentrations for the T1- and proton density-weighted spin-echo sequences at 3 T. These changes, however, were not pronounced and were even less well appreciated when signal intensity/gadodiamide concentration curves were calculated from T1 and T2 data. Considering these results, a substantial change in previously [5, 6, 17] proposed optimal concentration of gadolinium-based contrast agents at 1.5 T (2 mmol Gd/L) may not be required at 3 T for MR arthrography.

A mild decrease in concentration, down to 1-1.25 mmol Gd/L, may be considered to optimize SNR in T1-weighted sequences. In addition, the administration of additional iodinated contrast agents should be minimized, because they decrease signal intensity at 3 T more substantially than at 1.5 T. These effects are greatest for T2- and proton density-weighted images and for higher concentrations of gadolinium. It is possible for the addition of iodine to give a net SNR increase if combined with low concentrations of gadolinium and heavily T1-weighted imaging. Compared with the T1-weighted imaging sequences, proton density- and T2-weighted sequences had a significantly lower optimum SNR contrast concentration (< 0.5 mmol/L), with little or no shift in the intensity versus concentration curves between 1.5 and 3 T. For iodinated solutions, no peak SNR was observed for proton density- and T2-weighted sequences, with the signal increasing monotonically to the lowest gadolinium concentration measured.

Previously there was some discussion whether to use diluted gadolinium-based contrast agents or just to apply normal saline for MR arthrography, and studies examined whether a difference existed in image quality with saline or gadolinium-based contrast agents [18, 19]. Although Zanetti and Hodler [19] found better results for gadolinium-based contrast agents, Yao et al. [18] found a similar performance of saline and gadolinium-based contrast agents. Currently, however, most centers apply diluted gadolinium solutions, and a recent meta-analysis of 112 published studies found it to be a safe and efficient technique for diagnosing internal derangement of joints [5].

Although our study found the best SNR for a 2 mmol/L solution of gadolinium-based contrast agents, other investigators have suggested the use of higher gadolinium concentrations [20-24]. Kopka et al. [23] examined 2.5, 10, and 45 mmol/L solutions of gadopentetate dimeglumine and found the highest contrast-to-noise ratio for the 10 mmol/L solution. Those authors added 50% iotrolan to the 2.5 mmol/L solution, although they added less than 5% of iotrolan to the 10 mmol/L solution, which may explain these discordant findings compared to those of our study. Jacobson et al. [22] and Loew et al. [24] described the use of a 1:200 dilution, which corresponds to a concentration of 2.5 mmol/L. Pfirrman et al. [20] and Czerny et al. [21] used even higher concentrations of 4 mmol/L. If a large joint effusion is present, such as in the postoperative knee, or if a long delay (> 1 hr) between the injection and the MR examination is expected, the application of a higher gadolinium concentration may be justified to compensate for a subsequent dilution of the applied contrast agent in joint effusion or excretion from the joint with time.

Similar to the findings of Montgomery et al. [8], we observed that iodinated contrast material, although helpful in directing needle placement, may compromise the increase in signal intensity derived from gadolinium. This decline in signal intensity seen in our test solutions on all pulse sequences by adding iodine is most likely due to the well-known magnetic susceptibility of iodine (mass magnetic susceptibility, -4.40 × 10^-9 m³kg^-1. This susceptibility effect was more pronounced at 3 T than at 1.5 T and was also stronger on the proton density- and T2-weighted imaging sequences than on the T1-weighted sequences. To minimize the confounding susceptibility effect of additive iodine, the amount of iodinated contrast material injected into the joint should be minimized [23]. Nevertheless, several investigators described the use of high concentrations of iodinated contrast agents, which were mixed with the gadolinium solution and injected into the joint. Jacobson et al. [22] recommended the use of a solution of 10 mL saline, 10 mL iohexol, and 0.1 mL gadopentetate dimeglumine. Brown et al. [25] showed that these mixtures are safe and that no significant dissociation of the gadolinium ion is found. Similar mixtures of saline and iodinated contrast agents with gadolinium-based agents were also used in other studies [26, 27]. The advantages of using a single iodine-and-gadolinium solution are constant contrast agent concentrations and easier handling (without the required change of syringe and potential air bubble deposition), but the caveat is that a lower signal intensity of the applied contrast agent solution visualized on the MR images. Another potential advantage of a mixture of gadolinium and iodinated contrast material would be the production of an arthrogram in case the MR study is aborted or the patient becomes claustrophobic during the MR examination. The differences in signal intensity when using different concentrations of iodinated contrast material in gadolinium solutions should therefore be further investigated, and a concentration may have to be found that will provide a high signal at 3 T but that will also allow good depiction of the joint cavity at fluoroscopy.

Although gadodiamide concentrations that maximized signal intensity were determined in this in vitro study, an in vivo correlation is warranted. Factors such as further dilution of the injected concentration because of synovial fluid or large joint effusions, motion effects, and timing between injection and imaging, need to be evaluated at 3 T. We tried to simulate the effect of proteins found in synovial fluid by diluting the gadodiamide in an albumin solution as previously described [8] but found no substantial effect on signal intensity peaks at 1.5 and 3 T with the different sequences. Similar results were reported in the previous study performed at 1.5 and 0.2 T [8]. On the other hand, to eliminate the confounding effect of a large joint effusion, the radiologist could remove the native effusion before injecting the gadolinium mixture into the joint.

An interesting finding in that previous study performed at 1.5 and 0.2 T [8] is that the same contrast dilutions for MR arthrography (i.e., 1.25-2.0 mmol/L) were recommended for imaging at 1.5 and 0.2 T. Other investigators also favored the use of a single concentration of gadolinium-based contrast agents for imaging at 1.5 and 0.2 T [24, 28].

We determined SNR data as a quantitative measure for the observed signal effect of our test solutions on the MR images. These SNR data are the result of intrinsic relaxivities of the test solutions, relaxivity effects by the added contrast agents, and background noise. SNR data should not be confused with changes in R1 relaxivities due to the contrast agents. The generally higher SNR at 3 T compared with 1.5 T is due to the balance of a number of factors. These include the inherent increase in MR intensity with increased field strength, differences in receiver coil sensitivity between the two systems, and differences in the T1 and T2 relaxation times with or without added contrast agent. Not only our injected contrast agent solution, but also the surrounding joint tissues, will appear brighter at 3 T than at 1.5 T. Therefore, the resultant contrast between injected solution and surrounding tissues at different field strengths needs to be further investigated in in vivo studies.

Our results are limited to the use of gadodiamide, which is a nonionic gadolinium-based contrast agent, and must be proven for other small-molecular gadolinium chelates. We also did not study the effect of different dilutions of the iodinated contrast agent or different iodinated contrast agents. However, our findings at 1.5 T are consistent with these previously published in a contrast optimization study [8] and with clinical findings concerning contrast dilutions in MR arthrography of the shoulder at 1.5 T [5, 6, 17]. We therefore think that they should also be transferable to clinical examinations at 3 T.

Another potential limitation of this study is that we did not include any fat-containing media immediately adjacent to the contrast-media-filled vessels and thus did not investigate any fat-saturated sequences, which are routinely used in MR arthrography. We would not expect good fat-saturation pulses to significantly affect the relative SNR of different concentrations of contrast agents because these are set primarily by the T1 and T2 changes induced by the contrast media. However, associated effects, such as magnetization transfer or partial saturation of the water signal due to imperfect fat-saturation pulses, could affect the results in a true clinical setting. These effects may also affect the 1.5- to 3-T SNR comparisons for these sequences as a result of the differences in fat-saturation capabilities of the different MR systems. Because of the complex interactions involved in saturation pulses, only an in vivo study will adequately address this issue.

In conclusion, the results of this in vitro study suggest that similar concentrations of gadolinium-based contrast agents for MR arthrography may be used at 3 and 1.5 T, although a mildly increased dilution at 3 T should be considered to further optimize contrast. With a focus on T1-weighted sequences, concentrations of 2-1.25 mmol/L may be recommended. In addition, because SNR peak levels for iodinated contrast dilutions are relatively lower at 3 T than at 1.5 T, administration of iodinated contrast agents should be minimized during MR arthrography at 3 T.