General Imaging Characteristics of Hemorrhage

The appearance of ICH by CT can vary depending on the age of the hemorrhage and the hemoglobin level. The attenuation of blood is typically based on the protein content of which hemoglobin contributes a major portion. Therefore, the appearance of hyperacute/acute blood is easily detected on CT in patients with normal hemoglobin levels (approximately 15 g/dL) and typically appears as a hyperattenuating mass. This is because, immediately following extravasation, clot formation occurs with a progressive increase in attenuation over 72 hours due to increased hemoglobin concentration and separation of low-density serum. On the other hand, in anemic patients with hemoglobin less than 10 g/dL, acute hemorrhage can appear isoattenuating to the brain and can make detection difficult. Subsequently, following breakdown and hemolysis, the attenuation of the clot decreases until it becomes nearly isoattenuating to cerebrospinal fluid (CSF) by approximately 2 months. In the emergency setting, one should be aware of the “swirl” sign with unretracted clot appearing hypoattenuating and resembling a whirlpool; this may indicate active bleeding and typically occurs in a post-traumatic setting. It is important to recognize this sign, as prompt surgical evacuation may be required. The amount of mass effect on nearby tissues will depend on the size and location of the hemorrhage as well as the amount of secondary vasogenic edema that develops.

Administration of intravenous contrast material is usually not necessary for CT evaluation of ICH. If contrast is given, intra-axial hemorrhage can demonstrate an enhancing ring that is usually due to reactive changes and formation of a vascularized capsule, which typically occurs 5 to 7 days after the event and can last up to 6 months. Subacute and chronic extra-axial hematomas can also demonstrate peripheral enhancement usually due to reactive changes and granulation tissue formation. Unexpected areas of enhancement should raise concern, as active bleeding can appear as contrast pooling.

MR imaging has greatly revolutionized the evaluation of intracranial hemorrhage. The evolution of hemorrhage from the hyperacute to the chronic stage will have corresponding signal changes on T1-weighted images (T1WI), T2-weighted images (T2WI), fluid attenuated inversion recovery (FLAIR) images, and gradient-echo sequences. These properties can assist in detection and understanding of the time course of the injury. While beyond the scope of this chapter, description of the physics of the signal characteristics of blood products on MR is generally based on the paramagnetic effects of iron and the diamagnetic effects of protein in the hemoglobin molecule. The usual signal characteristics of hemorrhage and the general time course over which they evolve is summarized in Table 1-1.

Epidural Hemorrhage

Epidural hematoma is the term generally applied to hemorrhage that forms between the inner table of the calvarium and the outer layer of the dura due to its masslike behavior. More than 90% are associated with fractures in the temporoparietal, frontal, and parieto-occipital regions. CT is usually the most efficient method for evaluation of this type of hemorrhage. Epidural hematoma typically has a hyperdense, biconvex, or lens-shaped appearance. It may cross the midline but does not cross sutures (since the dura has its attachment at the sutures). This might not hold true if a fracture disrupts the suture. There is usually an arterial source, commonly due to tear of the middle meningeal artery and much less commonly (less than 10%) due to tear of the middle meningeal vein, diploic vein, or venous sinus (Figs. 1-1 and 1-2). The classic clinical presentation describes a patient with a “lucid” interval, although the incidence of this finding varies from 5% to 50% in the literature. Prompt identification of an epidural hematoma is critical, as evacuation or early reevaluation may be required. Management is based on clinical status, and therefore alert and oriented patients with small hematomas may be safely observed. The timing of follow-up CT depends on patient condition, but, generally, the first CT scan may be obtained in 6 to 8 hours and, if the patient is stable, follow-up may be extended to 24 hours or more afterward.

Figure 1-1 Epidural hematoma. A, CT shows usual biconvex, hyperdense acute epidural hematoma causing effacement of sulci and lateral ventricles and shift of midline structures. B, CT volume-rendered image shows nondisplaced fracture at vertex involving coronal suture. C, Coronal multiplanar reconstruction shows biconvex epidural hematoma crossing midline over the superior sagittal sinus.

Subdural Collections

Subdural hematoma (SDH) is the term generally applied to hemorrhage that collects in the potential space between the inner layer of the dura and the arachnoid membrane. It is typically the result of trauma (motor vehicle collision, assaults, and falls, the latter especially in the elderly population), causing tear of the bridging vein(s), and has a hyperattenuating, crescentic appearance overlying the cerebral hemisphere (Fig. 1-3). These hemorrhages can cross sutures and may track along the falx and tentorium but do not cross the midline. Inward displacement of the cortical vessels may be noted on a contrast-enhanced scan. There is a high association with subarachnoid hemorrhage. Acute SDHs thicker than 2 cm, seen with other parenchymal injuries, are associated with greater than 50% mortality. As the SDH evolves to the subacute (5 days to 3 weeks) and then chronic (more than 3 weeks) stage, it decreases in attenuation, becoming isodense to the brain and finally to CSF. Subacute SDH can have a layered appearance due to separation of formed elements from serum. Subacute hemorrhages may be relatively inconspicuous when isodense, and therefore it is especially important to recognize signs of mass effect, such as sulcal effacement, asymmetry of lateral ventricles, and shift of midline structures, as well as sulci that do not extend to the skull (Fig. 1-4). Bilateral isoattenuating SDHs can be especially challenging since findings are symmetric. Beware of this, particularly in the elderly patient who does not have generous sulci and ventricles. At this stage, the SDH should be conspicuous on MR imaging, especially on FLAIR sequences. A subacute SDH may also be very conspicuous on T1-weighted images due to the hyperintensity of methemoglobin.

Figure 1-3 Subdural hematoma. Mixed density layered pattern due to recurrent hemorrhages. Image (arrow) shows one method of measuring midline shift.

Figure 1-4 Isodense subdural hematoma. A, Sulcal effacement and midline shift to the right are clues to the presence of a left-sided subdural hematoma. B, Re-expansion of left Sylvian fissure and reduction in midline shift following evacuation.

Chronic subdural hematomas are collections that have been present for more than 3 weeks. Even a chronic hematoma may present in the emergency setting; for example, in a patient prone to repeated falls, brought in because of a change in mental status. On both CT and MR, these collections typically have a crescentic shape and may demonstrate enhancing septations and membranes surrounding the collection after contrast administration. Calcification of chronic SDH can occur and be quite extensive (Fig. 1-5). Areas of hyperdensity within a larger hypodense SDH may indicate an acute component due to recurrent bleeding, an “acute on chronic subdural hemorrhage.” Mixed density collections may also be acute, due to active bleeding or CSF accumulation due to tear of the arachnoid membrane. Chronic SDH is usually isointense to CSF on both T1WI and T2WI, but the appearance can be variable depending on any recurrent bleeding within the collection. The FLAIR sequence is typically very sensitive for detection of chronic SDH due to hyperintensity based on protein content. Hemosiderin within the hematoma will cause a signal void due to susceptibility effect, and “blooming” (appears to be larger than its true size) will be noted on gradient-echo sequence.

Figure 1-5 Calcified subdural hematomas. A, Colpocephaly configuration of lateral ventricles. B, Bone window/level settings more clearly show the calcified subdurals in this adult patient shunted as a child for congenital hydrocephalus.

Subdural hygroma is another type of collection and is commonly thought to be synonymous with chronic subdural hemorrhage. The actual definition of a hygroma is an accumulation of fluid due to a tear in the arachnoid membrane, usually by some type of trauma or from rapid ventricular decompression with associated accumulation of CSF within the subdural space. Many still use this term interchangeably with chronic subdural hematoma. CT demonstrates a fluid collection isodense to CSF in the subdural space. MR can be useful in differentiating CSF from chronic hematoma based on the imaging characteristics of the fluid on all sequences. Occasionally hygromas are difficult to differentiate from the prominence of the extra-axial CSF space associated with cerebral atrophy. Position of the cortical veins can be a helpful clue. In atrophy, the cortical veins are visible traversing the subarachnoid space, whereas with a hygroma they are displaced inward along with the arachnoid membrane by the fluid in the subdural space.

Subarachnoid Hemorrhage

Subarachnoid hemorrhage (SAH) fills the space between the pia and the arachnoid membrane, outlining the sulci and basilar cisterns. This can be due to a variety of causes, including trauma, ruptured aneurysm, hypertension, arteriovenous malformation, occult spinal vascular malformation, and hemorrhagic transformation of ischemic infarction. SAH is often associated with overlying traumatic subdural hematoma. Subarachnoid hemorrhages generally do not cause mass effect or edema. On CT, hyperdensity is seen within the sulci and/or basilar cisterns (Figs. 1-6 and 1-7).

Figure 1-6 Subarachnoid hemorrhage from ruptured aneurysm. A, Noncontrast CT shows ill-defined hyperdense subarachnoid hemorrhage in left Sylvian cistern (black arrow) and rim calcification in the wall of the aneurysm (white arrow). B, Volume-rendered image from CT angiography shows large aneurysm projecting above lesser sphenoid wing. C, Reconstruction from 3D rotational digital subtraction angiogram shows carotid-ophthalmic aneurysm to best advantage.

Figure 1-7 Subarachnoid hemorrhage and complications. A, Three CT images show diffuse hyperdense subarachnoid hemorrhage filling basal cisterns and cerebral sulci bilaterally. Diffuse loss of gray–white differentiation and effacement of sulci and cisterns probably preclude the need for further workup. B, Volume-rendered image from CT angiography demonstrates lack of intracranial flow consistent with expected elevation of intracranial pressure.

Although MR imaging may be as sensitive for the detection of acute intraparenchymal and subarachnoid hemorrhage, CT generally remains the modality of choice (and imaging gold standard). Sensitivity of CT for detection of SAH compared with CSF analysis can vary from up to 98% to 100% within 12 hours to approximately 85% to 90% after 24 hours of symptom onset. Other factors affecting sensitivity are hemoglobin concentration, size, and location of the hemorrhage. CT is widely available, rapid, and relatively inexpensive. In several small studies, MR has demonstrated sensitivity equivalent to CT for detection of acute parenchymal hemorrhage and SAH. In some cases of “CT-negative” (subacute) hemorrhage, MR has shown greater sensitivity. However, results may be confounded by artifacts from CSF pulsations, elevated protein (meningitis), or oxygen concentration (high fraction of inspired oxygen) in CSF on FLAIR images and presence of blood products from previous microhemorrhages on gradient-echo images.

Intraventricular Hemorrhage

Intraventricular hemorrhage (IVH) is typically caused by trauma in the adult population. It can result from extension of parenchymal hemorrhage into the ventricles or redistribution of subarachnoid hemorrhage. Primary intraventricular hemorrhage is uncommon and usually caused by a ruptured aneurysm, intraventricular tumor, vascular malformation, or coagulopathy (Fig. 1-8). Large IVHs are quite conspicuous on CT or MR. They may occupy a majority of the ventricle(s) and may result in hydrocephalus and increased intracranial pressure. Small amounts of IVH may be difficult to detect; one must check carefully for layering hemorrhage within the atria and occipital horns of the lateral ventricles. Normal choroid plexus calcifications in the atria of lateral ventricles, the fourth ventricle, and extending through the foramina of Luschka should not be mistaken for acute IVH.

Figure 1-8 Intraventricular hemorrhage. A, FLAIR MR image shows hyperintense hemorrhage isolated to frontal horn of right lateral ventricle. B, Image from right internal carotid artery digital subtraction angiogram in late arterial phase shows nidus (arrow) and early draining vein (arrowhead)diagnostic of an arteriovenous malformation.

Another less common type of extracerebral intracranial hemorrhage that may present acutely is pituitary hemorrhage. It is usually associated with pituitary apoplexy due to pituitary necrosis that may become hemorrhagic. Presenting symptoms may include headache, visual loss, ophthalmoplegia, nausea, and vomiting. Other causes of pituitary hemorrhage include tumors (macroadenoma, germinoma) and, less commonly, trauma.

Contusion

Parenchymal contusions result from blunt trauma and can occur in the cortex or white matter. Their locations are typically at the site of greatest impact of brain on bone, including the anterior/inferior frontal lobes and the lobes. They can be considered coup (occurring at the site of impact) or contrecoup (opposite the site of impact) types. On CT, a contusion typically appears as an area of hyperdensity with a surrounding rim of hypodense edema. It can initially appear as a focal area of subtle hypodensity and may blossom on follow-up exam at 12 to 24 hours with development of an obvious central area of hyperdensity and a larger surrounding zone of hypodense edema (Fig. 1-9). On MR, signal characteristics reflect the hemorrhagic and edematous components. Over time, the density and signal characteristics of the hemorrhage will evolve in a fashion similar to a spontaneous hemorrhage. Parenchymal hemorrhage due to penetrating trauma, such as from gunshot wound or impalement, will follow the same general pattern of evolution.

Figure 1-9 Blossoming contusion. A, CT shows thin left frontoparietal subdural hematoma tracking along anterior falx and mild sulcal effacement. B, Follow-up CT after 24 hours shows hyperdense parenchymal hemorrhage and surrounding edema in left frontal lobe and stable subdural hematoma. Notice mass effect on left lateral ventricle.

Diffuse Axonal Injury

Diffuse axonal injury (DAI) is another type of traumatic brain injury that may present with parenchymal hemorrhages and is distinct from parenchymal contusion. DAI is an injury to the axons caused by acceleration/deceleration injury with a rotational component (usually from motor vehicle collision or other blunt trauma to the head). There may be complete transection of axons with injury to the associated capillaries, or partial disruption of the axons. The lesions of DAI typically occur at the inter-faces of gray and white matter in the cerebral hemispheres, the body and splenium of the corpus callosum, midbrain, and upper pons. Lesions may also be seen in the basal ganglia. Patients sustaining DAI typically lose consciousness at the moment of impact. DAI may be suspected when the clinical exam is worse than expected based on the findings on initial CT scan. Usually, the greater the number of lesions, the worse the prognosis. Individuals who recover usually demonstrate lingering effects such as headaches and cognitive deficits. Initial CT scans in more than half of patients with DAI may be negative. CT findings include hypodense foci due to edema in areas of incomplete axonal disruption and hyperdense foci due to petechial hemorrhage where there is complete transection of the axons and associated capillaries (Fig. 1-10). MR is more sensitive than CT for detection of DAI. Approximately 30% of those negative by CT will demonstrate abnormal findings on MR. These findings include FLAIR and T2 hyperintensities (edema) and gradient-echo hypointensities (hemorrhages) (Fig. 1-11). Lesions may appear hyperintense on diffusion-weighted images. It is estimated that more than 80% of the lesions of DAI are nonhemorrhagic. Generally, if imaging is repeated within 3 to 5 days, more lesions will become apparent as the process evolves. A staging system for DAI based on locations of lesions on histopathology may be applied to MR findings. Stage 1 is based on subcortical lesions in the frontal and temporal lobes. Stage 2 will also show lesions in the corpus callosum and lobar white matter, and stage 3 will have lesions in the midbrain and pons. Diffusion tensor imaging, particularly helpful in evaluation of white matter tracts, has been shown to be more sensitive than conventional MR for detection of diffuse axonal injury and correlates more closely with clinical outcomes.

Figure 1-10 Shear hemorrhages of diffuse axonal injury. Noncontrast CT shows three small hemorrhages in the left superior frontal gyrus following blunt head trauma.

Figure 1-11 Diffuse axona linjury.A, FLAIR hyperintensities in posterior limb of internal capsule (arrow) and subcortical white matter (arrowhead). B, Gradient-echo hypointensities in subcortical white matter indicative of hemorrhages (arrows) were not evident on other sequences. Note IVH in occipital horns of lateral ventricles.

Subfalcine Herniation

Subfalcine herniation is due to displacement and impingement of the cingulate gyrus underneath the falx. It is usually caused by mass effect on the frontal lobe and is associated with ipsilateral lateral ventricle compression and obstruction of the foramen of Monro with dilatation of the contralateral ventricle (“trapped ventricle”). The degree of midline shift (not synonymous with subfalcine herniation) can be estimated by drawing a line between the anterior and posterior attachments of the falx and measuring the shift of the septum pellucidum relative to this line (see Fig. 1-3). Anterior cerebral artery territory infarct(s) may result from this type of herniation.

Transtentorial Herniation

Transtentorial herniations include two major types, which are the descending transtentorial herniation (DTH) and the ascending transtentorial herniation (ATH). An early DTH is known as uncal herniation, in which the uncus is displaced medially and occupies the ipsilateral suprasellar cistern. A later-stage DTH is caused by continued mass effect with displacement of the medial temporal lobe through the incisura, which completely occupies the suprasellar cistern (along with the uncus) and causes enlargement of the ipsilateral and effacement of the contralateral ambient cisterns. This phenomenon occurs because, as there is marked shifting of brain in the supratentorial compartment, there is also shift of the brainstem in the same direction. There is also compression of the ipsilateral cerebral peduncle. Occasionally, when there is marked mass effect, there can be compression of the contralateral cerebral peduncle against the tentorium, or “Kernohan’s notch,” which leads to ipsilateral motor weakness (this phenomenon may be a false localizing sign). Other imaging findings include a “trapped” temporal horn of the lateral ventricle contralateral to the side of the mass and Duret hemorrhages—hemorrhages of the midbrain and pons caused by stretching and tearing of the arterial perforators. In cases of bilateral mass effect, there can be displacement of both temporal lobes and midbrain through the incisura leading to effacement of the basilar cisterns bilaterally. Complications of this type of herniation include compression of the posterior cerebral artery and penetrating basal arteries with associated infarcts in these vascular distributions (see Fig. 1-2). In addition, there can be compression of the oculomotor nerve (CN III) with an associated palsy. ATH is less common and is caused by superior displacement of the cerebellum and brainstem through the incisura. It is usually due to mass effect in the posterior fossa (as from hemorrhage, tumor, or infarct), and on imaging there is compression on the posterolateral midbrain bilaterally with associated effacement of the ambient and quadrigeminal plate cisterns. There is usually hydrocephalus due to obstruction at the level of the cerebral aqueduct of Sylvius.

Figure 1-2 Epidural hematoma and complications. A, “Swirl” sign in this large epidural hematoma suggests continued bleeding. B, Pontine (Duret) hemorrhage (arrow) and effacement of basal cisterns as a result of downward herniation. C, Uncal herniation (arrow shows margin of left temporal lobe) and resultant left posterior cerebral artery territory infarct. The brainstem is distorted left and also abnormally hypodense. D,Infarcts in bilateral anterior, left middle, and left posterior cerebral artery territories as a result of herniations.

Tonsillar Herniation

Tonsillar herniation is caused by downward displacement of the cerebellar tonsils through the foramen magnum into the spinal canal (generally by more than 5 mm). On imaging, there is a peglike configuration to the tonsils with obliteration of the CSF space in the foramen magnum (Fig. 1-12). Complications include obstructive hydrocephalus from compression of the fourth ventricle. Mild tonsillar ectopia, Chiari I malformations, and sagging tonsils due to intracranial hypotension should not be mistaken for acute tonsillar herniation, but should be considered seriously when downward mass effect is expected based on brain edema, mass, or hemorrhage.

Figure 1-12 Tonsillar herniation. A, Sagittal T1 image shows pegged appearance of the cerebellar tonsils extending through the foramen magnum, simulating a Chiari I malformation. (This was a dramatic change from a previous exam.) B, Postgadolinium T1 image shows cerebellar leptomeningeal enhancement due to cryptococcal meningoencephalitis in this patient with AIDS.

Extracranial Herniation

An extracranial herniation is the displacement of brain parenchyma through a cranial and dural defect that is usually caused by trauma or craniectomy (usually performed to prevent downward herniation from acute cerebral edema). Complications may include infarct of the herniated brain tissue.

Transalar Herniation

Transalar herniation is uncommon and, by itself, does not cause symptoms. It is usually associated with subfalcine and transtentorial herniations. This type of herniation is caused by displacement of the temporal lobe anteriorly or frontal lobe posteriorly across the sphenoid wing. Look for anterior or posterior displacement of the middle cerebral artery to identify this type of herniation.

In the setting of severe head trauma, many of these different types of injuries may coexist. The mechanism of injury should correspond with the degree of injury. In cases when the reported mechanism is mild, nonaccidental trauma (“Trauma X,” “shaken-baby syndrome”) should be considered. Infants, children, those with mental or physical disabilities, and the elderly are particularly at risk. Skull fractures, SAH, SDH, contusions, shear injuries, infarcts, vertebral compression fractures, and retinal hemorrhages constitute the usual neuroradiologic spectrum of abnormalities. Injuries of different ages, metaphyseal and rib fractures, and visceral injuries are other common findings in child abuse.

Hemorrhagic Stroke: Spontaneous Parenchymal Hemorrhage

Approximately 10% to 15% of strokes present with an acute parenchymal hemorrhage. The most common cause is hypertension (Fig. 1-13). Coagulopathies, hematologic disorders including hypercoagulable states, amyloid angiopathy, drugs, vascular malformations and aneurysms, vasculitides, and tumors round out the usual list of etiologies. Hemorrhages resulting from illicit drug use and vascular malformations are commonly found in young adults (Fig. 1-14). Sickle cell disease and venous infarcts may also present with parenchymal hemorrhage. A ruptured intracranial aneurysm may occasionally cause a parenchymal hemorrhage in association with subarachnoid hemorrhage. Hypertensive hemorrhages most commonly occur in the basal ganglia and thalami but may also primarily arise within the cerebral hemispheres, brainstem, or cerebellum. Cerebral amyloid angiopathy (CAA) is another common cause of intracranial hemorrhages in patients over 65 years of age. CAA can be found in patients with mild cognitive impairment, dementia of Alzheimer type, and Down syndrome with extracellular deposition of beta amyloid occurring in the cortex and subcortical white matter. CAA can be hereditary (autosomal dominant, Dutch type), sporadic (presence of Apoε4 allele), or acquired (as from hemodialysis). The lobar hemorrhages of CAA typically occur in the frontal and parietal regions. MR is sensitive for the detection of hemosiderin deposition resulting from multiple microhemorrhages over the course of time appearing as small hypointense foci on gradient-echo sequences.

Figure 1-13 Spontaneous parenchymal hemorrhage secondary to uncontrolled hypertension. Noncontrast CT in a patient with uncontrolled hypertension shows a large hyperdense parenchymal hemorrhage arising in the basal ganglia with intraventricular extension. Note the ventriculostomy catheter in the right lateral ventricle.

Figure 1-14 Parenchymal hemorrhage due to illicit drug use—Ecstasy (3,4 methylenedioxymethamphetamine). A, CT shows hyperdense acute hemorrhage with minimal surrounding hypodense edema. B, T1-weighted image shows iso- to mild hyperintensity with hypointense edema. FLAIR (C) and fat-suppressed T2 gradient and spin echo (D) show hypointensity with surrounding hyperintense edema. E, Gradient echo shows peripheral rim of signal loss and blooming. In summary, signal changes on T1-weighted image and T2-weighted image are consistent with deoxyhemoglobin although the gradient echo suggests only a rim of deoxyhemoglobin.

Imaging of Acute Ischemic Stroke

Computed Tomography

The role of imaging in acute stroke diagnosis and management continues to evolve. Since the mid-1970s, unenhanced CT has been the first-line modality to determine the etiology of acute neurologic deficits. CT can offer the chance to detect an ischemic infarct, generally in the middle cerebral artery territory, within 3 hours in up to one third of cases based on findings of subtle parenchymal hypodensity, loss of gray–white matter differentiation (including loss of the insular ribbon or margins of basal ganglia; Fig. 1-15), and effacement of sulci. A hyperdense vessel sign may indicate the presence of an acute thrombus and support the diagnosis. The sensitivity for detection of acute stroke has been shown to increase with the use of an “acute stroke” window and level settings (see Fig. 1-15). A very narrow window width of 8 Hounsfield units (HU) and a level of 32 HU (compared with 80 and 20 HU, respectively) may increase the sensitivity of CT to approximately 70% without a loss of specificity. CT is currently used to screen patients who may be considered for treatment with intravenous recombinant tissue plasminogen activator (rt-PA) within 3 hours of onset based on guidelines from the National Institute of Neurologic Disorders and Stroke (NINDS) rt-PA trial. Beyond 3 hours, the risk of intracranial hemorrhage due to intravenous thrombolysis was shown to outweigh potential benefits. An association between larger stroke volumes (greater than one third of the middle cerebral artery territory) and reperfusion hemorrhage was initially reported. This criterion for the use of 100 mL estimated infarct volume has been commonly applied in stroke trials. The Alberta Stroke Program early CT score (ASPECTS), a 10-point topographic scoring system, was later developed to try to more easily quantify initial stroke volumes. This score has been shown to correlate with the initial National Institutes of Health stroke score (NIHSS).

Figure 1-15 Hyperacute infarct. A, Noncontrast CT with window/level of 80/40 shows subtle decrease in density of the right insular cortex. B, Insular “ribbon” sign is more conspicuous with stroke window/level of 40/40 (arrows). C, Infarct is much more conspicuous on diffusion-weighted image. D,Time-of-flight MR angiography shows attenuation of right middle cerebral artery distal branches.

Magnetic Resonance: Diffusion-weighted Imaging

MR, with diffusion-weighted imaging (DWI), which became widely available in routine clinical practice in the late 1990s, offers significantly greater sensitivity and specificity for the detection of acute stroke (greater than 90% compared with approximately 60% for CT). To put it very simply, energy depletion will trigger a cascade that will alter the internal cellular milieu and result in development of cytotoxic edema. The restriction of water molecule diffusion appears as hyperintensity on diffusion-weighted images. The diffusion “experiment” can be performed with a variety of rapid imaging techniques, such as echo planar imaging, and can acquire images of the entire brain in half a minute. This minimizes the effects of patient motion, especially important when the clinical presentation includes alteration of mental status. The apparent diffusion coefficient value is a quantitative measure that may be calculated from the diffusion-weighted images. Since diffusion-weighted images rely on both diffusion and T2 effects, it is wise to confirm that the apparent diffusion coefficient values are indeed reduced before diagnosing an acute infarct. This will reduce the number of false positives due to “T2 shine-through” effect from old infarcts (gliosis) or other T2 hyperintense processes such as vasogenic edema.

Acute ischemic infarcts may appear as hyperintense regions on DWI (see Fig. 1-15), and this can occur as quickly as 30 minutes after onset. Up to 100% sensitivity has been demonstrated in clinical studies. However, in routine practice, small lesions in the brainstem may not be perceived initially, only to be detected on a follow-up exam prompted by persistent symptoms. It is also possible that a region of ischemia (prior to completed infarction) may go undetected on an initial imaging study, resulting in a false negative result. False positives on DWI can be due to processes that mimic stroke and also cause diffusion restriction, such as certain neoplasms, multifocal metastatic disease, and abscesses. Presence or absence of associated findings on conventional MR sequences—such as loss of gray–white matter differentiation on T1WI, and hyperintense edema on FLAIR and T2WI—may help with diagnosis, although these signs may be inconspicuous for 6 to 12 hours after stroke onset. Blooming on gradient-echo sequences due to intravascular thrombus and loss of expected vascular flow voids are other useful clues.

Lacunar infarcts are generally less than 1 cm in diameter and presumed to be due to occlusion of small perforating branches due to embolic, atheromatous, or thrombotic lesions. They occur most commonly in the basal ganglia, internal and external capsules, immediate periventricular white matter (corona radiata), and, less frequently, the centrum semiovale. Occlusion of basilar artery perforators will result in lacunes in the brainstem. Diffusion imaging offers the ability to identify very small, acute infarcts even in the background of chronic white matter disease and remote lacunes (Fig. 1-16). While MR is still considered a relatively expensive technique, it has the potential to reduce the number of unnecessary hospital admissions for recurrent small vessel infarcts in many patients. It may also help to select the most appropriate pathway for patients with central embolic sources of infarcts based on detection of infarcts in different vascular territories.

Figure 1-16 Acute lacunar infarct. A, CT shows multifocal hypodensities. B, FLAIR image shows corresponding hyperintensities due to chronic small vessel disease. C, Diffusion-weighted image shows the acute infarct in the left lentiform nucleus/posterior limb of internal capsule in a patient with acute onset of right-sided weakness. Apparent diffusion coefficient map (not shown) confirmed restricted diffusion. Other periventricular white matter mild hyperintensities are the result of “T2 shine-through.”

MR is also valuable in the setting of neonatal hypoxic ischemic encephalopathy. Cranial ultrasonography and CT may be used to evaluate germinal matrix hemorrhages, periventricular leukomalacia, and hydrocephalus. Diffusion-weighted MR is most sensitive for evaluating the different patterns of injury. In preterm infants subjected to mild hypotension, the periventricular regions are most often affected. With more severe hypotension, the basal ganglia, brainstem, and cerebellum may be involved. In full-term infants with mild hypotension, infarcts in the border zones between anterior and middle cerebral arteries or between middle and posterior cerebral arteries may result. Severe hypotension may result in infarcts of basal ganglia, hippocampi, corticospinal tracts, and sensorimotor cortex.

Diffusion-weighted hyperintensity generally begins to decline after a few days, with the process of apparent diffusion coefficient (ADC) pseudonormalization usually taking place over the next few weeks. Final ADC values will vary based on degree of gliosis or cavitation of the infarct. It should be noted that infarction development depends on the magnitude and duration of ischemia and the metabolic demands of the affected tissue. While diffusion restriction due to ischemia almost always results in infarction, rare cases of spontaneous reversible diffusion abnormalities have been reported, as well as those occurring in the setting of thrombolytic therapy.

Computed Tomography: Perfusion Imaging

The advent and wide availability of helical and, more recently, of multidetector CT (MDCT) scanners has led to the use of CT angiography (CTA) and CT perfusion (CTP) imaging in the workup of suspected acute ischemic stroke. The ability to acquire a large 3D volume of data rapidly during bolus contrast administration combined with submillimeter spatial resolution results in CT angiographic images that in many instances approach the diagnostic quality of more invasive digital subtraction angiograms. Large detector arrays and cine or shuttle modes of scanner operation allow for generous coverage of the brain during perfusion studies. Detector arrays from 4 to 64 rows wide are available now and newer models with 256 and more are in development. Analogous to PWI, CT perfusion is based on measurement of tissue density as a function of time during a first pass of intravenous contrast and commonly uses deconvolution analysis. In contrast to PWI, as a result of the proportionate increase in tissue attenuation due to iodine concentration, quantitative (rather than relative) estimation of CBV (and therefore CBF) can be obtained by CTP.

CTP applied in the setting of acute stroke has been validated with clinical outcomes and follow-up imaging and also by comparison with diffusion-weighted imaging. Wintermark and colleagues have proposed that regions with CBV less than 2.5 mL/100 g be considered the “core infarct” and regions with CBF reduction of more than 34% be defined as the penumbra. These values are based on correspondence with initial DWI abnormality and final infarct size. Schaefer and colleagues have proposed other absolute values and the use of normalized CBV and CBF ratios to help distinguish ischemic tissue likely to infarct from that likely to survive following intra-arterial recanalization therapy. Perfusion maps can be visually compared to determine if a penumbra is present but may underestimate or overestimate the extent of tissue at risk (Fig. 1-17). Software packages are available that can automatically segment the processed perfusion maps into core infarct, penumbra, and normal regions based on thresholding techniques.

Figure 1-17 Perfusion CT—large mismatch. A, Noncontrast CT shows decreased density of left caudate head and putamen. B, CBF and mean transit time abnormalities are much larger than core infarct (CBV abnormality). (Color functional maps are more easily interpreted on a color monitor.) C, CT follow-up 1 week after conservative management shows that the final infarct volume closely matches the initial CBV abnormality. D, Initial CT angiogram showed abrupt occlusion of left middle cerebral artery stem (arrow) but with filling of distal branches due to robust leptomeningeal collateral circulation.

The ongoing MR RESCUE (MR and recanalization of stroke clots using embolectomy) trial is employing a similar, automated analysis of PWI data to determine eligibility for inclusion. Patients presenting with intracranial arterial occlusions within 8 hours of symptom onset and having a suitable penumbra are randomized to embolectomy with the Merci Retriever device (Fig. 1-18).

Figure 1-18 Mechanical thrombectomy. A, Acute right middle cerebral artery occlusion. B, Merci retriever deployed distal to thrombus. C, Recanalization of right middle cerebral artery following clot retrieval.

Data from the recently completed DIAS 2 trial, a phase III study that used either DWI/PWI/MRA or CT/CTP/CTA as eligibility criteria, are still being analyzed. However, initial reports suggest that the results were not as favorable as those of previous trials. It is hoped that analysis of outcomes with respect to imaging features may help to identify a subset of patients who might achieve therapeutic benefits with a reduced risk of symptomatic hemorrhage.

Hemorrhagic Transformation

Hemorrhagic transformation of an ischemic infarct, thought to result from reperfusion injury, can be a dreaded complication of therapy or may happen spontaneously within hours or after a period of several weeks (Fig. 1-19). If a large hematoma is present at the time of initial imaging, it may not be possible to distinguish a primary parenchymal hemorrhage from hemorrhagic transformation. The blood products generally cause substantial artifacts on DWI. However, if the hemorrhage is confined within a larger zone of restricted diffusion, the etiology may be clear. Petechial hemorrhage occurs very commonly within an ischemic infarct, is best detected by gradient-echo imaging, and does not usually lead to increased morbidity.

Figure 1-19 Hemorrhagic transformation of ischemic infarct. Noncontrast CT 24 hours postictus shows hyperdense hemorrhage within a large left middle cerebral artery territory infarct.

Cortical Laminar Necrosis

A pattern of gyriform T1 hyperintensity developing a week or two after an ischemic infarct may be attributed to cortical laminar necrosis (Fig. 1-20). It seems that gray matter is more vulnerable to ischemic necrosis than white matter (especially the third layer of the six cortical layers), and, although the signal changes may lead one to diagnose hemorrhage or calcification, in one histologic study neither was detected. The exact cause of the T1 shortening is uncertain, but it may be due to high concentrations of proteins and macromolecules.

Figure 1-20 Cortical laminar necrosis. T1-weighted image shows gyriform hyperintensity in a perisylvian distribution 1 week after middle cerebral artery infarct.

Cerebral Venous Infarction and Sinus Thrombosis

An uncommon (annual incidence estimates of less than 1 case per 100,000 population) but important cause of hemorrhage is cerebral venous thrombosis (Fig. 1-21). This may affect cortical veins and other portions of the superficial and deep venous drainage systems. Hypercoagulable states due to pregnancy and the postpartum period, or from oral contraceptive use, dehydration, regional infections, and trauma, are relatively common causes. Common presenting symptoms include headache, seizure, and focal neurologic deficits. Fluctuating symptoms and intracranial hypertension are common as well. Bilateral parenchymal hemorrhages or infarcts that do not obey usual arterial territorial borders can be clues to diagnosis. A working knowledge of the normal anatomy of the major venous structures and the common variants is necessary to avoid diagnostic pitfalls, especially false positives. On unenhanced CT, normal venous structures may appear denser than usual due to dehydration or elevated hematocrit, whereas thrombosis should appear hyperdense relative to arteries. A filling defect or occlusion may be detected on contrast-enhanced CT. CT venography can be performed with thin-section, volumetric technique allowing for creation of 2D and 3D reconstructions. Normal and thrombosed venous sinuses take on many different appearances on MR depending on scan parameters, flow velocity, and turbulence. Unexpected hyperintensity, loss of usual flow voids, and blooming on gradient-echo sequence are clues that flow-sensitive MR venography (MRV) should be performed. Time-of-flight MRV may be less sensitive than phase contrast technique due to shine-through of methemoglobin in a thrombosed sinus simulating flow in the vessel. Contrast-enhanced MRV may be more sensitive, although enhancement of chronic thrombus can be misleading. Associated findings of edema, hemorrhage, or ischemic infarct may help toward arriving at the correct diagnosis, but brain swelling without signal changes has been reported in up to approximately 40% of patients. Cavernous sinus thrombosis is discussed in relation to complex sinus and orbit infections in the section on head and neck imaging. Prompt diagnosis of cerebral thrombosis is critical, as many of the parenchymal changes may be reversible. Systemic anticoagulation and local catheter-based thrombolytic, mechanical, or rheolytic clot dissolution are treatment considerations. Intracranial hypertension and collateral formation leading to dural arteriovenous malformations are possible long-term sequelae.

Figure 1-21 Venous sinus thrombosis. A, T1-weighted image shows loss of usual flow voids in superior sagittal and straight sinuses (arrows). B, Two-dimensional phase contrast MR venography shows corresponding lack of flow-related signal. C, Three-dimensional phase contrast MR venography 1 week after systemic anticoagulation shows recanalization.