Background White matter hyperintensities (WMH), lacunes and microbleeds are regarded as typical MRI expressions of cerebral small vessel disease (SVD) and they are highly prevalent in the elderly. However, clinical expression of MRI defined SVD is generally moderate and heterogeneous. By reviewing studies that directly correlated postmortem MRI and histopathology, this paper aimed to characterise the pathological substrates of SVD in order to create more understanding as to its heterogeneous clinical manifestation.
Summary Postmortem studies showed that WMH are also heterogeneous in terms of histopathology. Damage to the tissue ranges from slight disentanglement of the matrix to varying degrees of myelin and axonal loss. Glial cell responses include astrocytic reactions—for example, astrogliosis and clasmatodendrosis—as well as loss of oligodendrocytes and distinct microglial responses. Lipohyalinosis, arteriosclerosis, vessel wall leakage and collagen deposition in venular walls are recognised microvascular changes. Suggested pathogenetic mechanisms are ischaemia/hypoxia, hypoperfusion due to altered cerebrovascular autoregulation, blood–brain barrier leakage, inflammation, degeneration and amyloid angiopathy. Only a few postmortem MRI studies have addressed lacunes and microbleeds to date. Cortical microinfarcts and changes in the normal appearing white matter are ‘invisible’ on conventional MRI but are nevertheless expected to contribute substantially to clinical symptoms.
Conclusion Pathological substrates of WMH are heterogeneous in nature and severity, which may partly explain the weak clinicoradiological associations found in SVD. Lacunes and microbleeds have been relatively understudied and need to be further investigated. Future studies should also take into account ‘MRI invisible’ SVD features and consider the use of, for example, quantitative MRI techniques, to increase the sensitivity of MRI for these abnormalities and study their effects on clinical functioning.
Statistics from Altmetric.com
MRI and clinical expression of small vessel disease
White matter hyperintensities (WMH), lacunes and microbleeds are regarded as MRI expressions of small vessel disease (SVD) and are commonly found on brain MRI of elderly subjects. WMH are visible as hyperintense areas on T2 weighted MRI scans (including FLAIR), while lacunes are identified on MRI as small cavities with a diameter of 3 mm to 10–15 mm, and signal intensities comparable with CSF. Lacunes are located in the white matter (WM) or subcortical grey matter and often have a surrounding hyperintense halo. Microbleeds are small, round, hypointense foci on gradient echo T2* weighted MRI and are mostly located in the basal ganglia or cortical–subcortical areas.1 2 Examples of these abnormalities are given in figure 1. Unfortunately, definition and quantification of these MRI expressions of SVD vary between studies. This warrants a standardisation of SVD rating on MRI, as extrapolation of results from different studies to more general conclusions may be severely hampered otherwise.
In non-demented elderly subjects, WMH, lacunes and microbleeds have been associated with cognitive decline, including reduced mental speed and impaired executive functions.3–7 WMH have also been related to other potentially disabling symptoms, such as gait disturbances, depression and urinary incontinence.8–11 SVD is even more common in subjects with Alzheimer's disease (AD), and it might interact with the neurodegenerative changes in AD and with their effect on cognitive decline.12 13
Thus SVD probably contributes significantly to clinical disability in the elderly. As it can potentially be treated or prevented, increased insight into the underlying pathological mechanisms of SVD is of paramount importance.
The value of postmortem MRI studies
The association between SVD features on MRI and clinical symptoms is modest. An explanation for this may be heterogeneity of the neuropathological substrates underlying SVD. T2 weighted MRI dichotomises the white matter as ‘hyperintense’ (WMH) or ‘normal’ whereas the hyperintense areas may reflect pathological tissue changes that vary in type and severity. It further reveals the presence of lacunes and microbleeds but it has been suggested that expressions of SVD that are not readily detectable on conventional MRI—that is, cortical microinfarcts and tissue changes in the normal appearing white matter (NAWM)—may play an even more important clinical role in terms of clinical symptomatology (figure 2). These abnormalities can now only be revealed post mortem.14
To better understand the pathological changes involved in SVD, postmortem MRI scanning and direct correlation with pathology is a valuable tool, as it bridges the gap between MRI findings and clinical studies.15 16 As it has been shown for other neurological diseases, such as multiple sclerosis,15 17 postmortem MRI histopathology correlation studies may help to solve the weak clinical–radiological associations in SVD.
This paper aimed to investigate the published pathological substrates of WMH and other features of SVD, by comprehensively reviewing studies that have directly compared postmortem MRI and histopathology. Another aim was to pinpoint the gaps in our knowledge and provide the readership with suggestions for further studies, which will hopefully contribute to the development of future treatment options for demented and non-demented elderly patients suffering from SVD.
A small proportion of patients with SVD features on their MRI suffer from genetic disorders such as cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL) or hereditary cerebral amyloid angiopathy (CAA). These diseases have a distinct aetiology, and we will only focus on SVD that is observed in ‘normal’ ageing and AD here.
Method (search strategy)
We have systematically searched PubMed for scientific reports correlating postmortem MRI and histopathological assessment of WMH, lacunes and microbleeds until December 2009. The following search terms were used: postmortem, MRI, magnetic resonance, white matter (hyperintensities/lesion(s)), lacune(s), lacunar infarct(ion), microbleed(s).
White matter hyperintensities
Studies having correlated postmortem MRI to histopathology of WMH are summarised in table 1. These studies confirmed in vivo studies, stating that WMH are highly prevalent (94%) in elderly populations.33 The first studies are small and descriptive. However, subsequent studies have specified WMH by distinguishing periventricular (PVL) versus deep (DWMH) WMH and the extent of DWMH (box 1).
Box 1 Sensitivity and specificity of postmortem MRI
All studies identified by the abovementioned search criteria have used formalin fixed brain specimens. Fixation duration and time to autopsy influence the reliability of postmortem MRI measurements. It has been shown that tissue fixation decreases both T1 and T2 relaxation times in both gray and white matter.42 43
Early descriptive studies have claimed that postmortem MRI of 0.25 T to 1.5 T can already visualise WMH with sufficient image quality.16 18 22 However, the sensitivity of postmortem MRI does appear to be dependent on the size of WMH. Smaller punctate WMH, thought to have little clinical impact, can be less clearly visible on postmortem MRI.24 44 45 A sensitivity of 95% (range 87–99%) and specificity of 71% (range 44–90%) was found for PVL on postmortem T2 weighted postmortem MRI, which could be directly compared with myelin loss in Luxol Fast Blue stainings. For DWMH, the sensitivity was 86% (range 79–93%) and specificity was 80% (range 72–88%).31 44 Overall, postmortem MRI was considered a valuable technique for translating pathological findings to the clinical setting.
Descriptive MRI and histopathology studies
Using direct post mortem MRI and histopathology correlations, a plethora of histopathological alterations in WMH was described in several studies, including studies with AD patients, patients with cortical infarctions and patients with Binswanger's disease. WMH was shown to reflect partial loss of myelin, axons and oligodendroglial cells, astrogliosis, dilatation of perivascular spaces, activated macrophages and fibrohyalinotic vessel changes.16 18 22 23 This range of tissue changes was suggested to be collectively suggestive of incomplete infarcts. Also, complete deep white matter infarcts were found, mostly in WMH with arteriolosclerotic vessel changes.16 19 23
Distinct types of white matter hyperintensities
In clinical studies using in vivo MRI, an attempt to improve specificity for WMH was made by distinguishing between periventricular WMH (thin hyperintense line, smooth halo or irregular bands/caps) and WMH in the deep WM (punctate, early confluent and confluent WMH).46 Postmortem MRI and histopathology correlation studies have described that each type of WMH reflects distinct pathological changes.24 26–28 47
Mild periventricular WMH presents with discontinuity of the ependyma, mild–moderate gliosis in the subependymal layer, loosening of the fibre network and myelin loss around so-called ‘tortuous venules’ and dilated perivascular spaces. No arteriolosclerotic vessel changes were found in these regions.24 30 Irregular PVL was shown to correspond to more severe, partly confluent, areas with varying fibre and myelin loss and reactive gliosis. Some complete infarcts were seen in irregular PVL regions, in combination with fibrohyalinotic and arteriosclerotic vessels.
Punctate, early confluent and confluent WMH in the deep WM were found to be associated with increasing severity of tissue changes. In punctate DWMH, tissue changes were generally mild and confined to the area around dilated perivascular spaces with myelin loss and atrophic neuropil around fibrohyalinotic arterioles. In early confluent DWMH, perivascular rarefaction of myelin was accompanied by varying degrees of axonal loss and astrogliosis. In confluent DWMH, diffuse areas of incomplete parenchymal destruction were observed, together with loss of myelin, axons and oligodendrocytes, astrogliosis, spongiosis and focal transitions to complete infarcts.24 26–28 29 31 Examples of pathological samples with periventricular and deep WMH, defined on postmortem MRI, are shown in figure 3.
The above described studies imply that smooth periventricular WMH and punctate WMH are mild forms of WMH and may therefore not be clinically relevant or even detectable.29 Irregular periventricular WMH and confluent DWMH, however, correspond to more severe tissue changes, probably of ischaemic origin, and are more likely to produce clinical symptoms.19 26 27 29 Of note is the dependence on subject selection: in the relatively healthy NUN study cohort, evidence of ischaemia was not found in extensive DWMH.30
Pathogenetic mechanisms underlying white matter hyperintensities
Recently, several studies further assessed possible pathogenetic mechanisms of WMH by quantitative assessment of immunohistochemical stainings. These studies include important work on the Medical Research Council Cognitive Function and Ageing Study (MRC CFAS) cohort that prospectively collects unselected brain specimens from a large community based cohort.48
Firstly, the role of hypoxia in the pathogenesis of WMH was investigated using specific markers for vascular morphology and tissue hypoxia.34 Thicker vessel walls and larger perivascular spaces were found in WMH. In DWMH specifically, capillary endothelial cells were found to be activated and an increased expression of several hypoxia markers was observed. Other studies on characterisation of afferent vessels showed arteriolar tortuosity and decreased vessel densities in WMH.32 49 These findings support an ischaemic pathogenesis of WMH, especially in DWMH.
Secondly, blood–brain barrier dysfunction was demonstrated in a proportion of WMH.19 35 This was shown by the presence of swollen, eosinophilic, glial fibrilary acidic protein positive astrocytes in both DWMH and PVL.50 These clasmatodendritic astrocytes stained positively for serum fibrinogen implying leakage of the blood–brain barrier.51 Furthermore, vascular integrity (as determined by CD31 staining) and P—glycoprotein, an important constituent of the blood–brain barrier, were decreased in WMH.37 Concentric collagen deposition in venular walls may cause intramural thickening, stenosis and eventually venous insufficiency. Venous collagenosis then induces ischaemic stress and may cause dysfunction of the blood–brain barrier.49 50
Thirdly, the role of microglial cells in the pathogenesis of WMH was investigated.36 Microglial cells in PVL showed a greater tendency to be immunologically activated than DWMH, as shown by the expression of MHC class II. DWMH contained microglial cells with an amoeboid morphology, which were less immune activated but were likely involved in the phagocytosis of myelin breakdown products. Alternative pathogenetic mechanisms of WMH included altered cerebral blood flow autoregulation, axonal depletion from Wallerian degeneration or toxic effects of amyloid on vascular permeability in AD patients.52
These findings have illustrated that MRI visible WMH are associated with various underlying pathological features and (patho)biological responses in the MRC CFAS cohort.53 This cohort, however, consists of a heterogeneous community based group of subjects, including healthy elderly, AD patients and subjects with other neurological disorders. The large heterogeneity encountered in this group may therefore be partly artificial and WMH may be pathologically distinct for patient (and control) groups.35
White matter hyperintensities in patients with Alzheimer's disease
MRI expressions of SVD are more prevalent in patients with AD than in non-demented elderly.12 In addition to the prototypical neuropathological characteristics of AD—that is, amyloid plaques and neurofibrillary tangles—cerebrovascular pathology is also more frequently observed in AD compared with the general elderly population.48 54 The impact of cerebrovascular pathology on cognitive decline in AD patients remains to be established. For WMH, some studies have suggested a synergistic effect with common AD pathology on cognitive decline13 55 whereas others have not found a distinct role for WMH in AD.56
Although WMH was found to be more extensive in AD patients than in controls, the nature of pathological correlates, including vascular morphological changes and specific markers for hypoxia were comparable between these groups.34 57 58 An exception is microglial activation, which was specific for WMH in AD patients.41 The severity of tissue changes, however, differed with more severe loss of myelinated axons, ependyma denudation, gliosis and thicker adventitia of the deep white matter arteries in AD patients.57 58 In vascular dementia (VaD) patients, the histopathological profile of WMH was comparable with that of AD patients.59 This generally comparable pathology suggests that WMH associated with ageing, AD and VaD does not have a distinct pathogenesis but instead may be part of a pathological continuum.34
A specific pathology possibly linking SVD and AD is CAA.60 61 CAA is characterised by amyloid deposition in the smooth muscle cells of cortical, subcortical and leptomeningeal small arteries and arterioles.28 62 Patients with CAA can present with intracerebral haemorrhage, transient neurological events and cognitive decline.62 In CAA patients, the severity of CAA was found to be associated with WMH severity,60 possibly due to global vascular dysfunction, which includes the vasculature in the white matter.60 In AD, some studies have found weak correlations between WMH and CAA in AD34 63 64 whereas other studies failed to find any correlations.19 23 54 57
Other expressions of SVD
In pathological terms a ‘lacune’ corresponds to small (lacunar) infarcts, dilated perivascular spaces or old small haemorrhages.65 66 However, the term ‘lacune’ in MRI studies is generally used for a lacunar infarct. These are focal CSF filled cavities, often surrounded by a hyperintense rim on FLAIR images. Lacunes are typically located in the areas supplied by the deep thalamoperforant, lenticulostriate or pontine paramedian arterioles—that is, basal ganglia, thalamus, internal capsule, pons and centrum semiovale.67 68
The few postmortem MRI studies that have focused on lacunes are summarised in table 2. On histological examination, MRI defined lacunes were found to correspond to irregular cavitations with scattered fat laden macrophages, which can be accompanied by surrounding reactive gliosis and myelin and axonal loss.22 66 71 72 With increasing age of the lacune, the density of macrophages diminishes and gliosis becomes more fibrillar.71 A subtype of lacunes may be seen that is not yet cavitated but shows selective neuronal loss with relative preservation of glial elements.73
Several postmortem MRI studies have compared lacunes to dilated perivascular (Virchow–Robin) spaces as these structures appear similar on MRI which generally hinders a clear distinction.18 26 28 The clinical relevance of enlarged perivascular spaces, if any, is not yet fully elucidated. In general, enlarged perivascular spaces are considered to be asymptomatic but a relation with SVD may exist. A discriminating feature between lacunes and enlarged perivascular spaces may be that lacunes are commonly larger (>3 mm) and can be accompanied by perifocal signal changes.20 70 Focal cavities in the anterior perforated substance and the lower part of the basal ganglia/putamen have been reported to generally refer to perivascular spaces rather than to lacunes.69 74 75
The most frequently reported cause of lacunes is acute arteriolar occlusion by arteriosclerosis/thrombosis but the existence of non-cavitated lacunes and the relationship with WMH suggest that there may be other pathogenetic mechanisms with a more gradual development.65 73 76–80 Possible alternatives include thromboembolism, general ongoing hypoxia or tissue damage by extravasated toxic serum proteins due to blood–brain barrier leakage.76 81 Future postmortem MRI studies with histopathological confirmation is warranted to further investigate these mechanisms.
Clinical MRI studies have generally regarded small foci of signal loss on gradient echo T2* MRI sequences as microbleeds.1 82 They are not only a predictor of future lobar intracerebral haemorrhage but are also independently associated with cognitive decline.6 7 83
Only a few studies have used direct postmortem MRI pathological correlations to establish the pathological changes responsible for these MRI hypointensities (see table 3).84–86 A recent study that systematically correlated susceptibility weighted imaging, an advanced T2* MRI sequence, to tissue pathology of hypointensities in AD patients87 found that most lesions indeed seem to be microscopic bleedings. A minority of these lesions, however, corresponded to small lacunes, dissections of a vessel wall or to microaneurysms. Microbleeds may also correspond to focal accumulations of hemosiderin containing macrophages in the perivascular space and there is evidence of haeme degradation activity with a surrounding inflammatory reaction with activated microglial cells, late complement activation and apoptosis.87 Microbleeds were found to be occasionally surrounded by gliosis and incomplete ischaemic changes. The walls of ruptured arterioles may show CAA related vascular damage, with thickened, acellular morphology, lack of the muscularis layer and β amyloid deposition. CAA related microbleeds tended to be localised at the grey–white matter junction and in superficial cortical layers of the parietal and occipital lobes. Microbleeds in hypertensive subjects, however, were more often seen in the basal ganglia, brainstem and cerebellum.88 Arteriosclerosis of the vessel walls was often present in these subjects.84 89
MRI ‘invisible’ expressions of SVD
As noted above, there is accumulating evidence that there are also pathological changes associated with SVD which are ‘invisible’ to conventional MRI, such as tissue changes in white matter areas appearing normal on postmortem MRI (NAWM) and cortical microinfarcts. Pathologically, NAWM may correspond to mild tissue changes with a slightly lower myelin density, activated endothelium, a looser but still largely intact axonal network and a normal glial density 31 33. Furthermore, it has been shown that the density of small afferent vessels is not only decreased in WMH but extends into NAWM and the cortex.32
Cortical microinfarcts are microscopically small lesions. They are attributed to ischaemia, consisting of complete or incomplete cavitation with myelin pallor and neuronal loss, surrounded by glial cells and/or macrophages.90 Cystic microinfarcts tend to be larger (up to 5 mm) than non-cystic microinfarcts (0.05–0.4 mm).90 91 Several population based prospective autopsy studies suggested that cortical microinfarcts are major determinants of dementia.90–92 Microinfarcts also have an independent influence on cognitive decline in the non-demented elderly, with only little or moderate AD changes on histology.14 93 Moreover, microinfarcts were associated with CAA in patients with VaD.90
All of these findings suggest that SVD is a widespread disease and has various expressions throughout the brain of which only some aspects can be visualised with conventional MRI. As illustrated by figure 2, MRI ‘invisible’ pathologies, including cortical microinfacts and tissue changes in the NAWM, hence contribute to the clinical–radiological association that is found in SVD.
Postmortem quantitative MRI: more specific for SVD related pathology than T2 weighted MRI?
To be able to draw conclusions on the clinical relevance of SVD, additional pathology specific tools are needed in vivo. Recently, quantitative MRI techniques (QMRI) have been suggested to be more specific for underlying pathology.31 44 59 It has been shown that magnetisation transfer imaging and diffusion tensor imaging (DTI) distinguish between WMH and NAWM in elderly subjects.52 94 When correlating postmortem QMRI and pathology, it needs to be taken into account that both the time to autopsy and fixation duration have an effect on T1 and T2 relaxation times and diffusivity measures.95 96
Although several postmortem studies using QMRI have been performed in patients with multiple sclerosis and other neurological diseases,97–99 postmortem QMRI studies in the elderly with SVD are still scarce. Two small studies showed that DTI measures in WMH seem to correspond to the degree of myelin loss. Also, the area of diffusivity and pathological changes was found to be more spatially extensive than indicated by the hyperintense areas on conventional T2 weighted MRI.39 40 Pathological correlates of fractional anisotropy in DTI and T1 relaxation time were established in a recent study on WMH in AD patients and controls. Fractional anisotropy reflected axonal loss whereas T1 relaxation time corresponded with axonal loss, myelin loss and microglial activation.41
These few studies reveal that QMRI techniques may be promising in assessing tissue damage in vivo because they sensitively and specifically reflect the severity of pathological substrates and reveal tissue changes in areas that appear normal on conventional MRI (NAWM).
Conclusions and considerations for future research
This review has considered the pathological correlates of SVD, as reflected on MRI. The available literature suggests that several explanations may exist for the weak clinical–pathological associations.
Firstly, pathological substrates of SVD expressions on MRI, such as WMH, lacunes and microbleeds, are heterogeneous in nature and differ in severity. Relative to WMH, postmortem MRI pathology correlation studies of lacunes and microbleeds are still scarce. For lacunes, pathological correlation studies are certainly warranted to be able to further investigate their hypothesised multiple aetiologies, including acute thromboembolism, continuing moderate ischaemia with eventual focal tissue loss and inflammation. For microbleeds, the pathogenetic mechanisms and their relationship with WMH and lacunes also needs to be further unravelled. It should be noted that in this review, we have only focused on normal ageing and dementia. In specific brain diseases, such as CADASIL and hereditary CAA, SVD features are also present but may differ with regard to their MRI and histopathology profiles.
Secondly, until recently, clinical MRI studies have often focused on separate aspects of SVD, such as WMH, and found only weak associations with clinical symptoms. However, not only WMH but also lacunes and microbleeds are bound to contribute to clinical symptoms such as cognitive decline.3 6 Furthermore, the previously discussed MRI ‘invisible’ lesions—that is, microinfarcts and tissue changes in the NAWM—may be clinically relevant, independently of MRI ‘visible’ characteristics of SVD.93 The combination of SVD features is therefore a better predictor of cognitive decline than separate SVD expressions.100 Moreover, SVD should be assessed together with frequently coexisting large vessel infarcts, which may improve insight into the mechanisms of vascular cognitive impairment. In addition to vascular disease, other (degenerative) brain changes—for example, AD pathology, CAA or cortical Lewy bodies—may interact and modulate their specific contributions to cognitive decline.92 101–104 Future studies should therefore consider the full spectrum of SVD expression, together with vascular and degenerative pathologies coexisting in the ageing brain.
Unfortunately, MRI sequences that are commonly used in radiological practice are insufficiently sensitive and specific to detect all the tissue changes related to SVD. Novel QMRI techniques, such as DTI, magnetisation transfer imaging, and T1 and T2 relaxation time measurements have two advantages over conventional MRI. Firstly, QMRI better reflects the severity of underlying pathological substrates. Secondly, it may better reveal clinically relevant tissue changes in the WM that appears normal on conventional MRI.39 40 QMRI techniques are therefore more adequate at reflecting the full range of SVD expressions. Several QMRI abnormalities were confirmed histopathologically, and these techniques are promising, pathology specific tools for future in vivo studies. It should be noted that, even though study logistics will become more complex, the use of fresh brain tissue in exploratory postmortem MRI studies is preferred above the use of fixed specimens, as formalin fixation has been shown to influence tissue proton relaxation characteristics. Unlike tissue changes in the NAWM, QMRI techniques have not yet been investigated to visualise cortical microinfarcts. The future use of high resolution MRI, using 7 T MRI scanners, may achieve in vivo assessment of these lesions, which are probably beyond the resolution of 1.5 or 3 T MRI.
We would like to acknowledge Miriam Cornella, MSc, for her help with the literature search.
Competing interests None.
Provenance and peer review Commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.