Multicontrast MR Imaging at 7T in Multiple Sclerosis: Highest Lesion Detection in Cortical Gray Matter with 3D-FLAIR

Discussion

Our results show that 3D-FLAIR detects the highest number of MS lesions at 7T, in total cortical GM and in total WM. On FLAIR images, signal from the CSF is nulled, increasing contrast between lesions and adjacent CSF. At a standard field strength (1.5T), this result meant an increase in MS lesion detection compared with conventional T2WI sequences, especially for juxtacortical lesions.^15,27,28 At 3T, FLAIR showed an increased lesion detection compared with 1.5T, and it showed superiority over conventional T2WI and T1WI sequences for WM lesion detection.^29,30 Regarding cortical GM lesions at lower field strengths, the highest lesion detection was found by DIR, which at 1.5T and at 3T, suppresses the signal of both CSF and WM, improving the visibility of the cortex and cortical abnormalities.^{7,15,16,31,32} This improved visibility of cortical abnormalities at DIR did not hold in our 7T dataset. The detection of total GM lesions was higher on 3D-FLAIR than on 3D-DIR, mostly due to a considerably superior detection of mixed lesions. The higher total GM and mixed-lesion detection at 3D-FLAIR was statistically significant when tested in a patient-wise analysis. Although the detection of purely intracortical lesions was higher at 3D-DIR compared with the other sequences, this did not reach statistical significance in the patient-wise comparison. The 10% gain in purely intracortical lesion detection on 3D-DIR compared with 3D-FLAIR concerned only 4 lesions.

In the present literature, there are several studies that focus on the detection of cortical GM lesions at 7T,^{17⇓–19,33,34} mostly reporting that T2*-weighted gradient-echo sequences improve detection. One of these studies by using T2*-weighted gradient-echo images even suggested that this sequence should be the new criterion standard for the detection of cortical lesions in patients with MS.³³ Another recent study reported that white matter signal attenuation at 7T is feasible and is able to detect cortical abnormalities at 7T.³⁵ To our knowledge, the value of the recommended sequences for clinical use at 7T has not yet been investigated nor has the use of DIR. Our results show that the highest total cortical lesion detection at 7T is gained with 3D-FLAIR. Future studies should compare the difference between these experimental T2*-weighted gradient-echo or white matter–attenuated sequences and clinical 3D-FLAIR sequences and elucidate which sequence has the highest sensitivity in terms of 7T cortical lesion detection. At even a higher field strength (9.4T), T2WI sequences were able to discriminate demyelinated and remyelinated areas in postmortem MS lesions.^36,37

The results of our 7T study are promising because in vivo imaging of cortical abnormalities in MS has high clinical relevance. Cortical lesions correlate with cognitive impairment^8,9 and could be used as an outcome measure in MS research or could be of help in the development and monitoring of treatment. Furthermore, the detection of GM abnormalities has prognostic relevance and will help to identify patients with clinically isolated syndrome who will eventually convert to clinically definite MS. It has been proposed that the sensitivity of the criteria for the diagnosis of MS will increase when GM lesions are included.¹³ These McDonald criteria are at present based solely on WM abnormalities, but GM abnormalities might be introduced in the criteria in the future. 7T MR imaging of cortical lesions can play a role in these different aspects and can be of additional value in the care of patients with MS. However, only a limited number of 7T scanners are available, worldwide around 40 7T systems, as mentioned in a recent review on 7T MR imaging.³⁸

As opposed to 1.5T and 3T, the highest total cortical GM lesion detection at 7T was found by the 3D-FLAIR sequence. To investigate this finding further, we reviewed a subset of patients and compared 3D-FLAIR and 3D-DIR sequences side by side (Fig 3). This comparison suggested several possible explanations involving the design of both sequences as well as our rating process. First, the design of sequences on our 7T system causes a multiple-layered appearance of the cortex, which differs between 3D-DIR and 3D-FLAIR²² and has recently been described as differing among anatomic regions in the brain.³⁹ The cortex is often relatively more hyperintense on 3D-DIR than on 3D-FLAIR because of the attenuation of WM at DIR (Fig 3). This might have caused diminished visibility of cortical lesions on 3D-DIR. Furthermore, due to the combination of 2 inversion pulses at 3D-DIR, besides CSF and WM, additional tissue components are also partly attenuated, including lesions and particularly the periphery of a lesion. This did not make small lesions completely invisible but often too inconspicuous to score, a phenomenon that has also been described at lower field strengths.¹⁵ Because T1 values are field-strength-dependent, this effect could be different at a high field and should be further investigated. In addition to these aspects, the resolution of 3D-FLAIR was better than that of 3D-DIR, 0.51 mm³ versus 0.64 mm³ voxel volumes respectively, which might affect the ability to detect small lesions.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Examples of scoring differences at 7T MR imaging: 3D-FLAIR versus 3D-DIR, found while reviewing a subset of images retrospectively. A, A juxtacortical lesion (closed arrows) that was scored on 3D-FLAIR but not on 3D-DIR because of differences in contrast between the lesion and the cortical gray matter. On 3D-DIR, the cortex looks relatively hyperintense compared with 3D-FLAIR because of the attenuation of white matter, which might have hindered lesion detection. The open arrows show a juxtacortical lesion that was detected at both sequences. B, The closed arrows show a mixed lesion that was scored on 3D-FLAIR but not on 3D-DIR because of conservative scoring/distraction by many small hyperintensities and its smaller size on 3D-DIR. C, A juxtacortical lesion that was scored on 3D-FLAIR but not on 3D-DIR images (closed arrows) because tissue was attenuated too much by using the DIR sequence, which decreases the size of the lesion.

Second, the scoring process might also have hampered lesion detection at 3D-DIR. There is an abundance of small signal abnormalities, mostly perivascular spaces, on 3D-DIR (Fig 1). This may have caused too much distraction while scoring lesions so that they were overlooked or not specified as being a lesion. Furthermore, we used a rather conservative scoring system: Focal hyperintensities were only specified as lesions if they were in line with the consensus guidelines for scoring of cortical lesions on DIR, which were developed by the MAGNIMS study group.²⁵ Hyperintensities or inhomogeneities in the cortex were interpreted as artifacts or as small vessels more easily on the 3D-DIR sequence than on the 3D-FLAIR sequence, which may have caused an under-representation of lesions at 3D-DIR.

Contrast ratios of our 7T 3D-FLAIR have been reported to be lower compared with 3D-DIR: GM/WM of 0.40 and 0.35 in healthy controls and patients respectively for FLAIR, compared with 0.93 and 0.87 for DIR. Lesion/WM contrast in FLAIR was 0.86, whereas it was 2.92 in DIR; lesion/GM was 0.91 in FLAIR and 1.40 in DIR.²² Hence contrast ratios are unlikely to have advanced lesion detection at 3D-FLAIR. Despite the higher relative contrast in DIR, the absolute contrast in terms of contrast-to-noise ratios could be less, due to the overall reduction in SNR by the extra inversion pulse.

The lower lesion counts detected by the 3D-DIR sequence cannot be explained by a reclassification phenomenon (ie, the shift from one anatomic region to another because of improved visibility of the cortex). This result was the case at 1.5T and 3T DIR, when a slightly lower sensitivity in juxtacortical lesion detection was counterbalanced by an increased detection of mixed or purely intracortical lesions.^16,31,32 However at 7T, with higher spatial resolution and different image contrast, visibility of the cortex was improved at all sequences; hence, a reclassification phenomenon was not evident in our data.

Concerning WM lesion detection at 7T, our results showed statistically significant differences in the juxtacortical region. At 3D-FLAIR, 3D-DIR, and 3D-T1WI, a higher number of juxtacortical lesions were detected compared with 2D-T2WI. Total WM lesion detection was highest at 3D-FLAIR when analyzed lesion-wise, though the difference with 3D-DIR and 3D-T1WI was negligible (1% higher). When we tested patient-wise, this increased WM lesion detection of 7T 3D-FLAIR, compared with other sequences, was not statistically significant.

Limitations of the study are that we could have taken more advantage of the multicontrast protocol by scoring lesions one on one on all 4 sequences simultaneously. The goal of this study, however, was to evaluate the performance of the sequences regarding overall lesion counts in various anatomic regions among the different sequences, so we chose to score the sequences separately. The scoring process was done by 3 raters in consensus because of the novelty of 7T MR images. A possible limitation could be that we have no intra- or interobserver comparison. In the future, it would be interesting to study this, possibly in addition to comparing interobserver agreement among different field strengths and among raters of different levels of experience. In addition, most important, the results in this study are specific for our MR imaging system and our sequences with mentioned parameters. Whether results can be generalized to other 7T MR systems remains a matter for future research.