The MRI appearance of hemorrhage is strongly dependent on the type of pulse sequence used (T1- or T2-weighted, conventional or fast spin echo, inversion recovery or gradient echo), the field strength of the magnet, as well as the magnetic state of blood products, which is dependent on the age of the hemorrhage (Gomori et al. 1985) (Fig. 23.3). In the hyperacute hematoma (encountered within minutes to hours), oxyhemoglobin is isointense on T1-weighted images and slightly hyperintense on T2-weighted images, but the more commonly encountered acute hematoma (hours to days) is comprised of intracellular deoxyhemoglobin and methemoglobin within intact red blood cells, appearing hypointense on T2-weighted images. There is usually a transient rim of edema surrounding the hematoma.

Lysed red blood cells in the subacute (days to weeks) hematoma cause increased signal on T1-weighted images from extracellular methemoglobin, typically starting from the outer rim and extending inward over time. In the chronic stage (weeks to months), phagocytosis results in accumulation of particulate iron in the form of hemosiderin and ferritin, visible as low signal areas on MRI. These breakdown products begin by forming a hypointense edge around the resolving hematoma and often persist for many years. In general, fast spin-echo imaging is slightly less sensitive in the detection of hemorrhage compared to conventional spin-echo images.

FLAIR is a strongly T2-weighted pulse sequence, and hence, the signal characteristics of intracerebral hematomas are similar to spin-echo T2-weighted images. FLAIR is very sensitive to hyperintense SAH, especially in the posterior fossa, base of skull, and the ventricles (Noguchi et al. 1995) (Fig. 23.4). CSF pulsation artifacts are a significant pitfall.

Hematomas appear hypointense on T2*-weighted gradient recalled echo (GRE) images, which are extremely sensitive to the effects of static magnetic field inhomogeneities produced by paramagnetic blood products. Hence, even small focal areas of microbleeding (MB) can be detected on GRE images: the so-called blooming effect due to the hypointense area being larger than the actual hemosiderin deposit (see Sect. 23.2.3.3) (Greenberg et al. 2009). GRE is also very sensitive to small amounts of hypointense SAH, especially in the ventricles or as superficial siderosis on the surfaces of the gyrus at the cerebral vertex. The higher oxygen partial pressure within the CSF delays conversion of blood products in a variable and less predictable way compared with hematoma in the brain parenchyma (Grossman et al. 1986). Although detection of acute SAH and hemorrhagic stroke is possible with MRI, CT is still the diagnostic procedure of choice (Irimia et al. 2011).

Hypertension is the presumed cause of the majority of intracranial hemorrhages, typically those deep in the basal ganglia, thalamus, pons, or the dentate nucleus of the cerebellum. Deep hemorrhages in hypertensive patients are often due to hypertension, whereas lobar hemorrhages in non-hypertensive elderly patients are often due to cerebral amyloid angiopathy (CAA) (Broderick et al. 2007). Although angiography is indicated in most young patients with suspicious features of calcification, abnormal vessels or blood in unusual locations, the yield declines in elderly hypertensives with deep hematoma, and it is rarely performed unless clinical suspicion is high (Broderick et al. 2007).

CAA is an important cause of nontraumatic hemorrhagic stroke in the elderly population, responsible for 15–20 % of intracranial hemorrhage in normotensive patients, such frequency increasing with age. CAA, which is associated with accumulation of the misfolded amyloid beta-protein (similar to senile plaques in Alzheimer’s disease) in vessel walls (Vinters 1987), typically causes multifocal lobar (especially frontal lobes) hemorrhages, MBs, superficial siderosis, and very rarely focal amyloidoma (Vonsattel et al. 1991). The hematoma caused by CAA is typically located peripherally in the cerebral lobes and less commonly involves the brainstem, cerebellum, or the basal ganglia (which are characteristic of hypertensive hemorrhage). CAA-associated bleeding from leptomeningeal vessels can also sometimes be detected as superficial siderosis, which is hemosiderin deposition in the subarachnoid space, visible as gyral hypointensity on GRE or susceptibility-weighted MRI. A combination of multifocal MBs, lobar hemorrhages of different ages, associated with superficial siderosis, is typical of CAA (see Fig. 24.11 in Chap. 24).

Microbleeds are defined as small “dot-like” round or ovoid (but not linear) hypointense brain lesions on GRE or susceptibility-weighted images, ­sometimes not visualized on other pulse sequences (Greenberg et al. 2009). The prevalence of these MBs is generally higher with age, in patients of Asian ethnicity, and with intracranial hemorrhage or ischemic stroke compared to populations without cerebrovascular disease (Koennecke 2006). Two common associations of MBs in the elderly include CAA and hypertension, but other causes of multiple small collections of hemosiderin should also be considered in the differential diagnosis, including traumatic injury, cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy, and multiple cavernoma (Hegde et al. 2012). CAA-related MBs tend to be lobar and are typically located in the cortical gray matter or in the border zone between gray matter and white matter.

The clinical implications of subclinical MBs are being actively investigated for both hemorrhage risk stratification and dementia. CAA may be associated with a higher rate of recurrent hemorrhage approaching 10 % per year (O’Donnell et al. 2000; Greenberg et al. 1993). Although lobar MBs are strongly linked to dementia in patients with Alzheimer disease, both types of MBs frequently coexist on MRI (see Chap. 24).

Spontaneous intracranial hemorrhage is a rare (around 0.3 % incidence) but deadly complication of thrombolytic therapy for acute myocardial infarction with a poor prognosis (up to 50 % mortality) (Vinters 1987). Age, hypertension, prior stroke, amyloid angiopathy, leukoaraiosis, and dementia may predict a higher risk of hemorrhage in patients treated with thrombolysis or warfarin after myocardial infarction (Vonsattel et al. 1991). Thrombocytopenia, hemophilia, leukemia, liver disease, fibrinolysis, disseminated intravascular coagulation associated with sepsis, and other rare diseases also predispose to intracranial hemorrhage (Wijdicks et al. 1994).

Vascular malformations include AVM and occult vascular malformations: those not demonstrated by angiography include cavernous angiomas, capillary telangiectasias, and venous malformations (Rigamonti et al. 1987). AVMs can cause hemorrhage in the cerebral hemispheres, and the nidus, dilated feeding arteries, and draining veins may be seen; signal loss on MRI in the nidus is due to a combination of hemorrhage, rapidly flowing blood, or calcification (Anonymous 1979). They are seldom seen in the elderly, being typically diagnosed before the age 40.

Cavernous angiomas are low-flow vascular malformations without enlarged feeding arteries or draining veins but are often incidentally detected on MRI (Robinson et al. 1993). On CT, they are nonspecific with focal hyperdensity containing blood or calcium with indistinct margins, but on MRI, they have a characteristic well-­circumscribed, lobulated “popcorn” appearance with heterogeneous signal intensity on both T1- and T2-weighted sequences and peripheral ring of hemosiderin hypointensity representing recurrent hemorrhage in various stages of evolution, (Hegde et al. 2012) (Fig. 23.5). Multiple, small cavernous angiomas may be seen in the familial form of cavernoma (typically become symptomatic between the third and fifth decades), which may be difficult to distinguish from CAA and other causes of developmental venous angiomas on MRI alone (Brunereau et al. 2000).

Developmental venous angiomas are clusters of deep medullary veins draining centripetally into a normal but prominent vein that does not cause intracranial hemorrhage. Developmental venous angiomas are incidental findings seen on contrast-enhanced MRI and do not require angiographic investigation nor surgical resection; however, these angiomas are associated with cavernous malformations and improve diagnostic confidence of the latter (Abe et al. 1998).

Bleeding into preexisting brain tumors accounts for 10 % of all brain hemorrhages, but only 1–2 % of primary (glioblastoma multiforme, oligodendroglioma, and ependymoma) or metastatic (melanoma, choriocarcinoma, bronchogenic, renal cell, and thyroid carcinomas) tumors present as intracranial hemorrhage (Atlas et al. 1987). Imaging red flags suggesting neoplasia include multifocal, heterogeneous stages of signal evolution present simultaneously, high signal intensity on T1-weighted images originating centrally within the hematoma, absent or incomplete hemosiderin rim, persistent and extensive perilesional edema, and contrast enhancement (see Chap. 22). In some cases, the hematoma may obliterate evidence of the neoplasm. Hypertensive hemorrhage is almost never multiple, and multiple hematomas are more likely due to bleeding diathesis, metastatic tumors, sepsis, CAA, cerebral venous thrombosis (see Sect. 23.7), or trauma.

Decreased attenuation from edema secondary to cerebral infarction is the most important sign in cerebral ischemia, and is typically evident by 24 h, although early signs may be detected as early as a few hours after ictus. However, CT has low sensitivity to cerebral infarction, especially small ones, and is also hampered by beam-hardening artifacts in the posterior fossa. Nevertheless, supporting signs such as loss of the gray matter-white matter differentiation (particularly the edge of the lentiform nucleus), insular ribbon sign, mass effect (effacement of sulci and ventricles), and the dense vessel (classically the MCA) signs might be helpful for detection (Fig. 23.6) (Schuknecht et al. 1990). MRI has a higher sensitivity than conventional CT for the detection of small lesions, infarction within the first hours of stroke onset, lesions in the posterior fossa, and documentation of vessel occlusion and brain edema (Irimia et al. 2011).

Cerebral infarction caused by embolism, hypertensive patients, and anticoagulant or thrombolytic therapy appears to have a higher risk of hemorrhagic transformation. After intra-arterial thrombolysis, independent predictors include a high National Institutes of Health Stroke Scale score, longer time to recanalization, lower platelet count, and higher glucose level (Kidwell et al. 2002). Imaging features that may predict increased risk include infarcts greater than one third of the MCA territory on CT (von Kummer et al. 1997), larger volume on DWI, early parenchymal enhancement (Vo et al. 2003), very low cerebral blood volume, and permeability measurements on perfusion studies (Schaefer et al. 2003).

MR angiography can be carried out with or without contrast injection, and hence, time-of-flight (TOF) and phase-contrast methods may be helpful for elderly patients with renal failure who might be at risk for nephrogenic systemic fibrosis (Kuo et al. 2007). In TOF-MR angiography, data are acquired in a series of overlapping thin sections suppressing stationary tissue signal and processed with a maximum intensity projection (MIP) algorithm, in which the brightest pixels represent inflow enhancement of flowing blood. Phase-contrast MR angiography is sensitive to velocity-dependent phase shifts subtracting stationary background tissue and uses velocity encoding to select for fast, intermediate, or slow flow. In contrast-enhanced MR angiography, a concentrated bolus of gadolinium contrast media is injected through an arm vein, and T1 shortening within the artery combined with techniques, such as bolus tracking or continuous fluoroscopic time-resolved techniques, closely approximates the appearance of the classic x-ray DSA standard. Three-dimensional TOF-MR angiography studies (which are less susceptible to flow-related artifacts than two-dimensional TOF due to smaller voxel size) may require 5–10 min, compared to the 10–30 s needed for contrast-enhanced MR angiography; the latter has the advantage of reduced patient movement, fewer artifacts, and improved signal-to-noise ratio (Nederkoorn et al. 2003).

Unenhanced TOF-MR angiography techniques are suitable for detecting intracranial stenoses and occlusions (Fig. 23.14a), and contrast-enhanced MR angiography, which does not overestimate the degree of stenosis, is ideal for imaging extracranial vessels (Bash et al. 2005) (Fig. 23.14c). Craniocervical arterial dissections can be detected with contrast-enhanced MR angiography plus a fat-saturated unenhanced T1-weighted MRI to depict a hyperintense subacute mural hematoma within a few days after ictus (Gelal et al. 2004). Vascular irregularity, narrowing, flow gaps, and cutoff may be seen in steno-occlusive cerebrovascular disease, with varying reconstitution of vascular signal downstream. However, not all patients will suffer complete infarction of the brain territory supplied by the occluded artery, due to the protective effect of collateral perfusion and other factors (Roberts et al. 1993). Collateral circulation is represented by alternative routes, most crucially the large caliber artery-to-artery anastomoses at the circle of Willis, but also including leptomeningeal ­anastomosis and those from the external carotid artery to ICA branches (Fig. 23.14, see also Sect. 23.6.1.3). The dynamic variability of the collateral status in each individual results in different morphologic and clinical outcomes even in patients with the same site of occlusion.

For conventional CT stroke imaging, intravenous iodinated contrast agent is unnecessary, but once hemorrhage is excluded, multimodal evaluation of acute ischemic stroke is possible using ­contrast-enhanced CT angiography and CT perfusion (see Sect. 23.9.2). CT angiography can demonstrate stenosis and occlusion from atherothrombotic or embolic disease. As post-processing software improves and the number of detectors increases in multidetector CT scanners, better spatial resolution and throughput speed result in CT angiography images superior to MR angiography and rapidly approach that of DSA (Bash et al. 2005) (Fig. 23.15).

Unfortunately, CT angiography is still hampered by being a static image of vascular anatomy and has a false-positive rate for occlusion when heavy atheromatous calcifications are present (Walker et al. 2002). On the other hand, assessment of the morphological features and density of atherosclerotic plaque may be helpful to determine increased stroke risk (U-King-Im et al. 2009). CT angiography source images have also been explored as an alternative marker for infarct core and leptomeningeal collateral vessels (Wintermark et al. 2008).

Doppler ultrasonography and transcranial Doppler are alternative, noninvasive, nonionizing methods of vascular imaging without the benefit of anatomical information about the brain parenchyma. Duplex sonography combines pulse-wave Doppler spectrum analysis and B-mode sonography; these methods can identify increased velocity of blood flow across a stenotic lesion (increased peak systolic velocity and end-diastolic velocity or increased ratios of ICA to common carotid artery peak systolic velocity) and display morphological features of the plaque and arterial wall (Moneta et al. 1993) (Fig. 23.16). Power Doppler imaging color-codes blood flow according to the amplitude of the Doppler signal. However, although duplex sonography is safe, inexpensive, and a useful screening tool, it overestimates the severity of stenosis and should not be used as the sole method for a definitive diagnosis (Latchaw et al. 2009). Transcranial Doppler can detect intracranial flow velocities, the direction of collateral flow, vessel occlusion, the presence of emboli, vascular reactivity, and cerebral vasospasm after SAH in the MCA, ACA, carotid siphon, vertebral artery, basilar artery, and the ophthalmic artery at the base of the brain. However, Doppler is limited to patients with suitable bony windows in the skull; it is improved by using contrast material such as saline with bubbles (Newell and Aaslid 1992).

DSA provides valuable information about collateral flow, perfusion status, and other occult vascular lesions that may affect patient management and remains the most accurate technique to diagnose AVM, vasculitis, and dissection and to evaluate the degree of stenosis to determine eligibility for carotid endarterectomy or carotid angioplasty and stenting (Barr 2004). However, invasive DSA is associated with a small (<1 %) risk of serious complications, such as stroke or death, and when there is concordance of noninvasive methods, may not be necessary (Irimia et al. 2011).

The incidence of SAH increases with age, reaching 78 per 100,000 in the eighth decade of life; there is a female preponderance in nearly all ages (Sacco et al. 1984). MR angiography and CT angiography are both very useful for the diagnosis or screening of cerebral ­aneurysms larger than 5 mm, and CT angiography may be as sensitive and specific as DSA for their detection and characterization (Preda et al. 1998; Irimia et al. 2011). CT angiography is also less invasive and less resource-intensive, especially in out-of-hours on-call situations, replacing DSA as a first-line screening tool that can be performed immediately after acute SAH is detected on unenhanced CT (Fig. 23.17). Some post-processing of 3D volumetric renderings may be necessary to visualize the complex relationships of individual arteries to eliminate false positive caused by infundibula and vascular turns, since unlike DSA, selective arterial injection is not possible.

MR angiography can reliably detect aneurysms >3 mm, but sensitivity declines with size; thus, TOF and phase-contrast studies without contrast injection can be an excellent method for surveillance of incidental asymptomatic small aneurysms (annual rupture risk of 0.5–2 % compared to 4 % for symptomatic) and follow-up of patients after treatment (Locksley 1996; White et al. 2000) (Fig. 23.18). Aneurysm assessment should include demonstrating the location (typically at vessel bifurcations), number (multiple in 12–31 %, often with mirror occurrence), size, morphology, and the architecture of the neck in relation to the parent artery. MRI is also ideal for assessing the curvilinear laminated thrombi of different ages and signal intensity in giant (>2.5 cm maximum diameter) aneurysms (Atlas et al. 1987). Peripheral aneurysms are rare and may suggest mycotic infectious, traumatic, or neoplastic causes; besides the typical saccular (or berry) aneurysms, fusiform, atherosclerotic, and flow-related (associated with AVM) aneurysms may be seen.