A simplified description of technical imaging considerations is included to highlight solutions for some common pitfalls that can be encountered. Axial images using 5 mm reconstructions are generally used in NCCT of intracranial hemorrhage, as this thickness has an adequate signal-to-noise ratio to allow for evaluation of the brain parenchyma, identification of intracranial hemorrhage, and assessment of intracranial mass effect. Unfortunately, in some cases intracranial hematoma can be obscured by nearby skeletal structures or be mistaken for dural prominence or calcifications [15]. Reviewing a wider “blood” window can improve the detection of intracranial hemorrhage which layers along the skull (Fig. 2a, b). Review of coronal and sagittal reformatted images is also paramount in all cases but particularly when findings of hemorrhage are equivocal. Reformatted images allow for visualization parallel and/or perpendicular to the long axis of the hematoma allowing for improved diagnostic accuracy [16] (Fig. 3a, b). Similarly, thin-cut images may help resolve any questionable areas (Fig. 2c). In the interest of time, thin section cuts should only be created and reviewed when there is a specific area of concern. Review of bone algorithm reconstructions is also suggested (particularly in cases of trauma) for identification of skull fractures which will heighten the readers’ suspicion of underlying intracranial hemorrhage and help to exclude mimickers of hemorrhage like small calcifications (Fig. 3c).

It is important to note that in the first minutes after hemorrhage, there is a partial overlap between the HU of blood products and that of brain parenchyma, creating a potential pitfall in diagnosis. Similarly, intracranial hemorrhage can remain occult in patients with low hematocrit because of isodensity to brain parenchyma [17]. In such cases, and in those where there is a high index of suspicion for intracranial hemorrhage without definite CT abnormality, MRI should be recommended for further evaluation.

The evaluation of blood products on magnetic resonance imaging (MRI) is much more involved than on CT. The appearance depends on a myriad of biologic factors which evolve with time and vary based on operator-dependent factors. A basic understanding of magnetic properties as well as the biologic and technical factors which affect the appearance of intracranial blood products aids in overcoming confusion and avoiding potential pitfalls. What follows is a description of the key factors involved and their effects on the appearance on intracranial blood products.

Magnetic susceptibility refers to the measure of substances’ interaction and response to an applied magnetic field [18, 19]. In simplified terms, the magnetic susceptibility of a substance can be attributed to the pairing of its valence electrons [19]. Those substances which have no unpaired valence electrons are termed diamagnetic. Paired electrons in an orbital have opposing spins which cancel each other, resulting in no magnetic dipole when placed in a magnetic field [19]. When a diamagnetic substance is placed in an applied magnetic field, its valence electrons do experience minimal alterations in their orbital motion, however [19]. According to Lenz’s law, this results in a weak induced magnetic field which opposes the applied field [20].

In contrast to diamagnetic substances, paramagnetic substances have unpaired valence electrons. When placed in a magnetic field, the spins of the valence electrons align with the magnetic field creating a magnetic dipole which augments the applied field [19, 20]. In human tissues, the magnetic dipole of paramagnetic substances allows them to undergo proton-electron dipole interactions with local protons (i.e., those in water molecules) [21]. These interactions have a profound effect on the magnetic microenvironments of human tissues, most significantly affecting the longitudinal relaxation of nearby protons, resulting in T1 shortening [20, 22]. For this discussion, it is important to note that the strength of dipole interactions varies inversely with the sixth power of the distance between the dipoles [23]. Also, as magnetic field strength increases, paramagnetic properties become more pronounced because paramagnetic substances augment the applied field more so than diamagnetic substances oppose it [24]. Relevant compounds are classified based on their magnetic susceptibility in Table 1.

Several biologic factors contribute to the characteristic appearance of intracranial blood products on MRI. As time passes, intracranial blood products are exposed to metabolic processes and variations in their local environment which alter their chemical and magnetic properties. What follows is a brief description of these processes and the resultant effects they have on the MRI signal produced.

The appearance of intracranial blood products can be significantly altered depending on the magnetic field strength and the pulse sequence used. Magnetic field strength imparts its influence by altering susceptibility-induced T2 effects. The inverse of the T2 relaxation rate varies quadratically with the field strength [24, 37, 38]. This relationship is a product of the relatively increased magnetization of paramagnetic compounds in relation to diamagnetic compounds at higher field strengths [24]. This exaggerated difference in magnetization results in greater magnetic field inhomogeneity between paramagnetic and diamagnetic microenvironments causing more pronounced T2 shortening (T2 darkness).

Gradient recall echo (GRE) sequences are highly sensitive to susceptibility effects caused by magnetic field in homogeneities [38–40]. GRE sequences are, in turn, highly sensitive for the detection of intracranial hemorrhage and highly accurate at predicting its extent [38, 41]. This makes GRE particularly valuable in the hyperacute and chronic periods, when blood products might otherwise remain inconspicuous [38, 39]. As a result, GRE sequences have become widely incorporated in standard MRI protocols for the evaluation of intracranial hemorrhage.

Fast spin echo (FSE) sequences have become a mainstay of standard MRI protocols, widely replacing conventional spin echo. By using shorter interecho times and closely spaced pulses, FSE allows for significant reductions in scan times [42, 43]. These shorter echo times have been shown to cause increased T2 prolongation by a variety of mechanisms [42]. An untoward result of this T2 prolongation is a decreased sensitivity to T2 susceptibility effects, although the clinical significance of this has been questioned [43]. Nevertheless, it is recommended that GRE sequences be included in standard MRI protocols to ensure that high sensitivity for small amounts of blood products is maintained despite the usage of FSE.

CT angiography (CTA) is a useful adjunct to both NCCT and MRI in the evaluation of patients with acute intracranial hemorrhage. CTA images can be rapidly acquired and reconstructed to allow intricate visualization of the intracranial vessels in multiple planes and projections. It has high sensitivity, specificity, and diagnostic accuracy in identifying vascular causes of intraparenchymal hemorrhage and can be beneficial in evaluating the vascularity of hemorrhagic neoplasms [44–48].

In addition to its role in diagnosis, CTA can provide valuable prognostic information in the patient with intracranial hemorrhage. The CTA “spot sign” is a high-attenuation focus within an intracerebral hematoma which represents a site of extravasation [49] (Fig. 6). In patients with intracerebral hematoma, the “spot sign” has been identified as a predictor of hematoma progression, longer mean hospital stay, in-hospital mortality, and poor overall outcomes in survivors [50–53].

Although CTA plays an important role in the evaluation of patients with intracranial hemorrhage, it has several notable limitations. Reconstructed images (i.e., volume-rendered and maximum-intensity-projection images) can enhance artifacts and are susceptible to data loss, potentially confusing the diagnosis of vascular pathology [54]. Pitfalls such as this can be avoided by reviewing the source images in all cases and using multiplanar reformatted images (which contain all source data), rather than relying on reconstructed images [54]. Acute extraluminal blood products can also obscure evaluation of the arteries, although this can be minimized with proper arterial enhancement, as the Hounsfield density of intravenous contrast is significantly higher than that of acute extraluminal blood. Unfortunately, other artifacts are not as easily avoided as these.

Evaluation of patients with previous placement of embolic materials, endovascular coils, and surgical aneurysm clips is often suboptimal secondary to marked streak artifact. Also, CTA is limited in evaluating small arteries as well as arteries near bony structures, secondary to limited spatial resolution, poor contrast between arteries and adjacent bones, and beam-hardening artifact [55, 56]. Because of these limitations, catheter-based digital subtraction angiography (DSA) remains the gold standard in the evaluation of intracranial hemorrhage [48, 57, 58].

Magnetic resonance angiography (MRA) is an extensive topic whose breadth is beyond the scope of this chapter. There are, however, important concepts which will allow for a more thorough understanding of its value and shortcomings in the diagnostic algorithm of intracranial hemorrhage. The three main types of magnetic resonance angiography are time-of-flight MRA (TOFMRA), phase-contrast MRA (PCMRA), and contrast-enhanced MRA (CEMRA). PCMRA allows for evaluation of hemodynamic flow, but its use in the anatomic evaluation of intracranial vasculature is at present dwarfed by TOFMRA. Therefore, this discussion will focus on the important strengths and weaknesses of non-contrast TOFMRA and CEMRA in the evaluation of intracranial hemorrhage.

TOF imaging, in its simplest form, takes advantage of the flow of blood. In TOF, radiofrequency (RF) pulses are applied at short TRs to a given slice thickness resulting in the saturation of all the magnetic spins within the slice and suppression of their longitudinal magnetization [59]. Spins that are moving perpendicularly into and out of the slice are not saturated, however. In a given length of time, the blood that has been saturated by the RF pulses will move out of the slice and be replaced by “new” unsaturated blood. When exposed to an excitatory pulse, this “new” blood is magnetized and can undergo longitudinal relaxation with resultant bright signal (Fig. 7). The relative signal of the flowing blood will be directly related to the slice thickness and the velocity of the blood within the vessel. Maximal signal is obtained when all saturated blood is replaced with unsaturated blood. This occurs when the velocity of flow is greater than or equal to the slice thickness divided by the repetition time, represented by the equation v ≥ z/TR, where v is the velocity of flow, z is the slice thickness, and TR is the pulse repetition time [60, 61]. Additional factors that affect the signal of flowing blood include the flip angle of the pulse, the T1 of the tissue in the slice, and the orientation of the vessel to the imaging slice [59, 61]. In the case of the intracranial circulation, flow signals are optimized by selecting axial slice acquisition, which is perpendicular to the average overall intracranial arterial flow. Similarly, optimal TR and slice thickness can be selected to ensure that the signal of the static tissue is suppressed and that the signal in the vessels is maximally enhanced.

TOF imaging identifies any flow within an image slice; therefore, both arterial and venous structures will produce signal. To selectively highlight the vascular structure of interest (artery or vein) and to minimize contamination of the other, saturation bands are employed. These pre-saturation bands are positioned a certain distance from the imaging slice to suppress the undesired flow signal before it arrives within the imaging slice [62] (Fig. 8).

Because of its dependence on flow and spin saturation, TOF is susceptible to many artifacts. When flow within a vessel is slow or turbulent, or the orientation of flow is parallel (or near parallel) to the imaging slice, the blood spins have a longer “dwell time” within the image slice [63]. Increased “dwell time” exposes the spins to more saturating pulses, leading to suppressed signal in these portions of the vessel, which can mimic vessel narrowing or occlusion [63]. Additional sources of artifact include bright T1 substances, tissue interfaces, pulsation, and foreign bodies. Methemoglobin (T1 bright) presents a particular challenge in TOF because it is often present in hemorrhage and/or vessel thrombus. The T1 signal from methemoglobin can be difficult to suppress, and it recovers its T1 signal rapidly, decreasing the conspicuity of vessels from background tissue and even mimicking flow where there is none [58, 64]. Metallic foreign bodies (surgical clips) and tissue interfaces can dephase spins in nearby vessels causing alteration in their apparent configuration and diameter, although these are a less significant source of artifact [65, 66].

The artifacts associated with TOF can make proper evaluation of aneurysms, arteriovenous malformations (AVM), and dural venous fistulas (DAVF) very challenging. Aneurysms can contain complex flow patterns which vary based on their size and morphology. Turbulent and slow flow within the aneurysm results in loss of signal on TOF, leading to underestimation of the aneurysm volume and neck size [67]. Also, the presence of methemoglobin in thrombosed aneurysms can lead to false-positive results [66]. Similarly, AVMs and DAVFs pose a formidable diagnostic challenge on TOFMRA because of their vessel tortuosity, complex angioarchitecture, and slow-flow draining veins. These properties cause TOFMRA to overestimate the size of small AVMs and underestimate the size of large AVMs [68]. Additionally, TOF is limited in correctly interpreting small feeding arteries and in identifying draining veins of AVMs and DAVFs, which can have significant treatment implications [68, 69]. The use of higher field strength has been shown to improve characterization, but many shortcomings remain [68].

CEMRA employs fast spoiled gradient-recalled echo-based sequences (FSPGR) and the paramagnetic properties of gadolinium to intensify the signal within vessels. FSPGR is used to spoil the transverse magnetization of background tissues which accentuates the marked T1 shortening of intravascular gadolinium [70]. FSPGR also employs short RF pulses allowing for fast sequence times [70]. Administration of gadolinium allows for improved signal-to-noise ratio and minimizes flow-related artifacts [70].

The saturation and spin dephasing effects that occur with TOF are of no consequence in CEMRA. It does, however, have its own limitations. Correct bolus timing is paramount to creating a diagnostic study but unfortunately can be very challenging to obtain [70]. When bolus timing is poor, significant artifact can limit evaluation of vascular pathology [71]. Artifacts including distortion of the size of the vessel, “ringing” within the vessel, and nontarget vessel enhancement can obscure pathology [70, 72]. In addition to bolus timing, the marked T1 shortening of methemoglobin can obscure contrast between vessels and background tissue, as it does in TOF images. Finally, limited spatial resolution in CEMRA makes it insensitive in the evaluation of small aneurysms (<3 mm) and in the evaluation of complex AVMs [73]. Artifacts and limitations of CTA and MRA are summarized in Table 3.