The term cerebral microbleed (CMB) refers to small, round dark-signal lesions detected by T2^*-weighted or gradient-echo (GRE) magnetic resonance imaging (MRI).1 CMBs were introduced to stroke physicians in the late 1990s and early 2000s after development of MRI techniques sensitive to paramagnetic effects.2 The clinical significance of CMBs has been actively investigated, especially in the stroke field and more recently in studies on cognitive impairment and aging.3 Histological investigation has shown that CMBs are tiny foci containing hemosiderin-laden macrophages and abnormal microvessels showing fibrohyalinosis.4,5 Clinical cases with frank symptoms caused by CMBs are uncommon. Because CMBs are manifestations of focal extravascular leakage of blood components, however, investigators have suggested that accumulation of CMBs reflects a bleeding-prone status in individuals with an elevated risk of cerebral hemorrhage. Clinical studies have found strong associations between CMBs and chronic hypertension and low cholesterol levels,6,7 and between the proximity and volume of CMBs and those of subsequent intracerebral hemorrhage (ICH).8,9 Longitudinal studies have found that CMBs are linked to subsequent hemorrhagic stroke in stroke survivors,10 and suggested that CMBs are related to antithrombotic-related hemorrhage.11,12 In this review, we discuss fundamental findings of CMBs, and the clinical implications of these observations for the field of cerebrovascular disease.

GRE sequences are more sensitive to susceptibility effects than are classical MRI sequences. Unlike classical T2-weighted imaging or echo-planar sequences, GRE sequences maximize the paramagnetic effects of blood components such as hemosiderin, deoxyhemoglobin, and ferritin.2,13 Due to the dephasing of MRI signals, GRE sequences tend to exaggerate lesion sizes. Therefore, sub-millimeter CMBs appear as signal-loss lesions of several millimeters, a phenomenon referred to as the blooming effect.

Susceptibility-weighted imaging (SWI) is a MRI sequence that maximizes sensitivity to magnetic susceptibility effects.14,15 SWI requires more time than does GRE, and also requires post-processing, but because SWI accentuates the magnetic properties of tissues, it enables visualization of areas such as CMBs containing deoxygenated blood substances.16 SWI permits visualization of a greater number of CMBs than can be seen with conventional GRE sequences, but the clinical implications of this increased sensitivity are not yet fully understood (Figure 1).17

Since CMBs are primarily a radiologic concept, their sizes and shapes are dependent upon the parameters of the GRE sequence. Visualization of CMBs is thought to be influenced by pulse sequence, sequence parameters, spatial resolution, magnetic field strength, and post-processing of images.1 Thus, lesions due to old parenchymal hemorrhage, cortical or deep-seated mineralizations, or vascular malformations potentially all can be misinterpreted as CMBs. Small cortical vessels, partial volume effects of the cerebellar cortex, and cavernous hemangiomas also should be considered (Figure 2).21 Moreover, multifocal bleeding spots caused by diffuse axonal injury are virtually impossible to distinguish from cortico-subcortical CMBs.

Comparison of magnetic field strength in detection of CMBs revealed 0.5 more microbleeds on average could be observed using a 3.0 Tesla MRI compared to a 1.5 Tesla unit (2.1 in 3.0 T and 1.6 in 1.5 T).18 No universal standard MR parameters exist for the detection of CMBs, and in addition, the definition of CMB size also varies. In published reports, the lower limit of CMB size is usually ≤2 mm, and the upper limit is usually between 5 and 10 mm.19 Greenberg et al. reported that the size distribution of microbleeds and macrobleeds is bimodal, and proposed that the most appropriate cut-off point between them is 5.7 mm.20 Those researchers suggested elsewhere that an upper size limitation is impractical, and that instead, a more desirable approach is to carefully exclude CMB-mimicking lesions such as those described above.1 Other investigators have proposed a specified rating scale for CMBs, the Microbleed Anatomical Rating Scale.21 In our view, setting an appropriate upper-size limitation, and determination of intra- and inter-rater agreement values, are necessary for clinical and radiologic research studies.

Pathologic-radiologic correlation studies have revealed that CMBs are focal accumulations of hemosiderin adjacent to abnormal blood vessels demonstrating fibrolipohyalinosis or amyloid microangiopathy.4,5 CMBs usually develop adjacent to capillaries, small arteries, or arterioles; hemosiderin-laden macrophages are usually present.22 A recent pathological investigation of CMBs in patients with cerebral amyloid angiopathy demonstrated extravasation of blood and hemosiderin through vulnerable vascular walls, with β-amyloid pigmentation and surrounding inflammation.23 These findings suggest that CMBs are the radiologic correlates of extravasation of blood components through injured or fragile vascular walls, or of frank small hemorrhage spots. In the context of cerebral amyloid angiopathy, Greenberg et al. suggested that increased vascular wall thickness was indicative of microbleeds.20

Because blooming artifact on GRE images exaggerates the size of CMBs, it has been proposed that the actual size of hemosiderin deposition is too small to cause apparent neurological deficit.22 Selected cases, however, have suggested that CMBs can cause neurological symptoms and signs.24,25

Further understanding of the characteristics of CMBs has generated interest in using detection of CMBs to enable hemorrhagic stroke risk stratification. A number of reports have described the hemorrhagic tendency of CMBs.24,63,64 A hemorrhagic transformation after multiple embolic infarctions occurred only in the site of the known CMB.63 A Hong Kong study followed 121 acute ischemic stroke patients, and observed that stroke survivors with CMBs on their initial MRI scans had a higher risk of subsequent hemorrhagic stroke.10 Hemorrhage counts in initial scans also were found to be pro portional to the elevated risk of a future hemorrhagic stroke.54 The increased risk of hemorrhagic stroke conferred by the presence of CMBs was also confirmed in a prospective study of 112 ICH survivors.65 Furthermore, association between CMBs and larger ICH volume has been suggested by two studies,9,66 and the predictive value of CMBs in ICH in patients with advanced white matter lesions also has been documented.67 The spot sign, an enhancing locus of contrast extravasation in a cerebral hematoma, suggesting ongoing bleeding, was reported to be negatively associated with the number of microbleeds; this result is not consistent with previous findings, and bears further investigation.68 Patients with CMBs also have been reported to be 2.8 times more likely to have a subsequent disabling or fatal stroke.69 A systematic review published in 2013 concluded that the presence of CMBs in patients with ischemic stroke was associated with greatly increased odds of a subsequent hemorrhagic stroke, but was only modestly linked to recurrence of ischemic stroke.56 This meta-analysis also noted that the strength of association between CMBs and subsequent ICH risk was modified by ethnic background, with a greater odds ratio in Asian cohorts than in Western cohorts (Figure 5).

The increased likelihood of cerebral hemorrhage associated with the presence of CMBs may allow prediction of hemorrhagic transformation after ischemic stroke. An earlier report suggested that hemorrhagic transformation after thrombolysis was associated with the presence of CMBs.70 In a case series of 100 acute ischemic stroke patients with imaging follow-up, the presence of CMBs was indicative of early hemorrhagic transformation.71 Embolic ischemic strokes occurring at the sites of previous CMBs were noted to become hemorrhagic.63 Contrary to these positive associations, however, a retrospective study of 279 acute ischemic stroke patients reported no association between CMB count and subsequent hemorrhagic transformation,72 and in an analysis of 70 stroke patients on thrombolytic treatment, CMBs failed to predict post-thrombolytic hemorrhagic transformation.73 No relationship between CMBs and prediction of hemorrhagic transformation was observed in a group of 1,034 acute ischemic stroke patients recruited in a single hospital.74 Finally, a pooled analysis of 570 acute ischemic stroke patients from 13 centers in Europe, North America, and Asia reported that symptomatic ICH after thrombolytic treatment developed regardless of the initial CMB frequency or extent.75 These studies suggest that for patients in need of thrombolysis, any increased risk of ICH attributable to CMBs is negligible, and unlikely to exceed the benefits from thrombolytic therapy. A subsequent meta-analysis indicated that the published analyses were vulnerable to publication bias and limited power, and identified a trend toward increased odds of symptomatic hemorrhage after thrombolysis (odds ratio, 1.98; 95% confidence interval 0.90-4.35).76 At present, identification of CMBs on baseline GRE images should not be considered a contraindication to thrombolysis treatment, but further investigation is warranted on this important issue.

Different findings regarding CMBs and the development of ICH or hemorrhagic transformation may be explained by the different pathological mechanisms of the two phenomena. Essentially, ICH involves rupture of a fragile microvascular wall affected by lipohyalinosis or microaneurysms under the chronic influence of hypertension.77 As CMBs have histological characteristics similar to those of vasculopathy, a correspondingly similar mechanism may underlie formation of CMBs and ICHs.22 In contrast, hemorrhagic transformation develops after acute lethal injury in relatively healthy microvasculature.

Considerable interest also exists in utilizing detection of CMBs to estimate the risks of hemorrhagic complications in patients on antithrombotic treatment. Two patients on warfarin were reported to have developed lobar hemorrhages right at the location of CMBs.64 I CMBs were found to be more frequent and extensive in patients with aspirin-associated ICH.11,78 In a cross-sectional study, CMBs were more common in patients taking antithrombotic agents, and aspirin use was found to be related to a lobar location.79 Results from our group demonstrated that patients with anticoagulation-associated hemorrhagic stroke complications are 3.6 times more likely to have CMBs than are age- and sex-matched controls.12 A recent pooled analysis involving 1,460 hemorrhagic strokes and 3,817 ischemic strokes concluded that the number of CMBs was greater in warfarin users who developed ICH.80 Given the strong association between CMBs and subsequent ICHs in stroke survivors, a prospective study is needed to assess the predictive power of CMBs in stroke patients on antithrombotic treatment.

CMBs were first identified as tiny, round dark-signal lesions on GRE MRI, and are frequently detected in patients with ischemic or hemorrhagic strokes. Pathological analysis demonstrated that CMBs are extravasations of blood components through fragile microvascular walls, and therefore reflect a bleeding-prone vasculopathy in brain. Several clinical studies have concluded that CMBs are associated with hemorrhagic stroke and hemorrhagic complications following antithrombotic medications. The currently available data do not support the exclusion of thrombolytic treatment based solely on CMB presence or extent. Prospective studies are warranted to confirm the clinical implications of CMBs, and to establish their use for predictive models of hemorrhagic stroke in various situations.