9. Neurovascular Anatomy

The arterial supply of the brain is best conceptualized not as a set of independent end-arteries but as a redundant, anastomotic network whose central node — the circle of Willis — exists precisely to equalize pressure between two carotid and one basilar inflow source. On CT angiography, the circle is resolved at the level of the suprasellar cistern, where contrast-opacified vessels appear as bright (typically 250–400 HU at peak arterial timing) tubular structures suspended in the cerebrospinal-fluid–attenuation cistern, an inherently high-contrast situation that makes the CSF a natural dark backdrop. The polygon is assembled from the terminal internal carotid arteries laterally, the proximal (A1) anterior cerebral arteries joined across the midline by the single anterior communicating artery (AcomA), and posteriorly the proximal (P1) posterior cerebral arteries connected to the carotid circulation by the paired posterior communicating arteries (PcomA). The cardinal teaching point, repeatedly validated in autopsy and CTA series, is that a textbook-complete, symmetric circle is present in only a minority of individuals — on the order of a third to a half — so that the expert reader treats asymmetry as the expected state and asks instead whether a given asymmetry is hemodynamically meaningful. The most consequential variants are A1 hypoplasia or aplasia (which makes both anterior cerebral territories dependent on one carotid through the AcomA, a configuration that concentrates shear at the AcomA and underlies its predilection for aneurysm), the fetal posterior cerebral artery (in which the PCA fills predominantly or entirely from the carotid via a dominant PcomA, leaving the occipital lobe carotid-dependent and therefore vulnerable in anterior-circulation embolism while paradoxically protected from basilar disease), and a hypoplastic or absent PcomA that closes off a primary collateral route.

Each major vessel must then be tracked into its territory and, critically, its perforators. The anterior cerebral artery, after the AcomA, courses around the genu of the corpus callosum to supply the medial frontal and parietal cortex and the anterior corpus callosum; its single most important perforator is the recurrent artery of Heubner, arising near the AcomA to supply the anteroinferior caudate, anterior limb of the internal capsule, and anterior globus pallidus — its occlusion produces a face-and-arm-predominant weakness with prominent dysarthria and, sometimes, abulia. The middle cerebral artery is the largest terminal branch and the most frequent site of large-vessel occlusion; its horizontal M1 segment gives off the lenticulostriate perforators, end-arteries that supply the bulk of the basal ganglia, internal capsule, and corona radiata and that lack collateral redundancy, which is why a proximal M1 embolus that occludes their origins produces a dense, eloquent striatocapsular infarct, and why chronic hypertensive lipohyalinosis of these same vessels yields lacunar infarction and the hypertensive hemorrhages of the putamen. The MCA bifurcates at the limen insula into superior and inferior M2 trunks supplying the lateral convexity; the 'hyperdense MCA sign' — acute thrombus appearing as intraluminal high attenuation (often 50–90 HU on non-contrast CT) — and its distal counterpart, the 'dot sign' in the Sylvian fissure, are the earliest CT evidence of occlusion and a high-yield, frequently missed finding. The posterior cerebral artery supplies the occipital lobe, inferomedial temporal lobe, and splenium, and gives thalamoperforators (including, when a single P1 perforator supplies both medial thalami, the artery of Percheron, whose occlusion causes the classic bilateral paramedian thalamic infarct with altered consciousness). The vertebrobasilar system — paired vertebral arteries uniting at the pontomedullary junction to form the basilar — supplies the brainstem, cerebellum (via PICA, AICA, and SCA), and, terminally, the PCAs; its perforators are unforgiving median and paramedian end-arteries, so basilar perforator occlusion produces devastating pontine syndromes. The disciplined CTA search pattern follows the column from the aortic arch and cervical vertebral/carotid origins upward, interrogates each circle-of-Willis segment for caliber and opacification, and deliberately revisits the posterior circulation and distal branches, because the dominant cognitive errors here are satisfaction of search after the first occlusion is found and inattentional neglect of the brainstem perforators that thin-section maximum-intensity projections can obscure.

The cerebral venous system is, for the diagnostician, the mirror image of the arterial system in both anatomy and error profile: it is more variant, more asymmetric, and lower in attenuation, and its principal clinical danger on CT is not under-detection of disease but the over-diagnosis of thrombosis from normal variants and physical artifacts. Functionally the system is divided into a superficial (cortical) compartment draining the convexity cortex and a deep compartment draining the periventricular white matter and deep gray nuclei, both converging on the dural venous sinuses — rigid, endothelium-lined channels enclosed between the periosteal and meningeal dural layers. On contrast-enhanced CT venography the sinuses opacify densely; on non-contrast CT a normal sinus may measure 50–70 HU and, when the hematocrit is high or the study is performed in a dehydrated or polycythemic patient, can approach the 60–80 HU range that mimics acute thrombus, the so-called pseudo-dense-sinus or hematocrit effect — a leading cause of false-positive interpretation that is resolved by comparison with arterial attenuation, by measuring the venous-to-arterial ratio, and by correlating with the clinical hematocrit. The superior sagittal sinus runs in the falx from the crista galli to the torcular Herophili (confluence of sinuses), receiving the superficial cortical veins; these cortical veins are notoriously variable, but two are named and worth knowing — the superficial middle cerebral (Sylvian) vein along the Sylvian fissure, and the anastomotic veins of Trolard (superior, connecting to the superior sagittal sinus) and Labbé (inferior, connecting to the transverse sinus), whose patency determines whether a sinus occlusion is clinically compensated. From the torcular, flow divides into the paired transverse sinuses, and here the single most important normal-variant trap arises: the transverse sinuses are frequently asymmetric, the right typically dominant, and a hypoplastic or atretic non-dominant transverse sinus — often accompanied by a high-splitting torcular or a focally narrowed segment — can be misread as thrombosis unless the reader recognizes the smooth tapering caliber, the corresponding asymmetry of the adjacent jugular foramen, and the absence of an intraluminal filling defect on CTV.

Arachnoid granulations compound this difficulty: these are normal CSF-filled or fibrous protrusions of arachnoid into the sinus lumen, appearing as well-defined, rounded, CSF- or soft-tissue-attenuation filling defects (typically a few millimeters, classically in the lateral transverse sinuses and the superior sagittal sinus) that must be distinguished from thrombus by their characteristic ovoid morphology, smooth margins, and stable appearance. The deep venous system is anatomically more constant and therefore more diagnostically reliable. The subependymal and medullary veins of the white matter drain centripetally into the paired internal cerebral veins, which run posteriorly in the roof of the third ventricle within the velum interpositum; the basal veins of Rosenthal sweep around the midbrain from the anterior perforated substance; and these unite at the level of the pineal region to form the great cerebral vein of Galen, a short, thick midline trunk that joins the inferior sagittal sinus to become the straight sinus, which courses to the torcular. The constancy of this deep architecture is precisely what makes deep venous thrombosis recognizable, because its territory — the medial thalami, basal ganglia, and periventricular white matter — produces a distinctive bilateral, often hemorrhagic, edema pattern that does not respect any single arterial territory and should immediately trigger interrogation of the internal cerebral veins, vein of Galen, and straight sinus for the dense-vein sign. The expert search pattern therefore explicitly includes the venous structures on every head CT — tracing the superior sagittal sinus on the midline sagittal reformation, comparing transverse-sinus calibers with attention to the jugular foramina, and inspecting the deep veins on the axial study — because the dominant cognitive failure in cerebral venous thrombosis is not difficulty seeing the finding but failure to look for it at all, an inattentional blindness reinforced by the protean, frequently non-territorial, and sometimes hemorrhagic presentation that anchors the reader on arterial or parenchymal explanations.

Collateral circulation is the physiological bridge between the static anatomy of the preceding sections and the dynamic, time-dependent reasoning of acute stroke imaging, and it is the single most important determinant of whether an arterial occlusion produces a small, survivable infarct or a malignant, hemispheric one. Collaterals are conventionally stratified into primary and secondary pathways by the order in which they are recruited as cerebral perfusion pressure falls. The primary collaterals are the pre-existing, large-caliber communications of the circle of Willis itself — the anterior communicating artery, which equalizes flow between the two anterior circulations, and the posterior communicating arteries, which couple the carotid (anterior) to the vertebrobasilar (posterior) circulation. These channels are instantaneously available, require no recruitment time, and are the reason an internal carotid occlusion can be clinically silent in a patient with a complete circle; conversely, their congenital incompleteness — the hypoplastic A1, the absent PcomA, the isolated configurations discussed earlier — converts an otherwise compensable occlusion into a symptomatic infarct, which is the deep clinical significance of the circle-of-Willis variants and the reason they must be reported, not merely noted. When the primary collaterals are absent or exhausted, the secondary collaterals are recruited: the leptomeningeal (pial) anastomoses that connect the distal cortical branches of the anterior, middle, and posterior cerebral arteries across the watershed border zones, the ophthalmic artery (which can reverse flow to supply the carotid territory retrogradely from the external carotid through the orbit), and dural-to-pial connections. These secondary routes are smaller, must dilate and recruit over minutes to hours, and are exquisitely sensitive to perfusion pressure, systemic blood pressure, and collateral 'reserve' — which is why permissive hypertension and avoidance of hypotension are central to acute stroke management, and why the leptomeningeal watershed zones (the cortical ACA–MCA and MCA–PCA borders, and the internal subcortical watershed of the corona radiata) are the territories that infarct first when global perfusion fails, producing the recognizable parasagittal and centrum-semiovale watershed patterns.

The imaging of collaterals is now quantitative and outcome-defining. On multiphase CT angiography the reader grades the extent and timing of pial vessel filling distal to an occlusion (commonly on an ordinal 0–5 collateral scale), and on CT perfusion the collateral state is read indirectly through the relationship between the truly infarcted core and the salvageable penumbra. The operational definitions used in 2026 endovascular-selection practice rest on transit-time and flow thresholds: the ischemic core is approximated by relative cerebral blood flow below roughly 30 percent of the contralateral normal, $\mathrm{rCBF} < 30\%$, while the penumbra — hypoperfused but still viable tissue maintained by collateral flow — is the volume with prolonged time-to-maximum, conventionally $T_{max} > 6\,\mathrm{s}$, that lies outside the core. The mismatch that defines a reperfusion candidate is then

$\text{Penumbra} = V_{T_{max} > 6s} - V_{\mathrm{rCBF} < 30\%},$

and a favorable mismatch ratio (penumbra-to-core, typically $\geq 1.8$ with an absolute mismatch volume $\geq 15\,\mathrm{mL}$ and a core below a threshold near $70\,\mathrm{mL}$) is what extends the thrombectomy window into the 6–24 hour period in selected patients. Underlying these numbers is autoregulation: across a plateau of perfusion pressure the brain holds cerebral blood flow constant by adjusting arteriolar resistance, $\mathrm{CBF} = \mathrm{CPP}/\mathrm{CVR}$, so that good collaterals effectively raise the local perfusion pressure and keep tissue on the autoregulatory plateau, slowing the conversion of penumbra to core and 'buying time' for reperfusion. The expert therefore reads collateral status as a prognostic and therapeutic variable, not a descriptive aside: robust pial collaterals predict slow infarct growth, a large mismatch, and benefit from late reperfusion, whereas poor collaterals predict a rapidly completing infarct and a malignant course. The dominant cognitive pitfalls are anchoring on the proximal occlusion while failing to grade the distal collateral filling, and over-relying on a single arterial-phase CTA acquisition whose timing can spuriously suggest poor collaterals when the contrast bolus has simply not yet reached slow-filling pial vessels — the specific error that multiphase or delayed acquisitions were designed to prevent.