The Future of Multiple Sclerosis Imaging

November 24, 2020
Pascal Sati, PhD

NeurologyLive, November 2020, Volume 3, Issue 6

Advanced imaging technology collides with artificial intelligence in what is shaping up to be a revolutionary period for the diagnosis and management of MS.

Brain imaging research in multiple sclerosis (MS) is experiencing an acceleration of technological progress. Fueled by the relentless pace of innovation in magnetic resonance imaging (MRI) and the recent outburst of artificial intelligence (AI), these advances are the crest of a new wave of scientific tools for the MS clinic.

Since its introduction nearly 3 decades ago, 2-dimensional (2D) multislice brain MRI sequences—such as proton-density, T2, fluid-attenuated inversion recovery (FLAIR), and T1 scans—have become the workhorses of any diagnostic or follow-up brain MRI exams for MS. The longevity of these scans, often referred to as conventional MRI, is explained by their high sensitivity in the detection of white matter lesions (WMLs) of the brain (represented as hyper- or hypointense focal areas) and their robust imaging on any 1.5-T and 3-T clinical scanners. However, recent advances in ultra-fast MRI sequences are about to completely transform the way these conventional brain scans are routinely acquired in hospitals and medical imaging centers.

Advances in Image Acquisition and Analysis

One of these new MRI sequences is simultaneous multislice (SMS), which collects multiple imaging slices concurrently.1 SMS can cut the scan time of conventional 2D multislice scans by a factor of 2 without degradation of image quality.2 Another technique, compressed sensing, collects and utilizes heavily undersampled data; this can reduce the scan time of any 2D or 3D anatomical scans by 20% to 50% while maintaining virtually equivalent image quality.3 Wave–controlled aliasing in parallel imaging is another technical breakthrough that combines an undersampling strategy with an efficient 3D encoding; this technique can massively accelerate any 3D T2, FLAIR, and T1 brain scan to reduce scan times to 1 to 2 minutes.4 Finally, AI-based image reconstruction techniques can improve the image quality of undersampled data, thus enabling shorter scans.5 This significant reduction in scan time will provide crucial benefits. First and foremost, patients will experience less discomfort as a result of a much shorter MRI exam, possibly as short as 10 minutes or even less. Second, a scan will be quickly repeated if motion affects image quality, enabling radiologists and neurologists to evaluate their patients’ images more consistently. Last, high-resolution 3D brain scans will replace conventional 2D brain scans, which produce thick-slice images, often containing large gaps between each slice. These important innovations will pave the way for widespread adoption of automated image analysis tools that provide invaluable information about MS disease activity.

Over the past decade, a multitude of image analysis techniques have been developed to automatically segment brain tissues and lesions using high-resolution 3D anatomical scans. These techniques rely on various approaches, including statistical methods (Method for Inter-Modal Segmentation Analysis6), machine-learning classifiers (Classification using DErivative-based Features7), and deep- learning algorithms (3D U-Net8). Currently, fully automated image analysis solutions powered by AI, including those from CorTechs Labs,9 Icometrix, and Quantib, are even offered as products that are approved for clinical use. Therefore, a flurry of imaging-based measures could be factored into the routine evaluation of patients with MS, including longitudinal volume changes of brain tissues, total/regional lesion load, and individual lesions.

Brain volume changes—which include whole-brain atrophy and regional brain atrophy (eg, cortical atrophy, thalamic atrophy)—are particularly useful as a measurement of brain tissue loss and have become widely accepted measures of neurodegeneration10; they are among the best predictors of physical and cognitive disability. Clinical trials evaluating new disease-modifying therapies (DMTs) for MS already incorporate brain atrophy measures as critical end points in their cohort studies. Further, ongoing efforts to improve the accuracy and reliability of these measures and build large normative reference data sets using cloud computing platforms will allow the ability to track brain atrophy routinely in individual patients. Tracking individual lesions, particularly the formation of new WMLs, is also useful to assess the biological response of a patient to DMTs targeting MS inflammatory activity. By automatically comparing a baseline 3D scan to a follow-up 3D scan, reliable detection of new lesions is now possible without using T1-enhancing lesions,11 thus preventing the repetitive use of macrocyclic and linear gadolinium-based contrast agents, a clinical practice that is being intensively debated because of possible safety concerns.12 Additionally, monitoring persistent WMLs will become clinically relevant as some of the chronic lesions slowly expand over time. These slowly expanding lesions are in the spot- light of MS investigators, as they are believed to reflect the demyelination and axonal loss as a result of a smoldering inflammation behind an intact blood-brain barrier.13 Investigators suspect that this chronic inflammation drives the disease worsening observed in patients with progressive MS.14 For this reason, an ongoing race exists in the MS pharmaceutical industry to deliver the first brain-penetrant therapy that could stop this smoldering inflammation.

Other features of MS lesions may also have clinical significance. Specific lesion morphological characteristics, such as shape and surface features extracted from high-resolution 3D brain MRI scans, could help differentiate MS lesions from nonspecific WMLs.15 Additionally, lesion phenotyping based on 3D morphology may open new avenues for developing outcome measures, which would reflect the biological activity of MS lesions such as myelin repair.16 Another feature, lesion intensity, is now assessed qualitatively in the clinic. As an example, chronic black holes that are known to be areas of permanent tissue destruction are defined as severely hypointense lesions on T1-weighted scans; however, depending on the type of MRI sequence and parameters of the T1 scan, the level of hypointensity can significantly change, thus hampering the proper identification of these black holes. A quantitative measure of the
T1 signal, referred to as T1 relaxation time, would provide a standardized metric of black holes, enabling their robust discrimination in the clinic.17 Initially dedicated to research, quantitative MRI techniques measuring T1 and T2 relaxation times have been significantly improved in terms of accuracy, reproducibility, and overall workflow. Some of these techniques, such as MAGnetic resonance image Compilation (MAGiC),18 2 inversion-contrast magnetization-prepared rapid gradient echo (MP2RAGE),19 and magnetic resonance fingerprinting,20 are moving into the realm of clinical imaging. These techniques may even become the future workhorses of clinical MS brain protocols thanks in part to their additional capability of synthesizing multiple image contrasts simultaneously.21

Innovative Imaging Biomarkers

The advances discussed so far rely on MRI techniques that produce conventional image contrasts (eg, proton-density, T1, T2, and FLAIR). Unfortunately, these types of contrasts lack specificity regarding the underlying pathological mechanisms involved in the development of MS lesions. This poor specificity frequently complicates the evaluation of patients suffering from radiological mimics of MS, putting these patients at risk of receiving a misdiagnosis when there is an overreliance on MRI findings. Brain lesions related to comorbidities in patients with MS could be falsely interpreted as MS lesions and mislead treatment decisions. Therefore, a crucial need exists for MRI techniques to deliver biomarkers specific to MS pathology; currently, 2 extremely promising imaging biomarkers are in the research pipe- line to help address this issue of specificity (FIGURE).

The first and most mature biomarker is the central vein sign (CVS),22 which corresponds to the presence of a small vein running centrally through a focal lesion. The perivenular formation of MS lesions is a well-known biological phenomenon initiated by the breakdown of the blood-brain barrier, which releases inflammatory cells that radially diffuse from the inflamed vein into the surrounding brain tissue leading to focal demyelination. Advanced MRI techniques sensitive
to the magnetic properties of the brain can image these small veins thanks to the difference in magnetic susceptibility between venous blood and parenchymal tissue. One such technique is susceptibility-weighted imaging (SWI), which was initially developed to perform cerebral venograms. However, SWI scans acquire thick-slice images that do not allow adequate visualization of central veins running in a multitude of directions. For a more sensitive and robust detection of the CVS, optimized susceptibility-based techniques can be used instead. Indeed, T2*-weighted 3D echo-planar imaging (or T2*-3DEPI) and FLAIR* (a combination of FLAIR and T2*) produce high-isotropic-resolution 3D images that facilitate the identification of brain lesions with central veins.23,24 From a clinical perspective, CVS will be helpful for the differential diagnosis of MS. Multiple clinical studies have recently demonstrated that patients with MS have a significantly higher proportion of brain lesions with CVS than patients with migraine,25 small vessel disease,26 neuromyelitis optica spectrum disorder,27 and other inflammatory vasculopathies.28 Encouraged by these results, an ongoing multicenter study is investigating the diagnostic potential of CVS in undiagnosed patients with typical or atypical onset. By comparing the current diagnostic criteria against the prediction based on CVS detection, this large-scale prospective study could demonstrate a significant gain in speed and accuracy of MS diagnosis.29 Another potential benefit of CVS could be the detection of the disease even before its clinical manifestation. Indeed, patients with radiological findings but without any history of clinical symptoms can have a significant proportion of their brain lesions with CVS, suggestive of subclinical MS disease.30

The second biomarker is the so-called paramagnetic rim lesions (PRLs). Using the susceptibility-based MRI techniques introduced above, along with advanced quantitative susceptibility mapping post-processing techniques, investigators very recently demonstrated that some MS lesions possess a paramagnetic rim located at their edges.31,32 This rim appears to correspond to the accumulation of iron-laden macrophages or microglia, which are thought to be the generation of the smoldering inflammation discussed previously. Although not all patients with MS have PRLs, those with PRLs clearly experience a more aggressive disease.33 These findings point to the potential value of PRLs as a prognostic biomarker. In addition, PRLs appear to be very specific to MS lesions34 and may be a complementary biomarker for MS diagnosis. To prepare for future clinical use of these biomarkers, automated solutions for the detection of CVS and PRLs are already in development using probabilistic methods,35 as well as deep-learning algorithms (CVSnet36 and RimNet37).

Advanced MRI Devices

Although imaging WMLs is the bread and butter of clinical MRI, cortical lesions are seen with increasing clinical importance and will become important radiological findings for predicting disease progression as well as evaluating new therapies for neuroprotection. Because of their specificity regarding MS pathology, cortical lesions are now part of the MRI diagnostic criteria for MS. However, in practice, these lesions remain difficult to detect reliably using conventional MRI scans. Newer techniques such as phase-sensitive inversion recovery38 and the very recent inversion-recovery susceptibility-weighted imaging with enhanced T2 weighting39 can improve the visualization of cortical lesions on 3T scanners, especially the detection of subpial lesions associated with cognitive decline and disease progression. Meanwhile, another powerful strategy for imaging cortical lesions with high sensitivity utilizes scanners with a stronger magnetic field, such as the 7-T scanner.

Seven-tesla MRI produces exquisite images of the human brain by achieving very high spatial resolution—on the order of hundreds of micrometers. This high-end imaging device has been intensively used in clinical research over the past 10 years to noninvasively investigate MS pathology developing in patients’ brains.40 This sustained research effort has brought new insights into the understanding of cortical and white matter MS lesions41,42 and sparked the discoveries of the CVS and PRLs. The recent FDA approval of the first 7-T MRI for clinical use is another major step forward for MS imaging, as it will enrich the clinician’s toolbox for diagnosing MS43 and speed up the clinical translation of future imaging biomarkers yet to be discovered with this cutting-edge imaging technology.

Without a doubt, MS imaging is set to undergo an exciting technological transformation. The next-generation MRI techniques discussed here will provide clinicians with an unprecedented wealth of information about this complex and multifaceted disease. By fully embracing these technological and scientific shifts on the horizon, neurologists will enhance their ability to diagnose earlier and more accurately, identify and anticipate the trajectory of the disease, and adapt treatment strategies to keep the irreversible progression of the disease at bay. Lastly, integrating rapidly innovative imaging biomarkers into the clinical trials of emerging drugs—especially those promoting neuroprotection and neurorepair—will help finally turn the tables on this debilitating disease. All in all, the future of MS imaging is very bright.

REFERENCES
1. Feinberg DA, Setsompop K. Ultra-fast MRI of the human brain with simultaneous multi-slice imaging. J Magn Reson. 2013;229:90-100. doi:10.1016/j.jmr.2013.02.002
2. Kean M, Coleman L, Josan S, Schmitt B. Epilepsy imaging with simultaneous multi-slice turbo spin echo. MAGNETOM Flash. 2015(63):96-99. https://cdn0.scrvt.com/39b415fb07de4d9656c7b516d8e2d907/1800000002474939/798edf786b9b/Epilepsy_Imaging_with_SMS_TSE_1800000002474939.pdf 
3. Mönch S, Sollmann N, Hock A, Zimmer C, Kirschke JS, Hedderich DM. Magnetic resonance imaging of the brain using compressed sensing - quality assessment in daily clinical routine. Clin Neuroradiol. 2020;30(2):279-286. doi:10.1007/s00062-019-00789-x
4. Polak D, Cauley S, Huang SY, et al. Highly-accelerated volumetric brain examination using optimized wave-CAIPI encoding. J Magn Reson Imaging. 2019;50(3):961-974. doi:10.1002/jmri.26678
5. Werth K, Ledbetter L. Artificial intelligence in head and neck imaging: a glimpse into the future. Neuroimaging Clin N Am. 2020;30(3):359-368. doi:10.1016/j.nic.2020.04.004
6. Valcarcel AM, Linn KA, Vandekar SN, et al. MIMoSA: an automated method for intermodal segmentation analysis of multiple sclerosis brain lesions. J Neuroimaging. 2018;28(4):389-398. doi:10.1111/jon.12506
7. Selvaganesan K, Whitehead E, DeAlwis PM, et al. Robust, atlas-free, automatic segmentation of brain MRI in health and disease. Heliyon. 2019;5(2):e01226. doi:10.1016/j.heliyon.2019.e01226
8. La Rosa F, Abdulkadir A, Fartaria MJ, et al. Multiple sclerosis cortical and WM lesion segmentation at 3T MRI: a deep learning method based on FLAIR and MP2RAGE. Neuroimage Clin. 2020;27:102335. doi:10.1016/j.nicl.2020.102335
9. CorTechs Labs receives US FDA clearance for quantitative brain image analysis. News release. CorTechs Labs. September 19, 2017. Accessed October 12, 2020. https://www.cortechslabs.com/company/news/fda-510k-clearance-2017/
10. Rocca MA, Battaglini M, Benedict RH, et al. Brain MRI atrophy quantification in MS: from methods to clinical application. Neurology. 2017;88(4):403-413. doi:10.1212/WNL.0000000000003542
11. Schmidt P, Pongratz V, Küster P, et al. Automated segmentation of changes in FLAIR-hyperintense white matter lesions in multiple sclerosis on serial magnetic resonance imaging. Neuroimage Clin. 2019;23:101849. doi:10.1016/j.nicl.2019.101849
12. McDonald JS, McDonald RJ. MR imaging safety considerations of gadolinium-based contrast agents: gadolinium retention and nephrogenic systemic fibrosis. Magn Reson Imaging Clin N Am. 2020;28(4):497-507. doi:10.1016/j.mric.2020.06.001
13. Elliott C, Arnold DL, Chen H, et al. Patterning chronic active demyelination in slowly expanding/evolving white matter MS lesions. AJNR Am J Neuroradiol. 2020;41(9):1584-1591. doi:10.3174/ajnr.A6742
14. Elliott C, Belachew S, Wolinsky JS, et al. Chronic white matter lesion activity predicts clinical progression in primary progressive multiple sclerosis. Brain. 2019;142(9):2787-2799. doi:10.1093/brain/awz212
15. Newton BD, Wright K, Winkler MD, et al. Three-dimensional shape and surface features distinguish multiple sclerosis lesions from nonspecific white matter disease. J Neuroimaging. 2017;27(6):613-619. doi:10.1111/jon.12449
16. Sivakolundu DK, Hansen MR, West KL, et al. Three-dimensional lesion phenotyping and physiologic characterization inform remyelination ability in multiple sclerosis. J Neuroimaging. 2019;29(5):605-614. doi:10.1111/jon.12633
17. Thaler C, Faizy T, Sedlacik J, et al. T1- thresholds in black holes increase clinical-radiological correlation in multiple sclerosis patients. PLoS One. 2015;10(12):e0144693. doi:10.1371/journal.pone.0144693
18. Tanenbaum LN, Tsiouris AJ, Johnson AN, et al. Synthetic MRI for clinical neuroimaging: results of the Magnetic resonance image Compilation (MAGiC) prospective, multicenter, multireader trial. AJNR Am J Neuroradiol. 2017;38(6):1103-1110. doi:10.3174/ajnr.A5227
19. Kober T, Granziera C, Ribes D, et al. MP2RAGE multiple sclerosis magnetic resonance imaging at 3T. Invest Radiol. 2012;47(6):346-352. doi:10.1097/RLI.0b013e31824600e9
20. Hsieh JJL, Svalbe I. Magnetic resonance fingerprinting: from evolution to clinical applications. J Med Radiat Sci. Published online June 28, 2020. doi:10.1002/jmrs.413
21. Hagiwara A, Warntjes M, Hori M, et al. SyMRI of the brain: rapid quantification of relaxation rates and proton density, with synthetic MRI, automatic brain segmentation, and myelin measurement. Invest Radiol. 2017;52(10):647-657. doi:10.1097/RLI.0000000000000365
22. Sati P, Oh J, Constable RT, et al. The central vein sign and its clinical evaluation for the diagnosis of multiple sclerosis: a consensus statement from the North American Imaging in Multiple Sclerosis Cooperative. Nat Rev Neurol. 2016;12(12):714-722. doi:10.1038/nrneurol.2016.166
23. Sati P, Thomasson DM, Li N, et al. Rapid, high-resolution, whole-brain, susceptibility-based MRI of multiple sclerosis. Mult Scler. 2014;20(11):1464-1470. doi:10.1177/1352458514525868
24. Sati P, George IC, Shea CD, Gaitán MI, Reich DS. FLAIR*: a combined MR contrast technique for visualizing white matter lesions and parenchymal veins. Radiology. 2012;265(3):926-932. doi:10.1148/radiol.12120208
25. Solomon AJ, Schindler MK, Howard DB, et al. “Central vessel sign” on 3T FLAIR* MRI for the differentiation of multiple sclerosis from migraine. Ann Clin Transl Neurol. 2015;3(2):82-87. doi:10.1002/acn3.273
26. Mistry N, Abdel-Fahim R, Samaraweera A, et al. Imaging central veins in brain lesions with 3-T T2*-weighted magnetic resonance imaging differentiates multiple sclerosis from microangiopathic brain lesions. Mult Scler. 2016;22(10):1289-1296. doi:10.1177/1352458515616700
27. Cortese R, Magnollay L, Tur C, et al. Value of the central vein sign at 3T to differentiate MS from seropositive NMOSD. Neurology. 2018;90(14):e1183-e1190. doi:10.1212/WNL.0000000000005256
28. Maggi P, Absinta M, Grammatico M, et al. Central vein sign differentiates multiple sclerosis from central nervous system inflammatory vasculopathies. Ann Neurol. 2018;83(2):283-294. doi:10.1002/ana.25146
29. Central vein sign: a diagnostic biomarker in multiple sclerosis (CAVS-MS). Updated October 5, 2020. Accessed October 10, 2020. https://clinicaltrials.gov/ct2/show/NCT04495556
30. Suthiphosuwan S, Sati P, Guenette M, et al. The central vein sign in radiologically isolated syndrome. AJNR Am J Neuroradiol. 2019;40(5):776-783. doi:10.3174/ajnr.A6045
31. Absinta M, Sati P, Gaitán MI, et al. Seven-tesla phase imaging of acute multiple sclerosis lesions: a new window into the inflammatory process. Ann Neurol. 2013;74(5):669-678. doi:10.1002/ana.23959
32. Kaunzner UW, Kang Y, Zhang S, et al. Quantitative susceptibility mapping identifies inflammation in a subset of chronic multiple sclerosis lesions. Brain. 2019;142(1):133-145. doi:10.1093/brain/awy296
33. Absinta M, Sati P, Masuzzo F, et al. Association of chronic active multiple sclerosis lesions with disability in vivo. JAMA Neurol. 2019;76(12):1474-1483. doi:10.1001/jamaneurol.2019.2399. Published correction appears in JAMA Neurol. 2019;76(12):1520. doi: 10.1001/jamaneurol.2019.4331
34. Maggi P, Sati P, Nair G, et al. Paramagnetic rim lesions are specific to multiple sclerosis: an international multicenter 3T MRI study. Ann Neurol. 2020;88(5):1034-1042. doi:10.1002/ana.25877
35. Dworkin JD, Sati P, Solomon A, et al. Automated integration of multimodal MRI for the probabilistic detection of the central vein sign in white matter lesions. AJNR Am J Neuroradiol. 2018;39(10):1806-1813. doi:10.3174/ajnr.A5765
36. Maggi P, Fartaria MJ, Jorge J, et al. CVSnet: a machine learning approach for automated central vein sign assessment in multiple sclerosis. NMR Biomed. 2020;33(5):e4283. doi:10.1002/nbm.4283
37. Barquero G, La Rosa F, Kebiri H, et al. RimNet: a deep 3D multimodal MRI architecture for paramagnetic rim lesion assessment in multiple sclerosis. Neuroimage Clin. 2020;28:102412. doi:10.1016/j.nicl.2020.102412
38. Mougin O, Abdel-Fahim R, Dineen R, Pitiot A, Evangelou N, Gowland P. Imaging gray matter with concomitant null point imaging from the phase sensitive inversion recovery sequence. Magn Reson Med. 2016;76(5):1512-1516. doi:10.1002/mrm.26061
39. Beck ES, Gai N, Filippini S, Maranzano J, Nair G, Reich DS. Inversion recovery susceptibility weighted imaging with enhanced T2 weighting at 3T improves visualization of subpial cortical multiple sclerosis lesions. Invest Radiol. 2020;55(11):727-735. doi:10.1097/RLI.0000000000000698
40. Schindler MK, Sati P, Reich DS. Insights from ultrahigh field imaging in multiple sclerosis. Neuroimaging Clin N Am. 2017;27(2):357-366. doi:10.1016/j.nic.2016.12.006
41. Absinta M, Sati P, Reich DS. Advanced MRI and staging of multiple sclerosis lesions. Nat Rev Neurol. 2016;12(6):358-368. doi:10.1038/nrneurol.2016.59
42. Louapre C, Treaba CA, Barletta V, Mainero C. Ultra-high field 7 T imaging in multiple sclerosis. Curr Opin Neurol. 2020;33(4):422-429. doi:10.1097/WCO.0000000000000839
43. Sati P. Diagnosis of multiple sclerosis through the lens of ultra-high-field MRI. J Magn Reson. 2018;291:101-109. doi:10.1016/j.jmr.2018.01.022
download issueDownload Issue : November 2020