Why CT first

Stroke is either a blocked artery (ischemic, about 85 %) or a burst one (hemorrhage, about 15 %), and the two look the same at the bedside: sudden weakness, numbness, speech or vision loss. The treatments are opposite. A clot-buster that opens a blocked artery makes a bleed worse. So the first scan does not need to show the stroke; it needs to show whether there is blood. Non-contrast CT does that in under a minute with near-perfect sensitivity for an acute bleed, which is why the stroke pathway runs straight from the door to the scanner (target: imaging within 20 minutes of arrival, needle within 60, ideally 45). Only a finger-stick glucose is needed before it, because low sugar mimics stroke.

How the picture is made

The tube on the rotor throws a thin fan of X-rays through the head to the detector arc opposite. The rotor turns about twice a second; each detector channel records how much of the beam got through at each angle, and the computer back-projects a thousand such views into a map of attenuation in Hounsfield units: water 0, air −1000, bone over +1000. Brain sits in a narrow band: CSF about 5, white matter about 28, gray matter about 38, fresh blood 50 to 80. A window maps a slice of that range to black through white. The brain window (width 80, center 40) spreads 0 to 80 HU across the gray scale; the narrower stroke window (width 40, center 35) doubles the contrast, so a few HU of early swelling becomes visible.

What each scan answers

• NCCT, non-contrast. Blood is bright from minute one. Ischemic brain looks normal at first: the infarct core only darkens as it takes up water (about 2 % an hour relative to the healthy side, each 1 % costing about 0.4 HU of gray matter), so in the first 2 to 3 hours many scans are read as normal. Early signs: loss of the gray-white border at the insula (insular ribbon) and lentiform nucleus, sulcal effacement, and a hyperdense middle cerebral artery, the clot itself (about 60 to 80 HU in the vessel). ASPECTS scores ten MCA regions at two levels; one point is lost for each region with early change. 10 is normal; 6 or more favors thrombectomy.
• CTA, CT angiography. 50 to 80 mL of iodinated contrast at 4 to 6 mL/s, scanned as the bolus reaches the neck and brain. Shows the occluded artery (here the left M1: contrast stops at the clot and the branches beyond stay empty), the route for a catheter, and in a hemorrhage the spot sign, contrast leaking into the clot, which predicts growth.
• CTP, perfusion. A second bolus scanned every 1 to 2 seconds for about a minute over the same slab. Each voxel's time-density curve gives blood flow, volume and transit time. Core: flow under 30 % of normal (dead or dying). Penumbra: Tmax over 6 s (starving but alive). Between 6 and 24 hours the core and the mismatch decide thrombectomy (DEFUSE 3: core under 70 mL, ratio at least 1.8, mismatch at least 15 mL).

Time windows

• IV thrombolysis: alteplase 0.9 mg/kg (10 % bolus, max 90 mg) or tenecteplase 0.25 mg/kg (max 25 mg), within 4.5 h of last known well, no blood on NCCT, blood pressure under 185/110.
• Thrombectomy for a large-vessel occlusion: up to 6 h on CT and CTA (ASPECTS 6 or more, recent trials extend to 3 to 5); 6 to 24 h with perfusion or clinical-core mismatch.
• Untreated, a typical large-vessel stroke loses about 1.9 million neurons a minute. The model here grows the core from 10 mL to 120 mL of a 130 mL territory over a day.

Keyboard

Focus the figure. Click it or press 1, 2, 3 for NCCT, CTA, CTP. ← → change hours since onset (Shift for 2 h), B switches bleed and clot, W switches window, Space stops the tour.

Teaching model, not for clinical use. Water uptake 30 % × (1 − e^(−t/12 h)); core 10 + 110 × (1 − e^(−t/10 h)) mL; hematoma 14 to 24 mL, 56 to 78 HU. A reader is assumed to see a patch once its contrast reaches 4 % of the window.