The internal carotid artery (ICA) enters the mastoid bone through the carotid canal. The vessel ascends in vertical direction, followed by a horizontal course (vertical and horizontal petrous (C2) segment) and then runs anterior to the tympanic cavity. The ICA is separated from the tympanic cavity by a 0.5 mm thick bony plate. It traverses the foramen lacerum of the sphenoid bone to enter the middle cranial fossa (C3 segment). Then the ICA traverses the cavernous sinus (C4 segment), after which it perforates the dura near the clinoid process (Fig. 1).

Some variations of the ICA should be recognized because they can influence surgical or interventional procedures [3]. An aberrant course of the ICA through the middle ear [4] should be recognized to prevent complications in middle ear surgery. Another aberrant course of the ICA is the lateral pharyngeal ICA (Fig. 2) in which the internal carotid arteries are elongated and course partially in the midline just posterior to the pharyngeal wall.

The circle of Willis provides a potential source of collateral blood flow between the carotid and the vertebrobasilar circulation in case of occlusive vascular disease. The anatomy of the circle of Willis is illustrated in Fig. 3. The ICA bifurcates into the middle cerebral artery (MCA) and the first segment of the anterior cerebral artery (ACA), also called the A1 segment. The bilateral A1 segments are connected to each other by the anterior communicating artery (Acom), after which they give rise to the A2 segments which course anteriorly. Some millimeters caudal from the top of the ICA the posterior communicating artery (Pcom) originates from the ICA and courses posteriorly to join the P1 segment of the posterior cerebral artery (PCA), which originates from the basilar artery. The circle of Willis is highly variable [5, 6] and is complete in only 20–25 % of people [7]. When incomplete, there is a potential risk of insufficient collateral blood flow in case of large vessel occlusion.

Figure 4 illustrates a well-known variant, a hypoplastic left A1 segment with filling of the left A2 via the Acom.

Most variations in the circle of Willis occur at the location of the Pcom; a complete posterior part of the circle is only present in 31 % of people [6]. Absence or hypoplasia of the Pcom is often seen. Another variant is the fetal PCA or fetal configuration of the PCA, in which the Pcom has the same caliber as the PCA and is combined with a hypoplastic P1 segment (Fig. 5) [8]. In these patients collateral blood flow to the PCA territory from the vertebrobasilar system in case of carotid occlusion is diminished or absent. This variant is even more important when combined with an ipsilateral absent A1 segment.

Another frequently encountered variation is the presence of an infundibulum of the Pcom. An infundibulum is a widening of the ICA located exactly at the origin of the Pcom. It is symmetric and should be smaller than 2 mm, and the Pcom originates from its apex. This variant should be distinguished from an aneurysm.

Normally two anterior cerebral arteries are present, one for each hemisphere. In case of an azygos anterior cerebral artery the ACA territories of both hemispheres are supplied by a single midline A2 segment (Fig. 6), which is rare and has a prevalence of 0.2–4.0 %. It is associated with aneurysms.

In case of a trifurcated ACA, the Acom gives rise to three anterior cerebral arteries, one of which probably is a persisting median callosal artery (Fig. 7). A bihemispheric ACA represents hypoplasia of one A2 segment, which only supplies the callosomarginal arteries. The contralateral A2 segment then provides the major arterial supply of the ACA territories bilaterally. A more rare persisting primitive olfactory artery [8] is illustrated in Fig. 8. Most ACA anomalies have a higher prevalence of aneurysms [9, 10].

The anterior part of the circle of Willis can also be incomplete. The Acom can be hypoplastic or absent. A hypoplastic A1 segment has a prevalence of 14 %, whereas the absence of an A1 segment has a prevalence of 2 % [6]. In a fenestration the arterial lumen divides into separate lumens that converge. In case of the basilar artery this represents a persisting non-fusion of two longitudinal neural arteries [9]. Fenestrations increase the incidence of aneurysms, probably because turbulent flow is created in the proximal and distal end of the fenestration [10]. However, in practice aneurysms are seldom found at the fenestration [11]. Intracranial arterial fenestration is more common in the vertebrobasilar arteries than in the arteries of the anterior circulation (Fig. 5) [11].

In the middle cerebral artery, duplications and fenestrations can occur, as well as an accessory MCA [8, 12]. These variants are infrequent and mostly clinically insignificant.

The venous drainage of the brain courses via the deep and the superficial venous system. The internal cerebral veins drain in the vein of Galen, which merges with the inferior sagittal sinus to form the straight sinus, which drains in the superficial venous system. The superficial venous sinuses are formed by the superior sagittal sinus, transverse sinuses, and sigmoid sinuses which drain bilaterally in the internal jugular vein. Fenestrations can also occur in the venous sinuses.

In the embryologic development of the intracranial circulation primitive carotid-basilar artery connections are present which usually regress completely [13]. Rarely, such connections remain patent through adult life and should be recognized before surgery or interventions. The trigeminal artery is the most common persisting carotid-basilar anastomosis [14]. This most cranially situated persistent carotid-basilar anastomosis has a prevalence of 0.1–0.6 % [15]. Two patterns are most commonly encountered. In the first pattern the persistent trigeminal artery connects with the upper basilar artery and supplies the posterior cerebral artery and superior cerebellar artery territories. Alternatively, one or two fetal type posterior cerebral arteries are present and the trigeminal artery supplies the posterior fossa (Fig. 9). In a variant persisting trigeminal artery a cerebellar artery originates from the ICA. A persistent trigeminal artery should be recognized before performing a Wada test to avoid infusion of barbiturates into the posterior fossa [16].

Another, more rare, persisting anastomosis is a primitive hypoglossal artery with a prevalence of 0.02–0.1 % [17]. A primitive hypoglossal artery represents a connection between the proximal cervical ICA and the basilar artery that traverses the hypoglossal canal. A persisting proatlantal intersegmental artery originates at a similar location but traverses the foramen magnum. This connection can also arise from the external carotid artery [8].

The most common variation of the vertebrobasilar system is the vertebral artery continuation as posterior inferior cerebellar artery (PICA), illustrated in Fig. 10 [19]. The reported prevalence is about 4.4 % [18]. Recognizing this variation is important before performing cerebral angiography. Duplications and fenestrations of the vertebrobasilar system can occur. In a duplication the two vessels have a different course (Fig. 11); in case of a fenestration the two lumina correspond to the same path [19].

For detailed imaging of intracranial vessels a high spatial resolution is mandatory. Therefore, the thinnest possible collimation is recommended. Imaging should be fast to minimize motion artifacts and to potentially reduce the length of the contrast bolus. Therefore fastest gantry rotation in combination with a slightly overlapping pitch should be chosen. Image quality in terms of signal-to-noise ratio is best if an overlapping helix of data is acquired. The degree of overlap of acquired image data can be set by the pitch, which is the table speed divided by the collimation. A tube voltage of 120 kV is commonly used, but kV settings of 100 kV have been proven to improve contrast-to-noise ratio (CNR). Images should be reconstructed with thin slices (e.g., 1 mm or less) with an overlap of 30 % (e.g., slice increment of 0.7). This allows the use of isotropic data for optimal post-processing (e.g., multiplanar reformation (MPR) in arbitrary directions).

To achieve optimal enhancement in CTA, the amount of iodine injected per second (iodine delivery rate (IDR) or iodine flux in gI/s) is the most important factor [20, 21].

Different concentrations of iodine can be used; typically the concentration ranges between 300 mg iodine/ml and 400 mg iodine/ml. For optimal vessel enhancement, an iodine delivery rate (IDR) of 1.6–2.0 g/s is recommended (e.g., 80–100 ml of iopromide 300 mg/ml at 6 ml/s). Contrast agent injection should be followed by a saline chaser with the same flow rate for better bolus shaping and for flushing the dead space from the site of injection to the region of interest.

Contrast media (CM) have a high viscosity dependent on temperature and osmolality. Increasing the CM temperature to body temperature prior to injection significantly decreases its viscosity [22]. Thus pre-warming CM prior to injection to body temperature is recommended at all times.

The preferred method for timing of the acquisition with respect to CM injection is bolus triggering. A region of interest (ROI) is placed in the internal carotid artery (ICA) just below the skull base. After CM injection this slice is continuously imaged and the Hounsfield units (HU) of the ROI are automatically measured. The scanner automatically starts scanning when the Hounsfield units exceed a certain threshold (mostly 120 HU). This method results in the best timing and optimal arterial enhancement. Adequate training of technicians is important, as manual scan start might be required if patient movement has occurred or the ROI was difficult to place due to anatomical variations.

Alternatively, a test bolus can be used. A small amount of CM (e.g., 20 ml) is injected at the same rate as the original bolus, and then a slice through the carotid arteries is scanned continuously. The time between injection and contrast arrival in the carotid arteries is measured. This delay is then used for the actual scan. The advantage of this method is its robustness; it is tailored to the individual patient and his or her cardiac output. However, additional CM is used, which does not contribute to the imaging.

CT imaging is also suitable to image the venous structures of the brain. Venous sinus thrombosis can be detected with CT as well as with MRI with equal sensitivity [23, 24]. For the detection of venous sinus thrombosis CT imaging is performed 40 s after the start of contrast injection. For venous imaging the injection speed can be lower than for arterial imaging, e.g., injection speed 3 ml/s, IDR 0.9 g/s followed by a saline chaser of 40 ml with 3 ml/s. The whole vertex should be included in the scanning volume to depict the whole superior sagittal sinus. Bone subtraction can be used to calculate maximum intensity projections (MIPs) of venous structures [25]; however source images should also be viewed to prevent misdiagnosis caused by bone subtraction artifacts due to the close relation of the dural sinuses with surrounding bone.

Dual-energy CT (DECT) is nowadays available on selected scanners. Scanning with two different energies simultaneously makes it possible to decompose materials with high and low atomic number by their attenuation differences for different X-ray energies [26]. In atoms with a low atomic number, Compton scatter is the main interaction with diagnostic X-ray photons. For atoms with a high atomic number, the photoelectric effect plays the main role. If an X-ray photon has slightly more energy than the k-shell electron-binding energy of an atom, photoelectric absorption can take place, which leads to a sudden increase in X-ray attenuation at this energy (k-edge) [27]. The k-shell electron is then ejected and its vacancy is filled up by adjacent electrons and characteristic X-rays are emitted [28]. The k-edges of iodine (33.2 keV) and calcium (4 keV) are sufficiently different from the k-edges of the constituents of soft tissue (0.01–0.53 keV) to be detected by dual-energy imaging. One of the energies is chosen at a low value, mostly 80 kV. The other energy is chosen higher, e.g., 140 kV. The spectrum of energies emitted by the X-ray source always forms a Gaussian bell curve, so also photons around 33 keV are present in an 80 kVp (kilo Volt peak) bundle. These are strongly attenuated because they approximate the k-edge of iodine. Based on the DECT-data iodine, calcium and hemorrhage can be discriminated. The characterization of iodine and calcium can be used for bone removal [27]. The characterization of iodine and hemorrhage can be used to calculate iodine only and virtual non-contrast images and can be used to differentiate CM from hemorrhage, e.g., after interventional procedures and for the analysis of intracranial hemorrhage [29–33]. Alternatively, DECT data can be used to calculate mono-energetic reconstructions, which can be applied for reduction of beam-hardening artifacts, for instance around dental restorations [34] and metallic implants [35]. The first potential applications of such reconstructions for angiographic applications have been reported for pulmonary CTA [36] and CTA of abdomen and extremities [37]. CNR for angiographic images can be increased for low kV reconstructions, potentially allowing for the use of less contrast agent [38, 39], which can also be applied in cranial CTA [29].

Bone subtraction can be performed to simplify vessel analysis. Bone subtraction can be performed in various ways. First, two scans can be performed (non-contrast enhanced–contrast enhanced) which are subtracted from each other [40, 41]. Secondly, various vendors developed software for threshold-based bone subtraction. A third alternative is DECT based bone subtraction [42–44].

Caution should be taken, as the degree of stenosis can be overestimated due to blooming artifacts or because of the suboptimal distinction of calcification and iodine due to poor enhancement of the artery [44]. Furthermore, segmentation errors may occur, especially at the skull base. Thus bone removal techniques can only be an add on—viewing of the source images, including post-processing, is always recommended.

Dynamic evaluation of the vessels is possible with time-resolved CTA, a relatively new application, also called time-invariant CTA, dynamic CTA, or 4D-CTA [45, 46, 47]. It has the potential benefit to demonstrate collateral vessels and drainage flow, next to the depiction of the cranial vessels [47, 48].

With the newest MDCT scanners, allowing for a large coverage in the z-direction or a back-and-forth sliding table (shuttle mode), dynamic CTA with high temporal resolution has become possible [49, 50]. The first applications of dynamic CTA (4D-CTA) for imaging of arteriovenous malformations have been described recently [46, 51–53]. Dynamic evaluation of the vessels is also described in the evaluation of ischemic stroke patients [50, 54, 55] and in venous occlusions [48].

Scanning with thin collimation and reconstructing thin overlapping slices allow for possibilities of reconstruction in arbitrary angles without sacrificing spatial resolution. Scan thin, review thick is the basic rule [56]. Several post-processing techniques are available.

Stroke, “a cerebrovascular accident (CVA),” is an acute neurological deficit.

Most strokes (80 %) are due to an ischemic event. Twenty percent of strokes are hemorrhagic.

The etiology of stroke varies with the stroke subtype [61] as well as its outcome and recurrence risks.

Etiological mechanisms are atherosclerotic strokes, small vessel disease, cardioembolic disease, or undetermined/other [62].

The literature is based mainly on extracranial atherosclerotic disease, located at the carotid bifurcation; less literature focuses on intracranial atherosclerotic disease. In proximal/extracranial atherosclerosis the vulnerable plaque with hemorrhage, ulceration, and calcification of plaques is frequently found, whereas in intracranial luminal stenosis of the middle cerebral artery the percentage of lipid area and the presence of intraplaque neovascularity probably plays a key role leading to ischemic stroke [68]. In a Chinese population the presence of intracranial calcifications was an independent risk factor of ischemic stroke [69].

The mechanisms of stroke in intracranial atherosclerosis can be divided into local thrombotic occlusion, artery-to-artery embolism, branch occlusive disease of the orifice of deep perforating arteries, hemodynamic (hypoperfusion), or a combination of these [70]. This can explain the different patterns of infarcts.

In situ thrombotic occlusion leads to the largest infarctions. Intracranial atherosclerosis usually is a slowly developing process, so collateral circulation can develop in chronic atherosclerotic disease. The size of the infarction is dependent on the presence of collateral circulation, the speed of the occlusion, and hemodynamics. Whole territory infarcts are therefore less frequent.

In artery-to-artery embolism emboli from the more proximal occlusion can be entrained with the bloodstream to distal branches and result in an occlusion with large strokes. Examples are thrombi from carotid bifurcation plaques and cardiac emboli (Fig. 12). The latter account for larger strokes due to larger emboli and less collateral circulation. Intracranial MCA stenosis can give artery-to-artery emboli to distal M2 segments.

Occlusions of the origin of a perforating artery can be due to an intracranial atherosclerotic plaque. The pathophysiology of these occlusions differs from the small vessel disease, but can lead to the same small infarctions (Fig. 12).

Imaging in acute stroke can be divided into differentiating hemorrhagic stroke and ischemic stroke, imaging of vessel occlusion (Figs. 12, 13, 15, and 17), and determining the core and penumbra of the infarction (Figs. 14 and 16).

In the acute setting a non-contrast-enhanced CT (NECT) of the brain can reveal hemorrhage. The ASPECTS score [71, 72] can be used to try to determine the extent of the infarction in the initial phase. Early signs of ischemic stroke are effacement of sulci, hypodensity of the brain parenchyma, loss of gray-white matter differentiation, and in the insular region called the “insular ribbon sign” and a hyperdense vessel sign [73] (Fig. 18).

Defining irreversibly damaged parts of the brain can be done by MRI with diffusion-weighted imaging (DWI) or with CT perfusion (Fig. 14). Vascular imaging can be performed with CTA [74], MRA [75], or DSA; the latter is the gold standard, but is preferably used in selected cases [74]. CTA can also be calculated from the VPCT dataset, thereby lowering the radiation dose [50, 76]. In case of an inconclusive CTA or MRA an adjunct MRA or CTA can lower the need for DSA [77].

The accuracy of CTA for evaluation of intracranial atherosclerotic disease is good, especially in stenosis larger than 50 % [74, 78], and is increasing with newer and more sophisticated scanners [79]. CTA can show the site and severity of the stenosis, the length of the thrombus, and the presence of collateral circulation [80] (Fig. 15). Time-resolved CTA or 4D-CTA (Fig. 19) can give additional information about the leptomeningeal collateral filling [45, 47] or the clot burden [54] (Fig. 16).