Brain microbleeds: what are they?

Brain microbleeds are primarily a radiological construct (ie, describing small MRI signal voids) but one that is meant to reflect a specific underlying microscopic pathology: perivascular collections of haemosiderin deposits (figure 1). The first radiological–histological comparison was performed by Fazekas et al in 1999.1 Only a few subsequent histological studies are available in the literature which hugely limits our understanding of brain microbleeds.

• Download figure
• Open in new tab
• Download powerpoint

Histology of a chronic cortical microbleed, haematoxylin–eosin stain. The lesion consists of central haemosiderin accumulation (brown pigment marked by an asterisk) surrounded by ring-shaped astrocytic and microglial proliferation (arrows) with active haemosiderin resorption. Scale bar=100 μm. Courtesy of Dr V Deramecourt.

What exactly do they look like?

A microbleed appears as a round or ovoid black signal on gradient echo T2* MRI, it displays blooming (ie, it is larger or more conspicuous on gradient echo than on spin echo MRI), it is devoid of T2 weighted hyperintensity (such as a cavernous malformation or haemorrhagic infarct) and it is surrounded by brain parenchyma (figures 2 and 3). Although their exact size cannot be extrapolated from the MR appearance because of the blooming effect, the upper size limit is usually taken as between 5 and 10 mm. Several questions remain unaddressed, in particular the operational rules (based on the history and MR appearances) needed to differentiate microbleeds from macrobleeds, as well as from haemorrhagic and non-haemorrhagic mimics such as calcification of the basal ganglia, diffuse axonal injury and small perforating arteries (figure 4). To improve intraobserver and interobserver agreement, morphological criteria2 and rating scales such as the Brain Observer MicroBleed Scale (BOMBS) (box 1)3 have been proposed.

• Download figure
• Open in new tab
• Download powerpoint

MRI, axial plane, T2* gradient echo sequence showing multiple lobar brain microbleeds as small black dots, without any lesions in the basal ganglia.

• Download figure
• Open in new tab
• Download powerpoint

MRI, axial plane, T2* gradient echo sequence showing multiple brain microbleeds predominantly in the basal ganglia region.

• Download figure
• Open in new tab
• Download powerpoint

Some mimics of brain microbleeds. (A) Axial T2* gradient echo sequence. Flow void mimicking a microbleed (arrow A1), enlarged on the right (circle A2). Note the vessel leading up to the black dot indicating that it is a vessel flow void in cross section. (B) Axial T2* gradient echo sequence. Partial volume artefact from left mastoid bone (arrow).

‘From small to big’ is an all too easy way of regarding microbleeds—in other words, as a single continuum. However, it seems that the volume of haemorrhages has a bimodal distribution.4 Interestingly, the cut-off point that best divides the microbleed and macrobleed peaks corresponds to a diameter of 5.1 mm, similar to the 5–10 mm maximum conventionally used. This suggests that microbleeds and macrobleeds represent two distinct entities, perhaps with different pathophysiology. However, these measurements were only performed in patients with presumed cerebral amyloid angiopathy and we cannot extrapolate these results to deep microbleeds and ICH due to small vessel disease.

How to search for brain microbleeds?

The MR imaging parameters of greatest influence on microbleed detection are pulse sequence, sequence parameters (echo time and flip angle), spatial resolution, magnetic field strength and postprocessing methods. There is huge variation in these parameters in the published studies.5 In most centres, gradient echo T2* sequences are used with 1.5 T magnet strength. However, increasing magnet strength increases the detectability of microbleeds,6 as does three-dimensional susceptibility weighted imaging.7 These variations in technique of course affect the results of different studies, the prevalence and numbers of microbleeds varying widely. To complicate matters, this field is evolving so quickly that it is impossible to recommend one technique over another. At some point, if indeed we are not already there (eg, with 7 T machines), it will be necessary, although difficult, to distinguish normal ageing processes from disease.

Who gets brain microbleeds?

Using conventional MR sequences, the prevalence of microbleeds is about 5% in healthy people but increases with age, and is about 34% in ischaemic and 60% in haemorrhagic stroke patients.5 The prevalence is less in first ever compared with recurrent ischaemic and haemorrhagic stroke patients, suggesting that microbleeds are a marker for the evolution or severity of the underlying vascular pathology. In Alzheimer's disease, one in five patients has microbleeds.8 In a cohort of people with cerebrovascular disorders, the risk factors associated with brain microbleeds were arterial hypertension, increasing age, leukoaraiosis and lacunes.5

Do they predict future intracerebral hemorrhage?

Some clinicians have raised obvious concerns about the safety of antithrombotic treatments in patients with microbleeds—are they at higher than average risk of future ICH? But at present there is no clear evidence that they are. Nevertheless, this may be a false negative conclusion as the numbers of recurrent strokes in the existing prospective studies9,–,13 is small: only 64 among 1212 patients have been reported to date during follow-up from 16 to 27 months. Moreover, there was insufficient power to stratify outcome on the basis of different antithrombotic drug treatments.

Should i thrombolyse a patient with brain microbleeds (if i know about them)?

Some case reports illustrate haemorrhagic transformation after intravenous thrombolysis at the site of a brain microbleed.14 But of course case reports may be misleading, they tell us that something can happen, not how often it happens. The BRASIL study gathered 570 acute stroke patients treated with intravenous thrombolysis; symptomatic ICH occurred in 3% of patients without versus 6% of patients with microbleeds (p=0.17).15 Even though this study was not big enough to detect a significant increase in bleeding (if one really exists), this risk would still not outweigh the benefit of intravenous thrombolysis shown in the trials where CT rather than MR was used (and so microbleeds were not even known about). However, two questions remain unanswered: do lobar microbleeds, revealing presumed cerebral amyloid angiopathy, have an increased risk of bleeding compared with deep microbleeds revealing presumed small vessel disease? And are patients with numerous microbleeds at increased risk? At the moment, these concerns are rather hypothetical as acute stroke patients seldom have an urgent MR but rather a CT scan.

Should i give warfarin to a patient with brain microbleeds?

Among patients with ICH, 14% occur in patients treated with warfarin.16 Given the rising prevalence of atrial fibrillation and the greater use of warfarin, oral anticoagulant associated ICH incidence is expected to rise. Therefore, any means of better identifying patients at particular risk is vital, and of course it is possible that microbleeds are one risk factor for anticoagulant associated ICH. We already know they are strongly associated with leukoaraiosis which itself is a risk factor.17 Moreover, microbleeds in a lobar distribution may be a marker for cerebral amyloid angiopathy, an increasingly recognised cause of anticoagulant associated ICH.2

Unfortunately, previous reports have been too limited by small numbers of cases and confounding associations with other risk factors for ICH.18 19 Therefore, until further data are available, warfarin should not be withdrawn in patients with microbleeds who have a clear indication for anticoagulation such as atrial fibrillation after ischaemic stroke or transient ischaemic attacks. Whether new oral anticoagulants such as dabigatran have the same risk in the setting of brain microbleeds remains to be seen.

Should i give aspirin to a patient with brain microbleeds?

Some studies have provided data about the use of antiplatelet drugs at the time of brain MRI examination. However, they consisted of cohorts of patients with cerebrovascular disease which may have acted as a confounder: microbleeds may have been due to the ongoing vascular disease and not to the treatment.5 In a population based sample of 1062 people, 60 years and older, free of dementia but some with vascular risk factors or previous history of stroke, brain microbleeds were more prevalent among users of antiplatelet drugs (odds ratio 1.71).20 However, data from such a cross sectional study do not necessarily prove that microbleed patients are at increased risk of bleeding when treated with antiplatelet drugs. Indeed, the presence of microbleeds on MRI does not even tell us when these bleeds actually occurred because haemosiderin deposits can remain visible in the brain for an undefined period (it is conceivable that they were there before the antiplatelet drugs were given).

On balance, antiplatelet treatment should not be contraindicated for secondary stroke prevention where a clear overall benefit has been established from the randomised trials and meta-analyses (even though these were almost all conducted before microbleeds could be diagnosed on MRI). For primary prevention, no data are available and it is possible that in patients with multiple brain microbleeds, the benefit/risk ratio is even less favourable than it is overall.

Are brain microbleeds useful in the diagnosis of cerebral amyloid angiopathy?

To date, microbleeds have not been validated in pathological studies as part of the diagnostic criteria for probable cerebral amyloid angiopathy. Indeed, few direct histological data are available: in the 11 brains reported by Fazekas et al, two had features of cerebral amyloid angiopathy and nine of advanced hypertensive lipohyalinotic changes.1

• Download figure
• Open in new tab
• Download powerpoint

Brain Observer MicroBleed Scale (BOMBS)

The presence of multiple strictly lobar ICHs in an older patient without other definite cause of haemorrhage, defined as ‘probable cerebral amyloid angiopathy associated ICH’ by the Boston criteria strongly correlates with pathologically advanced cerebral amyloid angiopathy.21 However, microbleeds were not taken into account: 24 out of the 39 patients were investigated with either CT (n=17) or MRI without GRE-T2* sequences (n=7).

Whether a lobar distribution of microbleeds can be used to make the diagnosis in the absence of macrobleeds remains to be determined but this idea has received indirect support from the correlation between isolated lobar microbleeds and the APOE ε4 allele in the population based Rotterdam Scan Study.22 Given the relationship between APOE ε4 and cerebral amyloid angiopathy, these results raise the interesting possibility that isolated lobar microbleeds might reflect the presence of advanced cerebral angiopathy. But these findings have to be confirmed in other populations.

Brain microbleeds in alzheimer's disease

Patients with Alzheimer's disease often have a history of arterial hypertension, and cerebral amyloid angiopathy is frequent at autopsy; both are associated with brain microbleeds. Indeed, microbleeds are frequent in Alzheimer's disease (one in five patients with the disease).8 Their anatomical distribution is very close to the distribution of cerebral amyloid angiopathy associated ICH; they tend to be lobar with posterior predominance.23 Therefore, it is conceivable that brain microbleeds are a key factor in the pathogenesis of Alzheimer's disease, connecting the main pathological contributors of amyloid accumulation with cerebrovascular damage.

Brain microbleeds may also predict a more aggressive form of the disease, being associated with a more than twofold increased mortality.24 Recently, patients with Alzheimer's disease with multiple microbleeds were compared with those without any.25 The microbleed group had more severe white matter hyperintensities (without any difference in degree of atrophy) and more severe cognitive impairment than could be accounted for by disease duration, degree of atrophy or white matter hyperintensities. Also, they had a lower CSF level of amyloid-β 1–42 suggesting a direct link between microbleeds and amyloid-β, one of the key proteins involved in Alzheimer's disease.

However, there are many unanswered questions: do the microbleed group behave differently from a cognitive point of view? Would they benefit from different treatment strategies? Curiously, patients with Alzheimer's disease with microbleeds are excluded from nearly all therapeutic trials.