15. Hemorrhage

Hemorrhage on CT is the imaging endpoint of a failed or overwhelmed hemostatic system, and the density that makes acute blood conspicuous is itself a downstream product of the coagulation machinery; understanding the molecular events therefore both explains the CT signature and frames the diagnostic reasoning. Physiologic hemostasis proceeds in tightly choreographed phases. Vascular injury first triggers reflex vasoconstriction and exposes subendothelial collagen and tissue factor. Primary hemostasis follows as von Willebrand factor bridges exposed collagen to platelet glycoprotein Ib, platelets adhere, activate, change shape, and aggregate via fibrinogen cross-bridging of activated GPIIb/IIIa receptors, forming a friable platelet plug. Secondary hemostasis is the proteolytic amplification cascade: the dominant initiating route in vivo is the tissue-factor (extrinsic) pathway, in which TF-VIIa activates factor X, and the resulting prothrombinase complex (Xa-Va on a phospholipid surface) converts prothrombin (II) to thrombin (IIa). Thrombin is the pivotal enzyme-it cleaves fibrinogen to fibrin monomers, activates factor XIII to covalently cross-link the fibrin mesh, and feeds back to amplify its own generation through factors XI, VIII, and V. The product is a stable fibrin-platelet clot that then undergoes retraction, an actomyosin-driven contraction of platelets that expresses serum and concentrates the cellular and protein constituents of the clot.

This last step is the direct physical origin of hyperattenuation. Fresh whole blood in vivo measures only modestly above brain, but as a clot forms and retracts, the extruded serum leaves behind a dense matrix of fibrin and packed erythrocytes whose globin protein concentration rises sharply. X-ray attenuation in the diagnostic energy range is dominated, for low-Z materials, by electron density and the photoelectric effect; the high protein (and thus electron) density of retracted clot-not its iron content, which is too sparse to contribute meaningfully-raises the linear attenuation coefficient. A useful first-order relationship is that attenuation scales nearly linearly with hematocrit and protein concentration, so that

$\mathrm{HU}_{\text{clot}} \approx \mathrm{HU}_{\text{serum}} + k\,[\text{Hb}]\,,$

with every $\sim$1 g/dL rise in intra-clot hemoglobin adding on the order of a few HU. Acute extravasated, retracted blood therefore settles at $+50$ to $+80\,\mathrm{HU}$, conspicuously above the $+20$ to $+40\,\mathrm{HU}$ of normal brain parenchyma. Two corollaries follow that the expert must hold. First, in severe anemia (hemoglobin $\lesssim 8$-$10$ g/dL) acute hemorrhage may be near-isodense to brain, a recognized cause of missed bleeds. Second, coagulopathy abolishes clot formation and retraction: in patients on anticoagulants, with thrombocytopenia, hepatic failure, or consumptive coagulopathy, the failure of the very cascade described above both predisposes to bleeding and yields a poorly retracted, lower-density, often heterogeneous and expanding hematoma. The Bayesian reasoner thus reads the attenuation of a hemorrhage as a partial readout of the patient's coagulation status, and uses a low or heterogeneous density as a prior nudging toward coagulopathic or hyperacute bleeding-with direct implications for reversal therapy. The cognitive trap is to anchor on the textbook 50-80 HU figure and dismiss a genuine isodense or hypodense acute bleed in an anemic or anticoagulated patient.

A hematoma is not a static, single-shot event but frequently a dynamic, propagating process, and hematoma expansion is the single most important modifiable determinant of outcome in spontaneous intracerebral hemorrhage (ICH). The governing pathophysiology is best framed as a competition between local hemostasis and ongoing extravasation. The classical conception of a solitary ruptured arteriole giving way once has been supplanted by the avalanche (secondary-bleeding) model: the initial hematoma mechanically shears and disrupts adjacent vessels at its expanding margin, each of which bleeds and in turn fractures its neighbors, so that bleeding propagates outward through a cascade of small-vessel ruptures. This biomechanical cascade is sustained where local hemostasis is impaired-by the very coagulation factors diluted and consumed within the lesion, by anticoagulation, by elevated blood pressure increasing transmural stress, and by the loss of tissue tamponade in low-resistance regions. The histopathologic substrate of spontaneous deep ICH is usually lipohyalinosis and Charcot-Bouchard microaneurysms of the lenticulostriate, thalamoperforating, and basilar perforating vessels driven by chronic hypertension; in the elderly normotensive lobar bleed it is cerebral amyloid angiopathy, with vascular fragility from beta-amyloid deposition.

Expansion is defined quantitatively, and the thresholds matter because they anchor both trials and management. Volume is estimated at the bedside by the ABC/2 method, an ellipsoid approximation

$V \approx \frac{A \times B \times C}{2},$

where $A$ and $B$ are the largest perpendicular in-plane diameters of the hematoma and $C$ is the number of slices containing clot multiplied by slice thickness; the divisor of two derives from the volume of an ellipsoid, $\tfrac{4}{3}\pi(\tfrac{A}{2})(\tfrac{B}{2})(\tfrac{C}{2}) \approx \tfrac{ABC}{2}$. ABC/2 systematically overestimates large, irregular, or lobar hematomas (where the ellipsoid assumption fails) and is best treated as a rapid approximation that semi-automated volumetrics refine. Significant expansion is conventionally defined as an absolute increase $\geq 6\,\mathrm{mL}$ or a relative increase $\geq 33\%$ on follow-up CT, typically within the first 24 hours and most often in the hyperacute window; using either criterion, expansion occurs in roughly a quarter to a third of patients scanned early. Each milliliter of expansion incrementally worsens mortality and functional outcome, which is why early imaging, blood-pressure control, and rapid coagulopathy reversal are the pillars of acute management.

The expert prediction of expansion is a Bayesian synthesis of imaging markers, each carrying a likelihood ratio. On CT angiography the spot sign-one or more foci of contrast enhancement within the hematoma, representing active extravasation-is the most robust single predictor, with reported sensitivity for expansion in the rough range of 50-60% and specificity of 80-90%; multiplicity, larger spot size, and higher spot attenuation grade up the risk, and the post-contrast leakage sign (delayed-phase enhancement increase) refines it further. On non-contrast CT, a constellation of density and shape markers signals an actively evolving, poorly retracted clot: the blend sign (a hypoattenuating region blending with adjacent hyperattenuating clot, $\geq 18\,\mathrm{HU}$ difference), the black-hole sign (an encapsulated hypodensity within hyperdensity), the swirl sign (irregular hypodense regions of unclotted fresh blood), the island sign (scattered separate small hematomas around the main clot), and irregular margins or fluid levels. Mechanistically these all reflect the same thing-fresh, unretracted, ongoing bleeding admixed with older retracted clot-and their presence should raise the pretest probability of expansion and lower the threshold for aggressive intervention and short-interval re-imaging. The principal failure mode is anchoring on a single baseline scan: a small, deceptively benign-appearing early hematoma in a coagulopathic or spot-sign-positive patient may double before the next routine CT, so the imaging report should explicitly flag expansion predictors rather than merely state a volume.

Once extravasation ceases, a hematoma enters a reproducible biochemical evolution that drives its changing appearance across modalities, and although the dramatic signal cascade is the province of MRI, the same underlying chemistry produces the more monotonic but diagnostically vital attenuation changes on CT. The trajectory is one of progressive deoxygenation, oxidative denaturation, and macrophage-mediated clearance of hemoglobin iron. At the moment of bleeding the clot contains oxygenated, intact erythrocytes with diamagnetic oxyhemoglobin. Within hours, as the closed environment exhausts oxygen, hemoglobin is reduced to paramagnetic deoxyhemoglobin sequestered within intact cells. Over the next days, continued oxidative stress converts the iron from the ferrous ($\mathrm{Fe}^{2+}$) to the ferric ($\mathrm{Fe}^{3+}$) state, yielding methemoglobin, first intracellularly and then-as erythrocytes lyse-extracellularly. Finally, infiltrating macrophages phagocytose the breakdown products and store iron as ferritin and hemosiderin, a superparamagnetic residue that may persist indefinitely. On MRI this sequence produces the well-known $T1$/$T2$ signature summarized below; on CT the chemistry matters chiefly because clot retraction and the high protein/electron density of the early clot give the hyperattenuation, while subsequent proteolysis, osmotic water influx, and clearance progressively dilute that density.

Phase	Time	Dominant blood product	Compartment	MRI T1	MRI T2	CT attenuation
Hyperacute	$<$ 24 h	Oxyhemoglobin	Intracellular	Iso/hypo	Iso/hyper	Hyperdense ($+60$-$80$)
Acute	1-3 d	Deoxyhemoglobin	Intracellular	Iso/hypo	Hypo	Hyperdense ($+60$-$80$)
Early subacute	3-7 d	Methemoglobin	Intracellular	Hyper	Hypo	Iso/slightly hyperdense, falling
Late subacute	1-4 wk	Methemoglobin	Extracellular	Hyper	Hyper	Isodense to brain
Chronic	$>$ 1 mo	Hemosiderin/ferritin	Extracellular (macrophage)	Iso/hypo	Hypo (rim)	Hypodense / slit-like cavity

The central mechanistic point for the CT reader is the dissociation between the iron chemistry and the attenuation curve. Iron in hemoglobin contributes negligibly to X-ray attenuation; what falls on CT is the protein and cellular density of the clot as fibrinolysis breaks down the matrix, erythrocytes lyse, and water is osmotically drawn in. Consequently the attenuation decline is comparatively smooth and slow, in contrast to the abrupt, paramagnetism-driven signal flips on MRI. Two CT phenomena deserve emphasis because they are diagnostically powerful. The hematocrit (sediment) effect appears when cells settle dependently within a poorly clotting or coagulopathic hematoma, producing a fluid-fluid level with a dependent hyperdense layer and a supernatant hypodense plasma layer-a strong marker of coagulopathy, anticoagulation, or rebleeding. The swirl sign of intermixed fresh hypodense blood signals incomplete clotting and ongoing or recent extravasation. The expert uses blood-product reasoning bidirectionally: to age a hemorrhage when the clinical timeline is unknown, and to flag the discordant case-an unexpectedly hypodense, layered, or heterogeneous acute hematoma-as evidence of impaired hemostasis that mandates a coagulation workup. The failure mode is the late-subacute isodense window, during which a resolving parenchymal or extra-axial hematoma transiently matches brain attenuation and can be overlooked entirely unless indirect signs (mass effect, sulcal effacement, a thin enhancing rim, displacement of the gray-white interface) are actively sought.

The defining clinical utility of CT in hemorrhage is that attenuation acts as a chemical clock, and the disciplined reader can place a bleed within a temporal window from density alone-an inference that frequently reconciles the imaging with an uncertain or unreliable history and that directs the differential. In the hyperacute and acute phases (first hours to roughly three days) the retracted, protein-dense clot sits at $+50$ to $+80\,\mathrm{HU}$, occasionally reaching the high 80s or low 90s in a polycythemic patient or a tightly packed clot. As fibrinolysis, erythrocyte lysis, and osmotic dilution proceed, attenuation declines at an approximate and clinically useful rate of $\sim$1-1.5 HU per day from the periphery inward, so that a parenchymal hematoma becomes isodense to brain at roughly 1-2 weeks and frankly hypodense by 3-4 weeks, ultimately resolving to a slit-like, fluid- or gliosis-filled cavity with surrounding encephalomalacia and, on MRI, a hemosiderin rim. The decline is not perfectly linear-it is fastest at the margin where clearance is most active and where surface-to-volume ratio is highest-which produces the characteristic centripetal pattern of a shrinking hyperdense core surrounded by an advancing hypodense rim.

Several temporally specific signatures sharpen this dating. A fluid-fluid (hematocrit) level in the acute setting points to coagulopathy or anticoagulation rather than the slow sedimentation of a chronic collection. In the subacute phase, breakdown of the blood-brain barrier and reactive neovascularity at the hematoma margin produce a thin, smooth, ring-enhancing capsule on contrast-enhanced CT, typically emerging around the end of the first week, peaking over weeks two to six, and slowly resolving; this is a benign expected evolution but is a classic mimic of a ring-enhancing neoplasm or abscess, and the distinction rests on the thin, regular, non-nodular character of the rim, the supporting blood-product evolution, the clinical context, and-where doubt persists-MRI or follow-up demonstrating expected involution. The same subacute window is where the resolving hematoma transits isodensity, the most treacherous moment in hemorrhage imaging, when a sizeable collection can vanish into the background gray scale and is betrayed only by mass effect or an enhancing margin. For extra-axial collections the same chemistry yields the well-known staging: a hyperdense acute subdural hematoma, an isodense subacute subdural (notoriously occult, especially when bilateral and symmetric, effacing sulci and shifting the gray-white interface medially), and a hypodense chronic subdural, with acute-on-chronic rebleeds producing layered densities.

The interpretive discipline is to read attenuation as one axis of a multidimensional inference rather than a literal timestamp, because several states perturb the curve: anemia lowers the starting density, polycythemia and hemoconcentration raise it, coagulopathy yields a heterogeneous and slowly retracting clot, and rebleeding resets the clock with admixed fresh hyperdensity. The principal technical pitfalls are partial-volume averaging, which can blur a small or peripheral bleed against adjacent dense bone or CSF (mitigated by thin sections and multiplanar review), and beam-hardening artifact adjacent to the petrous bones and at the vertex, which can both fabricate apparent hyperdensity and obscure true subarachnoid or extra-axial blood. The dominant cognitive failure is satisfaction of search-finding one obvious hemorrhage and stopping-when the entity of interest is a second, subacute, isodense, or extra-axial collection requiring active interrogation across optimized windows.

The lethality of intracranial hemorrhage derives less from the destruction of tissue at the bleeding site than from the secondary mechanical and physiologic consequences of an expanding mass within a fixed, non-compliant container, and these are governed by the Monro-Kellie doctrine: because the adult cranium is rigid, the sum of the intracranial volumes of brain, blood, and cerebrospinal fluid is constant, so any added volume-a hematoma plus its perihematomal edema-must be matched by displacement of an equal volume of the normal constituents. The initial compensation is buffering: CSF is displaced into the spinal thecal sac and venous blood is expelled, which keeps intracranial pressure (ICP) nearly flat over an early compensated phase. This reserve is finite. The relationship between added volume and pressure is the intracranial elastance curve, exponential in form,

$P = P_0\,e^{\,kV},$

so that once the CSF and venous buffers are exhausted, each additional milliliter of clot or edema produces a steep, decompensated rise in ICP. The clinical and imaging corollary is that two hematomas of identical volume can have utterly different consequences depending on where they sit on this curve-which is why posterior fossa and intraventricular hemorrhages, where small volumes rapidly obstruct CSF or compress the brainstem, are disproportionately dangerous, and why perihematomal edema, which can equal or exceed the clot volume over the following days, is a major and often underweighted contributor to delayed deterioration.

The CT signatures of mass effect form a graded hierarchy that the reader should interpret as a readout of position on the elastance curve and of impending herniation. Local effects-effacement of adjacent sulci and compression or displacement of the ipsilateral ventricle-give way to midline shift, conventionally measured at the septum pellucidum or the level of the foramen of Monro; a shift exceeding $\sim5\,\mathrm{mm}$ is generally significant and correlates with depressed consciousness, though the relationship is modulated by the chronicity of the lesion and the patient's atrophy reserve. Progressive shift produces the herniation syndromes, each with a characteristic CT appearance and vascular complication. Subfalcine herniation displaces the cingulate gyrus under the falx and can compress the anterior cerebral artery, causing medial frontal infarction. Uncal (transtentorial) herniation pushes the medial temporal lobe over the tentorial edge, effacing the suprasellar and perimesencephalic cisterns, compressing the third cranial nerve (a blown pupil), the posterior cerebral artery (occipital infarction, the classic Kernohan notch false-localizing hemiparesis from contralateral peduncle compression), and ultimately the brainstem; secondary Duret hemorrhages in the midbrain and pons follow from stretching of perforating vessels. Tonsillar herniation through the foramen magnum compresses the medulla and is rapidly fatal, while central (downward) herniation caudally displaces the diencephalon. The obliteration of the basal cisterns is one of the most ominous and reliable CT signs of critical mass effect, and obstructive hydrocephalus from intraventricular extension or fourth-ventricular compression adds a further, treatable volume load.

Expert interpretation integrates these findings into a management-directed synthesis. The reader estimates total mass (clot plus edema), localizes the patient on the compensation-to-decompensation spectrum using cisternal and ventricular morphology rather than midline shift alone, identifies the specific herniation pattern and its vascular territory at risk, and flags treatable contributors-expansion, hydrocephalus, and edema. Two failure modes recur. The first is underweighting infratentorial and intraventricular hemorrhage, where catastrophic physiology accompanies modest volume and where partial-volume and beam-hardening artifact at the skull base obscure the very findings that matter. The second is misjudging shift in the atrophic brain, where generous CSF reserve permits a large lesion with little shift, or conversely in the young patient where minimal reserve produces dangerous pressure at small volumes. The throughline is that mass effect, not the hemorrhage per se, most often determines whether and how urgently the patient goes to the operating room, and the report must therefore quantify and localize it explicitly.