Infratentorial lesions in multiple sclerosis (MS) are prevalent, with nearly a quarter of patients initially presenting with clinical symptoms related to brainstem or cerebellar lesions.1 The pons is the most frequent site of infratentorial involvement accounting for nearly half of lesions, followed by the midbrain, medulla, and cerebellar hemispheres.2 A wide range of symptoms may occur related to cerebellar dysfunction (eg, tremor, nystagmus, ataxia, hypotonia, cognitive impairment, etc) or brainstem dysfunction (eg, gait abnormality, sensorimotor symptoms, diplopia, vertigo, oscillopsia, trigeminal neuralgia, etc).3 Notably, patients with brainstem involvement in clinically isolated syndromes also have a significantly increased risk of progression to MS, with rates up to 60%.4

Although over 80% of MS patients display clinical symptoms referable to cerebellar or brainstem lesions, the radiologic correlation remains limited. For instance, only 63% of patients with brainstem or cerebellar symptoms had discernible lesions within the brainstem on magnetic resonance imaging (MRI).2 This discrepancy between clinical symptoms and MRI results has been dubbed the “clinicoradiological paradox.”3 Over time, efforts have focused on refining MRI techniques to enhance lesion detection in the posterior fossa. Variations of fluid-attenuated inversion recovery (FLAIR), double-inversion recovery (DIR), and T2-weighted spin echo imaging have been the mainstays despite their known limitations. Consequently, current MRI techniques still underestimate the extent of brainstem damage in MS that is detected by other clinical tests.5,6 Such discrepancy highlights the need for further advancements in the detection of infratentorial lesions in MS with MRI.

Ultra-high-field MRI has gained increasing interest in MS, particularly since the recent regulatory approval of 7 T MRI systems for clinical use. Because of the higher signal-to-noise ratio (SNR), 7 T MRI allows for higher spatial resolutions and provides better image contrast; therefore, it has shown better detection rates of cortical and juxtacortical lesions, improved sensitivity for the phase-rim sign, and superior detection of the central vein sign in MS imaging.7–10 Despite these advantages, 7 T presents greater challenges with imaging the posterior fossa versus its 1.5/3 T counterparts due to B1+ inhomogeneity, limited z axis coverage, larger B0 inhomogeneities, and greater sensitivity to dynamic field fluctuations.11–17 These cumulative effects lead to poor image contrast, low SNR, and image artifacts that frequently plague the brainstem, cerebellum, and upper cervical spinal cord. These effects are particularly pronounced for fast/turbo spin echo (TSE) type sequences, such as 2D and 3D sequences commonly used for T2-weighted, DIR, and FLAIR imaging. In addition, DIR and FLAIR are more challenging at 7 T due to a combination of high B1+ inhomogeneity, elevated specific absorption rate at ultra-high-field strength, and prolonged tissue longitudinal relaxation times of white matter (WM) and gray matter (GM), which introduces T1 effect and reduces the desirable T2 contrast.

In this study, we explore a novel approach to detecting demyelinating lesions in the posterior fossa and upper cervical spinal cord using a 7 T lesion-attenuated magnetization-prepared gradient echo acquisition (LAMA). This approach leverages the relative insensitivity to B1+ inhomogeneities of the 3D magnetization-prepared gradient echo (MPRAGE) sequence, high lesion-to-background contrast of signal-nulled demyelinating lesions, and the ability to achieve high spatial resolution in reasonable scan times at 7 T. First, we describe the optimization of the LAMA sequence as part of a magnetization-prepared 2 rapid acquisition gradient echo (MP2RAGE) sequence to yield simultaneous LAMA contrast and a conventional T1-weighted, MPRAGE-like image. Second, we evaluate the performance of the LAMA versus clinical reference standards of DIR and T2-weighted sequences.

MATERIALS AND METHODS Sequence Design

The lesion-attenuated contrast in LAMA is achieved by minimizing the signal of demyelinating WM lesions in a 3D MPRAGE sequence (Fig. 1). Although it can be implemented in a conventional single readout 3D GRE sequence, we opt for incorporating it in an MP2RAGE sequence with 2 gradient echo readouts at different inversion times (TIs).18 Compared with the traditional MP2RAGE sequence where the sequence parameters of the 2 acquisition windows are typically optimized for the GM-WM contrast in the calculated T1-weighted uniform images, incorporating LAMA into an MP2RAGE requires redesigning the sequence parameters of both readouts (ie, TI, FA, repetition time [TR], etc) to simultaneously optimize for both the LAMA contrast generated in the first readout window and the GM-WM contrast in the calculated T1-weighted images. By using an MP2RAGE, two contrasts, including both LAMA and the conventional T1-weighted images, can be generated from a single acquisition, which increases the overall data acquisition efficiency while still allowing a longer TR to be used to improve image SNR and resistance to B1+ inhomogeneity at 7 T.19

FIGURE 1:

Evolution of longitudinal magnetization (Mz) for gray and white matters in a single repetition time (TR) period, as well as that of lesions with T1 values from 1.6 to 2.4 seconds. The first and second readout windows (corresponding to the inversion 1 and 2 images) are highlighted in blue boxes, with their respective inversion times (TIs) indicated by vertical lines at the center. The proposed lesion-attenuated magnetization-prepared gradient echo acquisition (LAMA) contrast is obtained in the first readout window, where the TI is chosen around the nulling point of the average T1 value of demyelinating MS lesions to improve lesion contrast.

In this MP2RAGE-based implementation, the images acquired in the first readout after the inversion pulse (ie, INV 1 images) are designed to provide LAMA contrast. Figure 1 shows the evolution of the longitudinal magnetization (Mz) for GM and WM after an inversion pulse, as well as that of lesions with presumed T1 values from 1.6 to 2.4 seconds, respectively. To attenuate lesion signal and improve lesion contrast against surrounding WM, LAMA uses a TI at the nulling point of an average demyelinating MS lesion. The T1 value of an average WM lesion is assumed to be 1.9 seconds at 7 T, calculated as the average of brainstem and cerebellar lesion T1 values measured from T1 maps acquired on MS patients in our institution. Because the demyelinating MS lesion has a relatively broad T1 distribution, not all lesions will be perfectly nulled. However, by choosing the nulling point based on the average T1 value, the signals from lesions with a range of T1 can be attenuated (Fig. 1), which facilitates lesion detection.

Implementing LAMA at 7 T is challenging due to the severe B1+ inhomogeneity at ultra-high-field, therefore requiring intensive sequence optimization. The complex interaction of the transmit radiofrequency (RF) fields inside the head can cause deviation of the B1+ field in the infratentorial area, leading to nonideal image contrast and compromised uniformity in the area.20 To address this challenge, numeric simulations based on Bloch equations were performed to aid in the optimization of pulse sequence parameters.19,21,22 The effects of different sequence parameter combinations (TR, TI, flip angle, echo spacing, and bandwidth) on the signal intensities of GM, WM, and lesions were estimated. The optimal sequence parameters were then determined as the ones that allow uniform lesion signal attenuation in the first inversion image to achieve LAMA contrast within a range of B1+ inhomogeneity while maximizing the WM-GM contrast-to-noise ratio (CNR) in the resultant T1-weighted uniform image.

Clinical Study Design

This retrospective study was approved by Mayo Clinic institutional review board. We identified a consecutive series of patients evaluated from 2022 to 2023 with a 7 T brain MRI with MP2RAGE sequence having the LAMA sequence as the first TI, in addition to 3D SPACE DIR and 2D T2-weighted TSE sequences. Pulse sequence parameters used in the study are listed in Table 1. Images were acquired on a 7 T Magnetom Terra (Siemens Healthineers AG, Erlangen, Germany) with an 8-channel transmit, 32-channel receive head coil (Nova Medical, Wilmington, MA). All scans were performed in the circular polarized mode (“TrueForm” on Siemens platform). Although the coil is capable of performing parallel transmit, the use of parallel transmit carries elevated patient safety concerns due to higher SAR deposition.23–25 We therefore used the single transmit mode to reduce the related risks. Standard protocol also includes dielectric pads (built in-house)26 applied to the bilateral temporal and suboccipital regions.

TABLE 1 - Pulse Sequence Parameters TR, s TE, ms TI1/TI2, s FA, degrees FOV, cm2 Matrix Resolution, mm3 TA, min:s Sag 3D LAMA 4.5 2.2 0.95/2.5 6/4 230 × 230 288 × 288 0.8 × 0.8 × 0.8 8:44 Sag 3D SPACE DIR 8 408 0.62/3.4 T2 VFA 230 × 230 256 × 256 0.9 × 0.9 × 0.9 7:22 Ax 2D T2 TSE 5.75 56 NA 146 200 × 171 784 × 504 0.26 × 0.3 × 2 3:23

DIR, double inversion recovery; LAMA, lesion-attenuated magnetization-prepared gradient echo acquisition; FA, flip angle; TA, acquisition time; TE, echo time; TI, inversion time; TR, repetition time; TSE, turbo spin echo; VFA, variable flip angle.

Subjects were included if they were given a final diagnosis of MS after evaluation by 1 of 2 board-certified neurologists who specialize in MS and completed fellowship training in autoimmune neurology. The final diagnosis was in line with the 2017 revisions to the McDonald criteria.27 Images were manually assessed for quality, and patients were excluded if any of the 3 sequences had excessive motion artifacts.

Image Assessment

The imaging studies were independently reviewed by 3 board-certified neuroradiologists (A.A., J.V.M., E.H.M.). For the initial round of scoring, readers were provided only anonymized DIR and T2 sequences using a clinical Picture Archiving and Communication System (Visage Imaging, Richmond, Australia). In a subsequent session at a later date, the cases were anonymized again with randomized ordering compared with the previous session and only included the LAMA sequence for repeat scoring. The number of WM lesions was tallied within each of the following 4 regions: brainstem/middle cerebellar peduncle (central WM), brainstem/middle cerebellar peduncle (surface), cerebellum, and visualized cervical cord. Surface lesions included any lesion that had any part reaching the surface of the brainstem or peduncle (eg, ependymal surface of fourth ventricle or cisternal surface of brainstem), whereas central WM lesions did not have part of the lesion reaching the surface of the brain.

To quantitatively evaluate lesion conspicuity, a CNR was measured for 20 posterior fossa lesions in 5 randomly selected cases using the formula:

CNR=MLesion−MWMσLesion2+σWM2

where MLesion = mean intensity of the demyelinating lesion, MWM = mean intensity of normal WM, sLesion = SD of the demyelinating lesion, and sWM = SD of normal WM.

Statistical Analysis

The lesion count in each region was averaged among the 3 reviewers per case for both LAMA and DIR + T2. The distribution of this average lesion count was found to be not normally distributed in all regions, according to the method of D'Agostino. Thus, the nonparametric Wilcoxon signed rank test for paired samples was applied in its 1-sided form to test the hypothesis that LAMA increases lesion detection relative to DIR + T2. We also compared the total average lesion counts between the 2 MRI sequences, summed across all 4 brain regions. A Bonferroni correction was applied to control the false discovery rate. The individual raters' tallies were used to quantify reliability in lesion detection for both MRI sequences by estimating the intraclass correlation coefficients (ICCs) and their 95% confidence intervals (CIs) based on a single-rating, absolute-agreement, 2-way random-effects model. Lesion CNR was first tested among T2, DIR, and LAMA sequences using the Friedman test. Having rejected that null hypothesis, the CNR values for T2 and DIR were each compared against LAMA individually with the 1-sided Wilcoxon signed rank test for paired samples.

RESULTS

The optimized LAMA sequence parameters determined from numerical simulations are shown in Table 1, with the full protocol file shown in Table, Supplemental Digital Content 1, https://links.lww.com/RLI/A890. These parameters were implemented in the clinical study. An isotropic 0.8-mm resolution acquisition with full brain coverage can be obtained with a clinically acceptable scan time (~9 minutes), compared with 0.9-mm resolution in the DIR acquisition acquired in similar time. A higher resolution is possible with LAMA due to its higher intrinsic SNR compared with DIR.

Clinical Imaging Analysis

A total of 42 patients met inclusion criteria. The mean age was 44.9 ± 11.2 years, and there were 34 women (81%). There was good agreement between readers, with LAMA having a higher agreement than DIR + T2. For DIR + T2, the ICC was 0.61 (95% CI, 0.35–0.78). LAMA had an ICC = 0.75 (95% CI, 0.41–0.88).

LAMA detected significantly more lesions than DIR + T2 (mean 6.4 ± 5.8 vs 3.0 ± 2.7, respectively; P < 0.001) (Figs. 2A, 2B, 3, 4). Of the 6 patients with a consensus of no lesions on DIR + T2, 5 (83.3%) had at least 1 lesion on LAMA. In patients with identified lesions, 38 of 41 (92.7%) had more lesions detected on LAMA (Figs. 2A, 2B). Contrast-to-noise ratio was also significantly greater for LAMA (3.7 ± 0.9) than DIR (1.94 ± 0.7) and T2 (1.2 ± 0.7) (Fig. 2C; all P's < 0.001).

FIGURE 2:

A, Axial 7 T LAMA showing 2 distinct lesions (arrows) along the anterior and lateral surface of the right medulla. B, Axial 7 T 3D SPACE double-inversion recovery (DIR) shows faint hyperintensity (arrows) in the same regions. C, Axial 7 T T2-weighted turbo spin echo image is difficult to distinguish lesions (arrows) from artifacts within the anterior medulla (arrowhead).

FIGURE 3:

A, Sagittal 7 T LAMA showing numerous lesions (arrows) in the upper cervical spinal cord. B, Sagittal 7 T 3D SPACE DIR shows the limitations of TSE sequences with poor B1+ transmit limiting signal in the upper cord with none of the lesions clearly seen. C, Sagittal 7 T LAMA shows multiple lesions (arrows) in the upper cervical spinal cord. D, Sagittal 7 T DIR has limited conspicuity of the lesions.

FIGURE 4:

A, Mean total lesions in the posterior fossa and upper cervical spinal cord for DIR + T2 versus LAMA. B, Paired data plot showing, for each case, the change in mean total lesions detected in the posterior fossa and upper cervical spinal cord with DIR + T2 versus LAMA. C, Contrast-to-noise ratio between lesions and normal white matter for each of the sequences. *Box represents the lower quartile, median, and upper quartile, and whiskers include all data points within 150% of the interquartile range.

There were a significantly greater number of lesions detected on LAMA across all 4 regions (all P's < 0.001) (Fig. 5). The cervical cord had the greatest difference between techniques with an average of 3.5 times more lesions on LAMA versus DIR (Fig. 5). Surface lesions in the brainstem and peduncles were identified 2.1 times greater on LAMA and central lesions 2.0 times greater with LAMA. Cerebellar lesions were 1.8 times greater on LAMA.

FIGURE 5:

Mean total lesions across all 4 measured regions for DIR + T2 versus LAMA. Box shows the lower quartile, median, and upper quartile, and whiskers include all data points within 150% of the interquartile range.

DISCUSSION

In this study, we evaluate the performance of a novel approach for detection of infratentorial demyelinating lesions in MS on 7 T MRI. Clinical images demonstrate that the optimized LAMA sequence significantly improves lesion detection versus current standards due to its greater lesion contrast, robustness against B1+ transmit inhomogeneity at 7 T, and high spatial resolution, all within a reasonable scan duration. Given the importance of infratentorial lesion detection and historically suboptimal performance of MRI, LAMA has the potential to improve diagnosis and prognosis in patients with MS.

Infratentorial lesions are prevalent in MS, with 81.6% of patients experiencing brainstem or cerebellar symptoms and 22.5% of initial demyelination events occurring in the posterior fossa.1 Infratentorial lesions have a high specificity for MS and are a diagnostic marker of dissemination in space, as outlined in the 2017 revisions of the McDonald criteria.27 Patients presenting initially with demyelinating events in the brainstem or cerebellum are also at a greater risk of conversion to MS and more likely to develop disability.28,29 Notably, brainstem involvement also predicts the risk of conversion from relapsing-remitting MS to secondary progressive MS.30 Despite their importance, patients with brainstem or cerebellar symptoms had identifiable lesions in only 63% of cases.2 This well-described “clinicoradiological paradox” highlights the deficiencies in current methods for detecting infratentorial lesions.

Based on the reference standard imaging of DIR and T2, our results show a similar prevalence of posterior fossa and upper cervical cord lesions, with only 66.7% of patients having a lesion identified by all readers. Meanwhile, 92.9% of patients had at least 1 lesion identified on the LAMA sequence alone. Of the 6 patients with a consensus of no lesions, 5 (83.3%) had at least 1 lesion on LAMA. Our results highlight the added benefit of LAMA in the evaluation of MS and may aid in reducing this historic “clinicoradiological paradox.”

Compared with conventional techniques, namely, 3D DIR and 2D T2-weighted TSE, LAMA has several advantages. First, due to the higher intrinsic SNR with a 3D magnetization-prepared technique such as LAMA, a smaller isotropic voxel size can be achieved in permissible scan duration. High isotropic spatial resolution is more challenging with T2-weighted techniques such as DIR, as image SNR usually suffers from the use of longer TEs to emphasize T2 contrast, and the use of inversion pulses to improve tissue contrast also reduces signal. At 7 T, the T2-weighted method is more challenging due to the shortened T2 value.20,31 In addition, the 3D TSE readout used by DIR usually suffers from T2-induced blurring effect due to a long echo train, which can further reduce image sharpness. Second, due to the T1 prolongation effect at 7 T, LAMA can provide higher sensitivity for detecting MS lesions, as a greater T1 difference between the lesion and surrounding brain tissue can improve lesion contrast in LAMA. As shown in Figure 1, the optimized LAMA amplifies lesion contrast for a wide range of MS lesion T1 values ranging from 1600 to 2400 milliseconds. By contrast, the T1 prolongation effect can be detrimental for T2-weighted techniques such as FLAIR, as T1-induced signal change is opposite to the desirable T2 contrast, therefore reducing lesion contrast.32 Although DIR can partially mitigate the lesion contrast reduction, it suffers from low image SNR, resulting in inferior performance compared with LAMA.20 Third, unlike DIR and TSE sequences that heavily use radiofrequency-intensive refocusing pulses, the 3D GRE technique used in LAMA only uses small flip angle excitations followed by a single magnetization preparation pulse. It therefore can substantially reduce specific absorption rate and allow more data to be acquired in the same scan duration by improved acquisition efficiency. In addition, higher flip angles used in TSE sequences render them more susceptible to image nonuniformity issues due to B1+ inhomogeneity at 7 T. Fourth, LAMA can be integrated into the MP2RAGE sequence,9,33 which allows simultaneous acquisition of LAMA and traditional 3D T1-weighted images, further improving data acquisition efficiency. Lastly, the use of synthetic inversion images using T1 maps from the MP2RAGE acquisition has been shown to reliably produce multiple contrasts from a single acquisition (eg, edge-enhancing gradient echo, fast GM acquisition T1 inversion recovery, etc). LAMA images can theoretically be generated using a similar technique, while also allowing multiple other synthetic inversion images.34

There are some limitations to the current study. First, to evaluate the independent value of each sequence, they were read in isolation. This would be expected to result in a lower lesion detection rate and interrater agreement than if the sequences were reviewed in combination. Second, our study population was exclusively composed of patients with MS. Therefore, it remains uncertain whether other forms of WM disease would demonstrate similar superiority, necessitating additional research. Third, our evaluation was limited to LAMA at 7 T. Although this technique is theoretically adaptable to 1.5/3 T systems, further investigations are needed to assess its comparative advantage over other sequences at these reduced field strengths. Finally, our study did not directly compare LAMA with the traditional T1-weighted sequences. However, compared with the traditional T1-weighted images such as MPRAGE or the UNI image of MP2RAGE, LAMA is advantageous because the lesion signal is attenuated, which improves the relative contrast between the surrounding WM and lesions, facilitating improved lesion detection.

CONCLUSIONS

Infratentorial demyelinating lesions are a critical observation in MS, yet detection of these lesions has historically been poor. Although 7 T is highly advantageous in the evaluation of MS, the limited detection of infratentorial lesions is further compounded on 7 T. LAMA is a novel approach to attenuate lesion signal, providing high lesion-to-background contrast and greater robustness against many artifacts that plague 7 T TSE sequences, such as B1+ transmit heterogeneity. The ability to integrate this sequence with routine MP2RAGE acquisition provides an additional tool for accurately characterizing the extent of disease in MS.

REFERENCES 1. Poser S, Wikström J, Bauer HJ. Clinical data and the identification of special forms of multiple sclerosis in 1271 cases studied with a standardized documentation system. J Neurol Sci. 1979;40:159–168. 2. Nakashima I, Fujihara K, Okita N, et al. Clinical and MRI study of brain stem and cerebellar involvement in Japanese patients with multiple sclerosis. J Neurol Neurosurg Psychiatry. 1999;67:153–157. 3. Yang Y, Wang M, Xu L, et al. Cerebellar and/or brainstem lesions indicate poor prognosis in multiple sclerosis: a systematic review. Front Neurol. 2022;13:874388. 4. Fisniku LK, Brex PA, Altmann DR, et al. Disability and T2 MRI lesions: a 20-year follow-up of patients with relapse onset of multiple sclerosis. Brain. 2008;131:808–817. 5. Habek M. Evaluation of brainstem involvement in multiple sclerosis. Expert Rev Neurother. 2013;13:299–311. 6. Skorić MK, Adamec I, Madarić VN, et al. Evaluation of brainstem involvement in multiple sclerosis. Can J Neurol Sci. 2014;41:346–349. 7. Bruschi N, Boffa G, Inglese M. Ultra-high-field 7-T MRI in multiple sclerosis and other demyelinating diseases: from pathology to clinical practice. Eur Radiol Exp. 2020;4:59. 8. Schlaeger S, Li HB, Baum T, et al. Longitudinal assessment of multiple sclerosis lesion load with synthetic magnetic resonance imaging—a multicenter validation study. Invest Radiol. 2023;58:320–326. 9. Fartaria MJ, O'Brien K, Sorega A, et al. An ultra-high field study of cerebellar pathology in early relapsing-remitting multiple sclerosis using MP2RAGE. Invest Radiol. 2017;52:265–273. 10. Fartaria MJ, Sati P, Todea A, et al. Automated detection and segmentation of multiple sclerosis lesions using ultra-high-field MP2RAGE. Invest Radiol. 2019;54:356–364. 11. Okada T, Fujimoto K, Fushimi Y, et al. Neuroimaging at 7 tesla: a pictorial narrative review. Quant Imaging Med Surg. 2022;12:3406–3435. 12. Özütemiz C, White M, Elvendahl W, et al. Use of a commercial 7-T MRI scanner for clinical brain imaging: indications, protocols, challenges, and solutions—a single-center experience. AJR Am J Roentgenol. 2023. Publish ahead of print. 13. Opheim G, van der Kolk A, Markenroth Bloch K, et al. 7T epilepsy task force consensus recommendations on the use of 7T MRI in clinical practice. Neurology. 2021;96:327–341. 14. Platt T, Ladd ME, Paech D. 7 tesla and beyond: advanced methods and clinical applications in magnetic resonance imaging. Invest Radiol. 2021;56:705–725. 15. Young GS, Kimbrell V, Seethamraju R, et al. Clinical 7T MRI for epilepsy care: value, patient selection, technical issues, and outlook. J Neuroimaging. 2022;32:377–388. 16. Vachha B, Huang SY. MRI with ultrahigh field strength and high-performance gradients: challenges and opportunities for clinical neuroimaging at 7 T and beyond. Eur Radiol Exp. 2021;5:35. 17. Van de Moortele PF, Pfeuffer J, Glover GH, et al. Respiration-induced B0 fluctuations and their spatial distribution in the human brain at 7 tesla. Magn Reson Med. 2002;47:888–895. 18. Marques JP, Kober T, Krueger G, et al. MP2RAGE, a self bias-field corrected sequence for improved segmentation and T1-mapping at high field. Neuroimage. 2010;49:1271–1281. 19. Tao S, Zhou X, Greco E, et al. Edge-enhancing gradient-echo MP2RAGE for clinical epilepsy imaging at 7T. AJNR Am J Neuroradiol. 2023;44:268–270. 20. Burkett BJ, Fagan AJ, Felmlee JP, et al. Clinical 7-T MRI for neuroradiology: strengths, weaknesses, and ongoing challenges. Neuroradiology. 2021;63:167–177. 21. Deichmann R, Haase A. Quantification of T1 values by SNAPSHOT-FLASH NMR imaging. J Magn Reson. 1992;96:608–612. 22. Tao S, Zhou X, Westerhold EM, et al. Optimization of fast gray matter acquisition T1 inversion recovery (FGATIR) on 7T MRI for deep brain stimulation targeting. Neuroimage. 2022;252:119043. 23. Deniz CM, Vaidya MV, Sodickson DK, et al. Radiofrequency energy deposition and radiofrequency power requirements in parallel transmission with increasing distance from the coil to the sample. Magn Reson Med. 2016;75:423–432. 24. Zelinski AC, Angelone LM, Goyal VK, et al. Specific absorption rate studies of the parallel transmission of inner-volume excitations at 7T. J Magn Reson Imaging. 2008;28:1005–1018. 25. Zhu Y. Parallel excitation with an array of transmit coils. Magn Reson Med. 2004;51:775–784. 26. Fagan AJ, Welker KM, Amrami KK, et al. Image artifact management for clinical magnetic resonance imaging on a 7 T scanner using single-channel radiofrequency transmit mode. Invest Radiol. 2019;54:781–791. 27. Thompson AJ, Banwell BL, Barkhof F, et al. Diagnosis of multiple sclerosis: 2017 revisions of the McDonald criteria. Lancet Neurol. 2018;17:162–173. 28. Gaetani L, Fanelli F, Riccucci I, et al. High risk of early conversion to multiple sclerosis in clinically isolated syndromes with dissemination in space at baseline. J Neurol Sci. 2017;379:236–240. 29. Tintore M, Rovira A, Arrambide G, et al. Brainstem lesions in clinically isolated syndromes. Neurology. 2010;75:1933–1938. 30. Barzegar M, Najdaghi S, Afshari-Safavi A, et al. Early predictors of conversion to secondary progressive multiple sclerosis. Mult Scler Relat Disord. 2021;54:103115. 31. Uludag K, Muller-Bierl B, Ugurbil K. An integrative model for neuronal activity-induced signal changes for gradient and spin echo functional imaging. Neuroimage. 2009;48:150–165. 32. Zwanenburg JJM, Hendrikse J, Visser F, et al. Fluid attenuated inversion recovery (FLAIR) MRI at 7.0 Tesla: comparison with 1.5 and 3.0 Tesla. Eur Radiol. 2010;20:915–922. 33. Testud B, Fabiani N, Demortière S, et al. Contribution of the MP2RAGE 7T sequence in MS lesions of the cervical spinal cord. AJNR Am J Neuroradiol. 2023;44:1101–1107. 34. Middlebrooks EH, Tao S, Zhou X, et al. Synthetic inversion image generation using MP2RAGE T1 mapping for surgical targeting in deep brain stimulation and lesioning. Stereotact Funct Neurosurg. 2023;101:326–331.