Targeting Cerebral Small Vessel Disease With MRI

HomeStrokeVol. 48, No. 11Targeting Cerebral Small Vessel Disease With MRI Free AccessReview ArticlePDF/EPUBAboutView PDFView EPUBSections ToolsAdd to favoritesDownload citationsTrack citationsPermissionsDownload Articles + Supplements ShareShare onFacebookTwitterLinked InMendeleyReddit Jump toSupplementary MaterialsFree AccessReview ArticlePDF/EPUBTargeting Cerebral Small Vessel Disease With MRI Jaco J.M. Zwanenburg, PhD and Matthias J.P. van Osch, PhD Jaco J.M. ZwanenburgJaco J.M. Zwanenburg From the Deptartment of Radiology, University Medical Center Utrecht, the Netherlands (J.J.M.Z.); and Department of Radiology, Leiden University Medical Center, the Netherlands (M.J.P.v.O.). and Matthias J.P. van OschMatthias J.P. van Osch From the Deptartment of Radiology, University Medical Center Utrecht, the Netherlands (J.J.M.Z.); and Department of Radiology, Leiden University Medical Center, the Netherlands (M.J.P.v.O.). Originally published29 Sep 2017https://doi.org/10.1161/STROKEAHA.117.016996Stroke. 2017;48:3175–3182Other version(s) of this articleYou are viewing the most recent version of this article. Previous versions: January 1, 2017: Previous Version 1 Cerebral small vessel disease (SVD) is a severely debilitating disease affecting the smallest vessels in the brain. It is a major cause of cognitive decline, dementia, and functional disability in the elderly, and it is also responsible for ≈25% of all cerebral strokes.1,2 Still, treatment options are restricted to preventive risk management because there is currently no curative treatment available. Better understanding of both the pathogenesis of SVD and the subsequent pathophysiological mechanism by which SVD leads to macroscopic brain tissue damage (size ranging from mm to cm) could stimulate the development of more effective treatments.However, the nature of SVD hampers unraveling both its pathogenesis and the subsequent pathophysiological mechanisms that lead to macroscopic brain tissue damage. In the first place, SVD is a common term, grouping various pathology of the small arteries, arterioles, capillaries, and venules in the brain, regardless of the underlying pathogenesis.1 As a result, there are multiple potential cascades linking pathophysiological processes to SVD and to subsequent brain tissue damage and cognition loss. Most of these pathophysiological processes can coexist with mutual interaction, which makes it highly complex to unravel. Second, because SVD affects the smallest vessels in the brain, these vessels and their pathology are an order of magnitude below the imaging resolution of in vivo medical imaging equipment. Accordingly, current clinical markers of SVD do not reflect SVD itself but the consequences of SVD, such as loss of cognition. Similarly, image-based markers reflect macroscopic brain tissue damage secondary to SVD, such as (small) infarcts, white matter hyperintensities, perivascular spaces, microbleeds, and brain atrophy.3 These markers are marketed as indicators of SVD, even to such an extent that SVD has become synonymous with having macroscopic brain lesions on MRI.1 Trials testing the efficacy of new therapeutic treatments that rely on these indirect markers of SVD might fail as the surrogate markers of treatment response reflect irreversible brain damage, rather than the targeted disease process.Most of the image-based markers are obtained from standard anatomic MRI scans, such as T2-weighted, susceptibility-weighted imaging, and fluid attenuated inversion recovery images. Several recent advanced MRI methods have the potential to narrow the gap between current neuroimaging markers of SVD and the underlying pathophysiological processes. The aim of this review is to give an overview of these advanced MRI methods and to indicate how these methods link to pathophysiological processes involved in SVD.By reviewing the literature on SVD, several pathophysiological targets were identified (Table I in the online-only Data Supplement) that could potentially be probed by advanced MRI methods. The Table summarizes these advanced MRI techniques, together with the related (patho)physiological targets of interest for SVD. To remain concise, we refer to key papers and reviews that provide further background information on these techniques. In this review, we will highlight the potential of these recent developments to achieve faster adoption of these technologies in SVD studies. First, we will discuss the more conventional methods (numbers 1–3 in Table). Next, we discuss methods for assessing cerebral hemodynamics and vascular reactivity (numbers 4–8). Then, we discuss methods for quantitative mapping of other tissue parameters, such as magnetic susceptibility and the relaxation parameters T1, T2, and T2* (numbers 9 and 10), which is followed by reviewing several less established methods (numbers 11–14). Finally, we will identify challenges and opportunities to stimulate effective further developments of advanced MRI methods targeting the pathophysiology of SVD.Table. Overview of Available MRI Methods That Have the Potential to Assess the Pathophysiological Targets Defined in Table I in the online-only Data SupplementMRI Sequence ClassObtained Physical QuantityRelated Pathophysiological TargetsRefs1Anatomic imagingStructural images with contrast dominated by proton density, T1, T2, T2*, and inflow effects (time-of-flight angiography)Adaptive responses of the tissue (such as enlarged perivascular spaces, atrophy, etc) Brain tissue lesions (Large) vessel remodeling32Blood oxygen level–dependent (BOLD) contrastDynamic image series with temporal signal dynamics dominated by blood oxygenation level (while blood flow and volume also influence BOLD signals)Adaptive response of vasculature to challenges (from neuronal, systemic, or pharmaceutical origin)4–63Diffusion-weighted imaging (DWI)Quantitative maps of diffusion properties Semiquantitative maps of perfusion (IVIM)Adaptive responses of the tissue (such as changes in microstructure) Perfusion7,84Velocity phase-contrast (PC)Quantitative blood flow velocity Pulse wave velocityGlobal blood flow Vessel wall compliance95Arterial spin labeling (ASL)Quantitative maps of tissue perfusion, arterial blood volume, and arterial and tissue transit times Quantitative T2 of labeled spins (in combination with quantitative T2 mapping)Perfusion Small vessel remodeling Water transport across blood–brain barrier106Contrast agent–based perfusion MRI (dynamic susceptibility contrast [DSC] MRI and dynamic contrast enhancement [DCE] MRI)Quantitative maps of tissue perfusion, blood volume, and mean transit time Capillary transit time heterogeneity Quantitative maps of ktrans Vessel size imaging (DSC with combined T2 and T2* readout)Perfusion Small vessel remodeling Blood–brain barrier leakage11–137Vascular space occupancy (VASO)Dynamic imaging series with temporal signal dynamics dominated by blood volume changesAdaptive response of vasculature to challenges (from neurological, systemic, or pharmaceutical origin)148Reactivity measurementsChange in a physiological (eg, velocity) or MRI parameter (eg, T2*) in response to a vasodilatory or neuronal challengeVasodilatory capacity Hemodynamic reserve capacity159Susceptibility-weighted imaging (SWI), using signal phaseSemiquantitative maps of the magnetic susceptibility for both tissue and bloodAdaptive responses of the tissue (such as changes in tissue microstructure and content) Quantitative blood oxygenation levels Iron deposits1610Quantitative relaxation parameter mappingQuantitative maps of the relaxation parameters T1, T2, or T2* for both tissue and bloodAdaptive responses of the tissue (such as changes in tissue microstructure and content) Quantitative blood oxygenation levels1711Magnetic resonance elastography (MRE)Quantitative maps of tissue elasticity properties Volumetric strain of tissue over the cardiac cycle, induced by pulsation of blood volume within the tissueAdaptive responses of the tissue (such as changes in tissue microstructure and content) Vessel wall compliance18–2012Magnetization transfer (MT)Structural images with contrast weighting dominated by MT between different water poolsAdaptive responses of the tissue (such as demyelination, loss of macromolecules, and changed free water content)21,2213Chemical exchange saturation transfer (CEST)Structural images with contrast weighting dominated by MT between different chemical compoundsAdaptive responses of the tissue (including changes in metabolism)2314Spectroscopic imaging(Low resolution) maps of the concentration of several metabolitesAdaptive responses of the tissue and metabolism24,25The references refer to reviews or key papers. IVIM indicates IntraVoxel Incoherent Motion.Conventional MRI MethodsAlthough anatomic imaging, diffusion-weighted imaging, functional MRI, and angiography (see the Table) can be considered conventional because of their widespread use, they are still in vivid development. Increasingly, detailed tissue anatomy and subtle damage can be studied as a result of the ongoing advancement of imaging hardware, including the introduction of ultrahigh field MRI (7 Tesla and beyond) and high-density receive arrays for signal reception. Examples are the improved depiction of penetrating arteries and small draining veins, enlarged perivascular spaces, microbleeds, microinfarcts, and using diffusion-weighted imaging, even acute microinfarcts. Figure 1 illustrates the joint imaging of microbleeds and the adjacent penetrating artery in a patient with a hypertensive hemorrhage.Download figureDownload PowerPointFigure 1. Coronal minimum (A) and maximum (B) intensity projections over a thin (4-mm thick) volume, generated from the same scan, showing microbleeds (arrows) and a penetrating artery (B). Combination of these projections (C) reveals a direct correlation between the microbleeds and the artery. H indicates head (direction); and L, left. Reproduced with permission from Biessels et al.26The use of these conventional methods is further enhanced by the ongoing increase in computational power, which enables increasingly sophisticated data analysis. This allows studying the brain as a network in various ways. Structural connectivity can be studied by performing fiber tracking on diffusion-weighted imaging data, whereas functional connectivity can be studied with blood oxygen level–dependent (BOLD) images that are acquired while the subject is awake but in rest (resting state functional MRI). Another example is structural network covariance analysis that allows studying normal brain development and disease-related degeneration at the network level, by using ordinary T1-weighted anatomic images from a large number of patients.27 Structural network covariance analysis has shown that several distinct anatomic areas (structural networks) change together in healthy development, ageing, as well as during the development of dementia. Finally, lesion-symptom mapping allows analyzing the relationship between the location of a lesion and the clinical symptoms of the patients.28 Using lesion-symptom mapping, it has been shown that lesions at strategic locations can explain considerably more variance of cognitive performance than simple lesion measures like total lesion volume.28 Similarly, it has been shown that recent small subcortical infarcts are related to regional subsequent atrophy.29,30 The increasing ability of MRI to assess more subtle tissue damage and to relate this to the brain as a network has the potential to elucidate part of the pathophysiological processes that tie SVD to subsequent brain dysfunction.MRI Methods for Assessing Blood Flow and Other Baseline Hemodynamic ParametersBeyond structural imaging, MRI is capable of obtaining functional measures targeted at the hemodynamics (methods 4–8; Table). Quantitative measurement of blood flow velocities is obtained by phase-contrast MRI. Traditionally, phase-contrast MRI is used in the main brain-feeding arteries only. Regional hypoperfusion, for example, in the border zones or deep white matter, can therefore not be detected. A second limitation is that the flow in the main arteries does not tell how much tissue is supplied by these arteries. Normalization by total brain volume allows for measuring mean whole brain perfusion.31 New developments in MRI technology enable to perform phase-contrast measurements in much smaller intracranial arteries. Recently, pulsatility measurements in white matter perforating arteries <300 µm have been reported (Figure 2). This allows to study the relationship between SVD and pulsatility relatively close to the microvessels where the impact of the hammering effects of the cardiac cycle is assumed to be the largest.32Download figureDownload PowerPointFigure 2. Blood flow velocity measurements in the white matter perforators of a healthy human subject, obtained with high-resolution 2-dimensional phase-contrast MRI, similar to Bouvy et al.32A, Mean magnitude over the cardiac cycle showing the perforators as hyperintense dots. Enlarged detail: magnitude and corresponding velocity map at single cardiac time point. B, Mean velocity time curve (average over all detected perforators [n=55]) and corresponding 95% confidence interval. Image courtesy: Lennart Geurts.MRI offers 2 families of techniques to measure blood flow further downstream in the vascular tree. Arterial spin labeling MRI exploits the blood itself as a noninvasive tracer by inverting the magnetization of blood in the brain-feeding arteries. Interestingly, the water transport time across the blood–brain barrier, which may reflect blood–brain barrier function, can be measured by recording the transverse relaxation time of the arterial spin labeling.33The second family of magnetic resonance perfusion techniques uses an exogenous contrast agent to measure perfusion parameters, either by using dynamic T2(*)-weighted imaging (dynamic susceptibility contrast [DSC] MRI) or dynamic T1-weighted imaging (dynamic contrast enhanced MRI). From these measurements, multiple hemodynamic parameters can be obtained, including cerebral blood flow, cerebral blood volume, and mean transit time (for DSC and to a lesser extent for dynamic contrast enhanced), blood–brain barrier permeability (dynamic contrast enhanced), mean vessel size (DSC when acquired with combined T2- and T2*-weighted imaging), capillary transit time heterogeneity (DSC), and timing parameters (DSC and dynamic contrast enhanced).11–13 Probably, the most promising application of contrast agent–based perfusion MRI is the measurement of subtle blood–brain barrier damage, for which no validated noninvasive alternatives are currently available. The use of perfusion MRI in studies of SVD has been recently summarized.34 Finally, MRI offers also the possibility to measure cerebral blood volume noninvasively with an inversion recovery sequence timed to null the blood signal (vascular space occupancy).14 Applications of vascular space occupancy have, however, been mainly limited to neuroscience.An interesting twist to an existing conventional technique is the use of BOLD MRI for studying the hemodynamics of the (micro)vasculature. BOLD functional MRI is normally used to study brain neuronal function under the assumption of comparable hemodynamic response functions between and within subjects. However, alterations in hemodynamics with aging and disease affect the neurovascular coupling and complicate the interpretation of functional MRI results.35 This can be turned around by using BOLD functional MRI for studying the hemodynamics, for example, by applying a neuronal stimulus that is largely independent of the performance of the subject, such as a checkerboard visual stimulus. Recent examples show the high potential of this approach,36,37 as also illustrated in Figure 3.Download figureDownload PowerPointFigure 3. Blood oxygen level–dependent (BOLD) response to a visual stimulus, representative for vessel reactivity within the occipital lobe. The BOLD response clearly shows reduced vascular reactivity in (A) symptomatic hereditary cerebral hemorrhage with amyloidosis-Dutch type (HCHWA-D) mutation carriers and (B) presymptomatic HCHWA-D mutation carriers vs age-matched controls. Blue boxes show the time of activation of the visual stimulus. Data points are mean group responses and shaded areas are SEs. Reproduced with permission from van Opstal et al.37Reactivity MeasurementsOne of the most direct efforts to quantify SVD relies on measuring the response of the microvasculature to a physiological challenge (also named reserve capacity). Reduced reserve capacity could point to either an exhausted autoregulation (ie, the microvasculature is capable to change its caliber and thus its resistance, but the vessel are maximally dilated) or to a failure in vasodilatory capacity, which may be because of loss of smooth muscle cells, stiffening of the vessel walls, or poor functioning of the pericytes (recent insights have attributed the main vasodilatory capacity to the pericytes instead of the more traditionally assumed arterioles38).Typical challenges for stressing the hemodynamic system, such as a tilting table test or squatting, are not feasible inside an MRI scanner. The most widespread used challenges in MRI are injection of acetazolamide, hypercapnic challenges by manipulations of the air breathed by the subject, breath-holding, and neuronal activation. More experimental options include lower-body-negative-pressure challenges, compression tests, in-bore cycling, and cold-pressor tests. As readout modality, many different MRI sequences can be used, each targeting another aspect of the cerebral hemodynamics. Most popular is BOLD imaging that provides high signal-to-noise ratios but at the expense of a more complicated interpretation of the results due the mixed origin of the BOLD effect. Also, arterial spin labeling, phase-contrast MRI, and vascular space occupancy are commonly applied readout modalities. Current available studies on cerebrovascular reactivity MRI are summarized in a recent meta-analysis.39 Recently, it was shown that spontaneous low frequency fluctuations in BOLD signal provide a measure for cerebrovascular reactivity without the need for external challenges.40MRI Methods Quantifying Other Tissue ParametersWhereas MRI is based on signal from water (ie, from the hydrogen atoms in water), the properties of this signal (the relaxation parameters T1, T2, and T2*) are determined by the microenvironment in which the water resides. Alterations in, for example, the oxygen content or iron concentration will subsequently affect these relaxation parameters. MRI is unique in that it can quantitatively map these parameters, which provides potential probes for assessing oxygen consumption and iron deposits in SVD. Also the magnetic susceptibility of tissue (the amount by which the tissue locally changes the magnetic field) depends on tissue properties and can be quantitatively assessed by MRI and advanced image processing.Key challenge here is to design the MRI methods in such a way that the observed variation in magnetic susceptibility, T1, T2, or T2* can be uniquely tied to a single physiological target, such as the oxygenation level of blood. For blood, a straightforward way to do so is to quantify the difference between arterial and venous blood. Any difference in susceptibility, T2, or T2* will reflect the difference in oxygenation,41,42 which in turn can be studied between patient groups, between different veins, over time, or in response to a stimulus or drugs. (T1 is less sensitive to changes in oxygen level, at least at 1.5 T, and rather depends on the hematocrit level.43) Drawback of this approach is that it measures the oxygenation in relatively large draining veins, which does not provide detailed spatial information on where the oxygen is consumed. Recent advances in modeling have enabled to obtain quantitative susceptibility maps from the phase signal of T2*-weighted scans, which generally have a submillimeter resolution, thus allowing to study smaller veins. A different approach that can provide information on oxygen on a tissue level is quantitative BOLD.44 Here, the different behavior of gradient echoes (T2* weighted) and spin echoes (T2 weighted) in response to a change in the susceptibility (oxygenation) of the microvasculature is used to estimate the oxygenation of the microvasculature. One should keep in mind that this method assumes that the observed changes in transverse relaxation are because of different concentrations of deoxyhemoglobin, whereas other sources of changes in magnetic susceptibility (such as iron or myelin) are not taken into account. By combining such approaches with velocity-selective spin labeling, the origin of signal can be restricted to veins, thereby improving the specificity.45When focusing on tissue rather than on the vasculature, regional magnetic susceptibility differences reflect differences in iron accumulation or in myelin concentration, which both are relevant factors for understanding SVD. Although amyloid β does shorten the T2 and T2* of the MRI signal, it does only do so when deposited in the fibrillar amyloid conformation. Diffuse amyloid β plaques do not significantly alter the relaxation parameters,46 which makes MRI of limited value for studying amyloid β directly.Less Established MRI Methods in SVD ResearchThis section briefly lists 4 different MRI techniques that are not yet commonly used in studying SVD (numbers 11–14 in the Table). For brevity, we only elaborate on the first technique, magnetic resonance elastography. The second technique, magnetization transfer, has been widely used to study demyelination in multiple sclerosis. Magnetization transfer also has potential in studying tissue degeneration in ageing and dementia, for which we refer to reviews.21,22 The third technique, chemical exchange saturation transfer, has the potential to assess tissue pH and a limited number of metabolites, such as myo-inositol (a marker for glial cell proliferation) and glutamate. Chemical exchange saturation transfer is much more sensitive than spectroscopy by using exchange phenomena, but less specific, which makes it more complicated to interpret. We refer to the following reviews for more details.23,47 Applications of chemical exchange saturation transfer in dementia have just started with a few studies in rodents. The last technique mentioned here is spectroscopic imaging that can quantitatively assess many metabolites. Spectroscopy has been thoroughly discussed in relation to various types of dementia in several reviews.24,25,48 An intriguing recent finding that deserves follow-up in SVD is that diffusion-weighted spectroscopy may provide a marker for glial reactivity in response to inflammation.49Magnetic resonance elastography can quantify mechanical (viscoelastic) properties of brain tissue, such as stiffness. It can be seen as an innovation of palpation, being both quantitative and able to assess nonpalpable tissue like brain within the skull. Conventional magnetic resonance elastography uses an external motion driver to induce shear waves in the (brain) tissue (typical driver frequencies range from 10 to 100 Hz). With dedicated MRI, these waves can be imaged from which mechanical tissue properties can be derived.20 A change in tissue microstructure, because of subtle (disease) processes, alters the tissues mechanical properties. Brain tissue softens with various diseases, including multiple sclerosis, Parkinson disease, and Alzheimer disease.20 Interestingly, a close relationship between the stiffness and local perfusion was found in the striatum.50 Besides, brain softening with healthy ageing has been shown, which occurs at a faster relative rate than brain volume loss with ageing.51 Thus, magnetic resonance elastography has the potential to assess tissue alterations in SVD before the onset of clinical symptoms.An intriguing development is the use of the heartbeat itself as intrinsic driver for magnetic resonance elastography.18,19 First, the hammering of cardiac pulsations on the arterioles and capillaries has been proposed as an important mechanism for vessel wall changes, as well as a key parameter in drainage of waste products.52 Therefore, by taking the heartbeat as driver, one focuses on a potential causal factor of pathological changes in SVD. Second, these cardiac-related pulsations also reveal properties of the (micro)vascular bed from which they originate. For example, one can measure the pulsation in blood volume of the microvascular bed as shown in Figure 4. This has great potential for assessing how the pulsations in the microvasculature alter with ageing when vessel wall compliance decreases and, thus, dampening of the pulse pressure in the large elastic arteries is diminished. At the same time, one should note that breathing and changing body posture potentially induce larger challenges to the brain's microcirculation than the heartbeat.53,54 Future work should assess these effects too.Download figureDownload PowerPointFigure 4. Magnetic resonance elastography using the heartbeat as intrinsic driver can measure the subtle volume change of brain tissue that is induced by the pulsation of the small blood vessels within the tissue. A, Volumetric strain (change in volume relative to original volume) for a region of interest in the white matter, averaged for 6 healthy volunteers. B, Schematic representation of brain tissue (gray) permeated by blood vessels (black) during systole. The MRI signal is integrated over the entire voxel (dashed line), giving rise to an effectively measured displacement during the heartbeat. The inflow of blood during systole (black arrows) causes dilatation or positive volumetric strain (gray arrows). C, Outflow of blood during diastole causes compression (deflation) of the tissue. Assuming that the brain tissue is incompressible, that is, the size of the gray area does not change, a change in tissue volume reflects a change in blood volume. Given a typical blood volume of 2% in white matter, the observed strain would correspond to a blood volume change of only 2.8×10–4/0.02=1.4% in these healthy subjects. Figure adapted with permission from Hirsch et al.19 Authorization for this adaptation has been obtained both from the owner of the copyright in the original work and from the owner of copyright in the translation or adaptation.Challenges and OpportunitiesThe extensive list of MRI techniques and corresponding pathophysiological targets clearly show the potential of advanced MRI techniques in SVD research. These techniques have the potential to narrow the gap between the invisible SVD-related alterations of the smallest vessels in the brain on one hand and perceivable tissue damage and clinical symptoms on the other hand. Many of these advanced techniques may not evolve to widespread use in routine clinical diagnosis because they require either ultrahigh field scanners, long scan time, advanced image processing with mathematical modeling, or all of these limiting factors. Nonetheless, they may provide clues that help unraveling the pathophysiological processes involved in SVD when applied in carefully designed studies on selected patient groups. They may also allow to study target involvement in drug trials. In the long run, they may allow the community to change the currently so rightful consensus opinion as formulated in the STandards for ReportIng Vascular changes on nEuroimaging (STRIVE), which endorsed the principle that terms and definitions should indicate imaging characteristics as descriptively as possible, avoiding presumptions of mechanism or pathological links.3Main challenge for existing methods is to avoid confounding factors because virtually all methods are not exclusively sensitive for a single target. Studies aiming to assess a certain target should explicitly state the underlying assumptions and provide insight in the performance by proper validation and simulations. An example is the validation of arterial spin labeling perfusion MRI to positron emission tomography imaging in target subjects, such as patients with Alzheimer disease.55 Another challenge is to translate the advanced MRI methods to applications in patient studies. By far most research is done with relatively basic MRI methods. Although these methods are still powerful and sound, the advanced MRI methods, which continuously evolve, hold even more promise for better understanding of SVD.One of the biggest unmet needs is a technique that is capable to target the waste clearance system of the brain. Ideally, this method should be based on intrinsic contrast because injection of exogenous contrast agents beyond the blood–brain barrier is both challenging as well as invasive.In conclusion, there is a wide range of advanced MRI methods that potentially can narrow the gap between current neuroimaging markers of SVD and the underlying pathophysiological processes. Future work should aim to optimize these methods for specific targets related to SVD and to apply these in selected patient groups to advance the understanding of this disease.AcknowledgmentsWe thank Lennart Geurts for providing us with Figure 2.DisclosuresDr Zwanenburg is supported by the European Research Council, grant agreement no. 33733. Dr van Osch is supported by the European Union (Horizon2020, project number 634541) and the Netherlands Organisation for Scientific Research (project number 016.160.351). This study was part of the CAVIA project (Cerebral Amyloid Angiopathy: Vascular Imaging and fluid markers of Amyloid deposition, project number 733050202), which has been made possible by the Netherlands Organisation for Health Research and Development (ZonMW), part of the Dutch national Deltaplan for Dementia: zonmw.nl/dementiaresearch.FootnotesThe online-only Data S