20. Neurocritical Care CT

Intracranial pressure is governed by the Monro-Kellie doctrine: the cranium after fusion is a rigid box of fixed volume (~1400-1700 mL) containing brain parenchyma (~80%), blood (~10%), and cerebrospinal fluid (~10%), and because the contents are nearly incompressible, any increase in one compartment must be matched by an equal decrease in another or pressure rises. The relationship is not linear but exponential: $\mathrm{ICP} = P_0\,e^{kV}$, so that early volume loading is buffered—chiefly by displacement of CSF into the spinal thecal sac and by collapse of the low-pressure venous compartment—and pressure stays near the normal $7$–$15\,\mathrm{mmHg}$ along the flat part of the curve, but once this reserve is exhausted the curve turns sharply vertical and a small additional volume produces a large pressure jump. This is the mechanistic reason a patient can tolerate a slowly growing meningioma with a normal examination yet a smaller acute hematoma decompensates abruptly: it is compliance, not absolute volume, that has run out. Cerebral perfusion pressure, the driver of flow, is $\mathrm{CPP} = \mathrm{MAP} - \mathrm{ICP}$, so rising ICP directly throttles perfusion, and when CPP falls the Cushing reflex (hypertension, bradycardia, irregular respiration) signals brainstem compromise.

CT cannot measure pressure directly; it reads the structural consequences of spent compliance, and the basal cisterns are the most prognostically loaded structure on the scan. The reader interrogates, in order of lethality, the basal cisterns (suprasellar, interpeduncular, ambient, quadrigeminal, prepontine) for effacement, then the cortical sulci for crowding, then the ventricles for compression or trapping, then the gray-white junction for the diffuse blurring of global edema. Effacement of the perimesencephalic cisterns is the imaging fingerprint of transtentorial pressure and carries the steepest mortality gradient in the Marshall and Rotterdam traumatic-brain-injury classifications, where absent cisterns and midline shift greater than $5\,\mathrm{mm}$ independently predict death. Chronic elevation leaves a different signature—an empty or partially empty sella, posterior globe flattening with optic-nerve-sheath distension and protrusion of the optic papilla, vertical tortuosity of the optic nerves, and slit-like ventricles in idiopathic intracranial hypertension—whereas acute elevation manifests as the cisternal and sulcal effacement above. A useful quantitative anchor is the optic nerve sheath diameter measured $3\,\mathrm{mm}$ behind the globe, which correlates with invasive ICP and exceeds roughly $5.7$–$6.0\,\mathrm{mm}$ when pressure is elevated.

The differential for the CT picture of raised ICP is approached by Bayesian updating from the clinical context. In trauma, effaced cisterns with shift point to an evacuable mass or malignant swelling; in a known cirrhotic with hyperammonemia, diffuse swelling suggests fulminant hepatic edema; in the post-arrest patient, the loss of gray-white differentiation with the pseudo-subarachnoid sign (apparent hyperdensity in effaced cisterns produced by the contrast between dark edematous brain and relatively dense vessels) indicates global hypoxic-ischemic injury and a grim prognosis. The pseudo-SAH sign is the cardinal mimic: it must not be read as true subarachnoid blood, and the distinction rests on Hounsfield measurement (true blood is $50$–$70\,\mathrm{HU}$; pseudo-SAH measures lower, typically $30$–$40\,\mathrm{HU}$, and tracks with diffuse low parenchymal density) and on the diffuse, non-aneurysmal distribution. Management is dictated by the curve: the scan that shows effaced cisterns mandates immediate measures—head elevation, osmotherapy (hypertonic saline or mannitol), and, when a mass or hydrocephalus is the cause, surgical decompression or CSF diversion—because the patient is on the vertical limb where minutes matter. The failure modes are technical (motion and beam-hardening that blur the cisterns and counterfeit or conceal effacement) and cognitive (anchoring on a conspicuous hematoma and failing to register that already-effaced cisterns signal imminent herniation, the satisfaction-of-search error that turns a survivable scan into a missed emergency).

Cerebral edema is an increase in brain water, and the interpretive task is to classify it by mechanism because mechanism dictates compartment, CT appearance, distribution, reversibility, and treatment. Four categories are distinguished. Cytotoxic (cellular) edema follows energy failure: when ischemia or hypoxia collapses the $\mathrm{Na}^+/\mathrm{K}^+$-ATPase, sodium and water move intracellularly, cells swell, and the diffusion of water becomes restricted; on CT this manifests as loss of gray-white differentiation and sulcal effacement with a modest fall in attenuation, because edematous tissue is more watery and water sits near $0\,\mathrm{HU}$—each roughly $1\%$ increase in tissue water lowers attenuation by approximately $2$–$3\,\mathrm{HU}$, which is why the early infarct darkens by only a few Hounsfield units and demands a narrow window to detect. Cytotoxic edema characteristically respects vascular territory and the gray-white boundary is lost because the cortex, with its higher baseline blood volume and metabolic rate, swells first. Vasogenic edema arises from blood-brain-barrier breakdown, with protein-rich plasma filtrate accumulating in the extracellular space and tracking preferentially along white-matter tracts; on CT it is finger-like, low-density, spares the overlying cortex, and surrounds tumors, abscesses, and contusions. Interstitial (transependymal) edema is pressurized CSF forced across the ependyma in acute hydrocephalus, producing a smooth periventricular halo. Osmotic edema follows a fall in serum osmolality (hyponatremia, dialysis disequilibrium) that drives water down its gradient into the brain diffusely.

The quantitative and temporal evolution is decisive for the neurocritical reader. In ischemic stroke the cytotoxic edema is detectable within hours as subtle hypoattenuation, becomes frankly hypodense and space-occupying over $24$–$72\,\mathrm{h}$, and in the malignant middle-cerebral-artery syndrome produces a swollen hemisphere that effaces the ipsilateral ventricle and drives midline shift—pineal-level shift exceeding $5\,\mathrm{mm}$ within $48\,\mathrm{h}$ predicts fatal herniation, and the pooled DESTINY/DECIMAL/HAMLET data established that decompressive hemicraniectomy within $48\,\mathrm{h}$ reduces mortality and severe disability, making the serial CT measurement of shift a direct surgical trigger. Tumoral vasogenic edema is exquisitely steroid-responsive, whereas cytotoxic edema is not—a discriminator with therapeutic teeth. In posterior reversible encephalopathy syndrome the edema is vasogenic, posterior-predominant (parieto-occipital), usually symmetric, and reversible with blood-pressure control, and its recognition prevents the misdiagnosis of bilateral PCA infarction.

The ranked differential and its Bayesian logic: territorial gray-white loss in a vascular distribution is cytotoxic infarction until proven otherwise; finger-like white-matter low density sparing cortex and surrounding an enhancing mass is vasogenic; a smooth periventricular rim with enlarged ventricles is interstitial; diffuse loss of differentiation across both hemispheres with a low serum sodium is osmotic; and posterior-predominant symmetric vasogenic edema in a hypertensive, eclamptic, or calcineurin-inhibitor-exposed patient is PRES. The pretest context moves the posterior sharply—the same parieto-occipital edema is PCA infarction in an embolic source and PRES in a patient spiking to $220\,\mathrm{mmHg}$. The mimics are the chronic small-vessel leukoaraiosis that counterfeits vasogenic edema (but lacks mass effect) and the dilated perivascular spaces that mimic lacunar edema. Failure modes include the technical trap of a wide window erasing the few-Hounsfield-unit cytotoxic change, and the cognitive trap of attributing early malignant infarct swelling to artifact and thereby missing the window for hemicraniectomy.

Herniation is the mechanical displacement of brain from one dural compartment into another once a pressure gradient overwhelms regional compliance, and each pattern has a defined anatomic pathway, a specific CT signature, and a brainstem consequence that the reader must name because the syndrome, not the inciting mass, is what kills. Subfalcine (cingulate) herniation is the commonest and earliest: an expanding supratentorial mass pushes the cingulate gyrus beneath the free edge of the falx, and CT shows shift of the septum pellucidum and the falcine structures across the midline. Shift is quantified at two landmarks—the septum pellucidum anteriorly and the pineal gland posteriorly—with the convention that displacement greater than $5\,\mathrm{mm}$ is clinically significant and worsens prognosis stepwise; the danger is not the shift itself but compression of the ipsilateral foramen of Monro (trapping and dilating the contralateral lateral ventricle) and, critically, stretching and compression of the anterior cerebral artery against the falx, producing a secondary medial frontal infarct.

Descending transtentorial herniation has two forms. Uncal (lateral) herniation drives the medial temporal lobe (uncus and parahippocampal gyrus) medially over the tentorial incisura into the suprasellar and ambient cisterns; the CT hallmarks are effacement and asymmetric widening of those cisterns, with the classic clinical triad of an ipsilateral blown pupil (third-nerve compression as it runs in the incisura), contralateral hemiparesis, and decreased consciousness. Two named consequences are testable and lethal: compression of the posterior cerebral artery against the tentorial edge causes occipital (PCA-territory) infarction, and contralateral displacement of the midbrain crushes the opposite cerebral peduncle against the tentorium—the Kernohan notch phenomenon—producing a paradoxical hemiparesis ipsilateral to the mass and a false-localizing sign the unwary reader and clinician misattribute. Central (axial) transtentorial herniation is the symmetric downward displacement of the diencephalon and midbrain through the incisura from diffuse swelling or bifrontal mass; CT shows symmetric effacement of the perimesencephalic and suprasellar cisterns, inferior displacement of the pineal, and effacement of the third ventricle, and as the brainstem is forced caudally the penetrating paramedian basilar perforators are stretched and shear, producing the Duret hemorrhages—small midline hyperdensities in the rostral pons and midbrain tegmentum that are a near-terminal, irreversible finding.

Upward transtentorial herniation reverses the direction: a posterior-fossa mass pushes the superior vermis and cerebellum up through the incisura, effacing the quadrigeminal cistern (the smooth, flattened or obliterated cistern is the sign), compressing the tectum and aqueduct (causing obstructive hydrocephalus), and kinking the PCAs and superior cerebellar arteries. Tonsillar herniation drives the cerebellar tonsils down through the foramen magnum, crowding the cervicomedullary junction; CT shows effacement of the cisterna magna and tonsils projecting below the foramen magnum, and the consequence is medullary compression with respiratory and cardiovascular collapse—the reason a strategically placed cerebellar lesion in the rigid posterior fossa is a neurosurgical emergency out of proportion to its volume. External (transcalvarial) herniation through a craniectomy defect or fracture is iatrogenic or traumatic and is read directly as parenchyma bulging beyond the inner table. The Bayesian search is direction-aware: a supratentorial mass mandates inspection of the septum, foramen of Monro, suprasellar/ambient cisterns, and PCA territory; a posterior-fossa mass mandates the quadrigeminal cistern, fourth ventricle, and foramen magnum. Mimics include congenital low-lying tonsils (Chiari I, without acute cisternal effacement) and benign asymmetry of the temporal horns. The failure modes are technical (posterior-fossa beam-hardening obscuring tonsillar position and Duret hemorrhage) and cognitive (the false-localizing Kernohan notch and PCA infarct leading to wrong-side or wrong-cause reasoning, and premature closure on the mass without staging the herniation it has produced).

Hydrocephalus is an active distension of the ventricular system by cerebrospinal fluid resulting from a mismatch between production and circulation/absorption, and on CT it is the upstream dilatation that localizes the obstruction and the periventricular halo that timestamps its acuity. CSF is produced principally by the choroid plexus at roughly $20\,\mathrm{mL/h}$ (~$500\,\mathrm{mL/day}$) against a total volume of ~$150\,\mathrm{mL}$, so the system turns over several times daily and any obstruction manifests rapidly as proximal enlargement. The cardinal interpretive distinction is obstructive (non-communicating) versus communicating hydrocephalus, and it is made by reading the pattern of dilatation against the serial anatomy of the ventricular pathway. In obstructive hydrocephalus the block is within the ventricular system and the spaces proximal to it dilate while those distal stay normal or are effaced: a colloid cyst or intraventricular hemorrhage at the foramen of Monro dilates one or both lateral ventricles with a normal third; aqueductal stenosis or a tectal/pineal mass dilates the lateral and third ventricles with a normal fourth (the single most localizing pattern on the scan); and a fourth-ventricular or posterior-fossa mass dilates all four ventricles by obstructing the outlet. Communicating hydrocephalus, by contrast, follows impaired subarachnoid CSF resorption at the arachnoid granulations—post-subarachnoid-hemorrhage, post-meningitic, or carcinomatous—and dilates the entire ventricular system, fourth ventricle included, often with relative effacement of the cortical sulci.

The quantitative anchors are the temporal horns and the third ventricle, and the acuity marker is transependymal flow. Normally slit-like temporal horns that become rounded and visible are an early, sensitive sign of obstruction, often appearing before frontal-horn ballooning; the frontal horns enlarge and the normally concave caudate-indented margin becomes convex, and the third ventricle loses its slit shape and bows. The Evans index—the ratio of the maximal width of the frontal horns to the maximal inner-table diameter at the same level—exceeding $0.30$ is the conventional cutoff for ventriculomegaly, and the callosal angle (measured on coronal reformats at the posterior commissure) helps separate normal-pressure hydrocephalus (narrowed, $<90^{\circ}$) from atrophy. The decisive sign of acute decompensation is interstitial (transependymal) edema: a smooth, low-density periventricular halo, most marked capping the frontal horns, where pressurized CSF is forced across the ependyma—its presence converts ventriculomegaly into an emergency mandating CSF diversion.

The ranked differential and its mimics: ventriculomegaly with a transependymal halo and effaced sulci is acute hydrocephalus and is treated; ventriculomegaly with proportionate sulcal enlargement is ex-vacuo (atrophic) and is not; ventriculomegaly out of proportion to sulci, with a narrow callosal angle and dilated Sylvian fissures but without a transependymal gradient, in an elderly patient with gait apraxia, urinary incontinence, and cognitive decline is normal-pressure hydrocephalus, where the imaging supports shunt candidacy. The Bayesian leverage is large: identical ventricular size means a shunt in the patient with a halo and a posterior-fossa mass, and no shunt in the patient with diffuse atrophy. The mimics are atrophy (the chief confounder), benign enlargement of the subarachnoid spaces in infancy, and the dilated but non-obstructed ventricle of holoprosencephaly or agenesis. The failure modes are technical—axial-only acquisition that hides the callosal angle and underestimates temporal-horn rounding—and cognitive: attributing genuine acute hydrocephalus to atrophy (and discharging a patient who needs a drain), overlooking the colloid cyst or intraventricular clot at the foramen of Monro that produces sudden, sometimes positional, obstruction, and failing to register the transependymal halo as the line between chronic and emergent.

Vasospasm and the broader syndrome of delayed cerebral ischemia are the leading preventable cause of death and disability in the days after aneurysmal subarachnoid hemorrhage, and their imaging detection is a defined neurocritical-care task with named thresholds and a guideline-fixed temporal window. The pathophysiology is initiated by subarachnoid blood: spasmogenic products of hemolysis (notably oxyhemoglobin) and the breakdown of nitric-oxide signaling, together with cortical spreading depolarizations, microthrombosis, and impaired autoregulation, narrow the large conductance arteries and compromise the microcirculation. The risk is dose-dependent on the volume and distribution of cisternal blood, which is why the modified Fisher scale—graded on the initial noncontrast CT—is the reader's first prognostic instrument: grade 1 is thin focal/diffuse subarachnoid blood without intraventricular hemorrhage, grade 2 thin subarachnoid blood with intraventricular hemorrhage, grade 3 thick subarachnoid blood without intraventricular hemorrhage, and grade 4 thick subarachnoid blood with intraventricular hemorrhage, with the incidence of symptomatic vasospasm rising stepwise across the scale (the modified scale corrected the paradoxical non-monotonic risk of the original Fisher grading by explicitly accounting for intraventricular blood). The crucial temporal fact is that angiographic vasospasm is uncommon before day 3, peaks at days $7$–$10$, and resolves by days $14$–$21$, so the reader interprets every delayed scan in this window against the day of ictus.

The imaging is layered. CT angiography demonstrates the large-vessel narrowing directly and is graded by the percentage reduction in luminal diameter relative to the adjacent normal segment—mild ($<25\%$), moderate ($25$–$50\%$), and severe ($>50\%$)—with severe proximal narrowing carrying the highest ischemic risk; CTA is sensitive for proximal, severe spasm but less reliable for distal and mild disease, and digital subtraction angiography remains the standard when intra-arterial rescue therapy is contemplated. Because large-vessel caliber correlates imperfectly with tissue fate, CT perfusion adds the decisive physiologic layer: delayed cerebral ischemia is fundamentally a perfusion diagnosis, and a prolonged mean transit time (the most sensitive parameter, commonly using thresholds such as $\mathrm{MTT} > 6.4\,\mathrm{s}$) with regionally reduced cerebral blood flow (approaching the ischemic range of $<25$–$30\,\mathrm{mL}/100\,\mathrm{g}/\mathrm{min}$) identifies tissue at risk before infarction is fixed. The 2023 AHA/ASA guideline and the Vergouwen multidisciplinary consensus define delayed cerebral ischemia clinically as a new focal deficit or a drop of at least two points on the Glasgow Coma Scale lasting at least one hour and not attributable to another cause; imaging is deployed to confirm the substrate and exclude mimics, and the diagnosis explicitly does not require angiographic spasm, because microcirculatory failure can produce DCI with patent large vessels.

The ranked differential for delayed neurologic decline after aSAH—and the Bayesian reasoning—runs: in the day $4$–$14$ window with a high modified Fisher grade, vasospasm/DCI is the prior-dominant diagnosis and is confirmed by territorial perfusion delay with corresponding large-vessel narrowing; but the reader must actively exclude rebleeding (new or expanded hyperdense subarachnoid/intraparenchymal blood), acute hydrocephalus (ventricular enlargement with a transependymal halo), seizure (often with hyperperfusion rather than hypoperfusion), metabolic derangement, and the iatrogenic confounders of sedation and hyponatremia. The pretest probability set by blood burden and post-bleed day moves the posterior decisively—the same hemiparesis is far more likely DCI on day 8 of a Fisher 4 hemorrhage than on day 1. Management hinges on detection: induced normovolemic hypertension to augment CPP, and, for refractory focal deficits with a vascular correlate, endovascular rescue with intra-arterial vasodilators or balloon angioplasty—oral nimodipine improves outcome and is given to all, though it does not reliably reverse angiographic narrowing. The mimics are perfusion artifact from motion or poor bolus timing and the chronic hypoperfusion of prior infarct. The failure modes are technical (truncated bolus and motion degrading CT perfusion maps, and CTA overcalling distal spasm or missing it) and cognitive (attributing a new deficit to sedation and missing evolving DCI within its predictable window—the anchoring and premature-closure errors that let a salvageable penumbra infarct).

Death by neurologic criteria is the irreversible cessation of all functions of the entire brain, including the brainstem, and it is fundamentally a clinical diagnosis established by demonstrated coma of known and irreversible cause, absent brainstem reflexes, and apnea, codified for the United States in the 2023 AAN/AAP/CNS/SCCM consensus practice guideline. Imaging enters as an ancillary test, not a substitute for the clinical examination: it is indicated only when a component of the clinical or apnea examination cannot be safely or reliably completed—severe facial or cervical-spine trauma precluding reflex testing, high cervical cord injury, the inability to perform apnea testing because of hemodynamic instability or hypoxemia, or the presence of confounders that cannot be fully cleared. The neurocritical reader must therefore understand both what the structural CT contributes to establishing irreversible cause and how the physiologic CT angiogram is interpreted when an ancillary study is required.

The noncontrast CT establishes a catastrophic, irreversible mechanism and excludes a reversible mimic. The findings that accompany whole-brain infarction are diffuse loss of gray-white differentiation, effacement of the cisterns and sulci, and the pseudo-subarachnoid sign—apparent hyperdensity in the basal cisterns and along the falx produced by the contrast between profoundly hypodense, edematous brain and the relatively higher density of the patent dural sinuses and superficial vessels; it measures lower than true blood (commonly $30$–$40\,\mathrm{HU}$ versus $50$–$70\,\mathrm{HU}$ for hemorrhage) and must not be read as subarachnoid hemorrhage. Crucially, the guideline requires that confounders be excluded before any death determination: core temperature must be near-normal (the examination is invalid in hypothermia), severe metabolic and endocrine derangements corrected, and central-nervous-system depressants and neuromuscular blockade accounted for—because sedatives, profound hypothermia, and barbiturate coma can all counterfeit the clinical and even the perfusion picture of brain death and render the determination invalid.

The physiologic principle underlying perfusion-based ancillary tests is the intracranial circulatory arrest that accompanies brain death: as global cerebral edema raises ICP to or above arterial pressure, cerebral perfusion pressure falls to zero and intracranial blood flow ceases. CT angiography exploits this by demonstrating absent opacification of the intracranial arteries despite a patent extracranial circulation, and it has become the most widely used ancillary study where radionuclide flow imaging is unavailable. Interpretation uses a validated opacification scoring system: the original 7-point scale assesses bilateral cortical (M4) middle-cerebral branches, the pericallosal arteries, the internal cerebral veins, and the great cerebral vein of Galen, but the simplified 4-point scale—requiring absent opacification of both M4 segments and both internal cerebral veins—has higher sensitivity (reported around $85$–$96\%$ with near-$100\%$ specificity) and better interobserver agreement, and is the preferred construct. The pivotal pitfall is the decompressed skull: a craniectomy, open fracture, or ventricular drain provides a route for pressure relief that may preserve some intracranial flow and produce a false-negative study, so a known skull defect must be factored into interpretation. The ranked alternative explanations for absent flow that the reader must exclude before attributing it to brain death are technical bolus failure or poor cardiac output (a non-diagnostic, not confirmatory, study) and the pharmacologic and thermal confounders above. Other validated ancillaries—radionuclide perfusion scintigraphy showing the hollow-skull sign with absent parenchymal uptake, and four-vessel digital subtraction angiography showing arrest of contrast at the skull base—are used per institutional protocol. The management consequence is absolute and irreversible: a valid ancillary study consistent with the clinical examination establishes legal death, with implications for withdrawal of support and organ donation, which is precisely why the failure modes carry such weight. The technical failure is mistaking inadequate contrast delivery for circulatory arrest; the cognitive failure is allowing the ancillary test to substitute for, rather than supplement, a rigorous exclusion of confounders and a complete clinical examination—the determination of death must never rest on imaging alone.