Every argument at a stroke alert — lyse or don't, fly the patient or don't, extend the window or don't — is secretly an argument about one question: how much brain in there is dead, and how much is merely dark? The vocabulary for that question is fifty years old, and it still runs the field.

Two thresholds, three tissues

Astrup and colleagues formalized what electrophysiology had shown: as cerebral blood flow falls, neurons cross two distinct thresholds. At the first — roughly a third of normal flow — synaptic transmission fails: the neuron falls silent, the deficit appears, but the cell keeps its membranes and its future. At a second, deeper threshold, ion pumps fail, potassium floods out and calcium floods in, and infarction begins.1 Between the thresholds lies the ischemic penumbra: tissue that is electrically dead and biologically alive — the "not working yet not dying" ring around a core that is doing both.

Three tissues, then, at any moment of an occlusion: core (dead or irreversibly dying), penumbra (silent, salvageable, generating much of the visible deficit), and oligemia (underperfused but safe). Reperfusion therapy has exactly one purpose: convert penumbra's fate from core's to oligemia's before the core annexes it.

Collaterals: why the clock lies in both directions

The penumbra is not a static ring; it is a siege being provisioned. Leptomeningeal collaterals — pial back-channels from neighboring territories — determine how long the siege holds. The patient with robust collaterals is a slow progressor: hours in, the core is still small, and treatment remains handsomely rewarded. The patient with sparse collaterals is a fast progressor: the same M1 occlusion consumes its territory in under an hour, and even an early arrival may find little left to save. Same artery, same clock, opposite truths — which is why chronological time is a proxy, and an increasingly disrespected one.

Saver's arithmetic made the average cost of delay unforgettable: a typical untreated large-vessel stroke destroys about 1.9 million neurons and 14 billion synapses per minute.2 The number is an average across exactly the collateral spectrum above — which is the correct way to hold it: run the system like every minute costs two million neurons (see the golden hour), and read the individual patient like the number might be wrong by an order of magnitude in either direction.

The clock estimates the physiology. The collaterals are the physiology.

Why imaging selection works

If penumbra size at arrival is what varies, then measuring it should beat assuming it — and that is precisely the story of the modern trials. Perfusion imaging estimates the badly hypoperfused volume; diffusion (or CT proxies like ASPECTS) estimates core; a large mismatch is a penumbra held open by collaterals. The late-window thrombectomy paradigm — patients selected 6–24 hours out by mismatch or clinical-core mismatch — treats exactly the slow progressors, and its outsized treatment effects are the physiology cashing the check. The same logic underlies imaging-selected late thrombolysis and the wake-up stroke pathways now embedded in guidelines.3 Selection details live on the thrombectomy eligibility page and the IV thrombolysis page; the trial arc is traced in trends in stroke care.

The cascade, in usable terms

Inside the failing tissue, the sequence is worth knowing at the level that explains clinical facts. Energy failure → membrane depolarization → massive glutamate release → excitotoxic calcium influx → mitochondrial failure, free-radical generation, protease activation → necrosis in the core, slower programmed death at the margins — amplified over hours by spreading depolarization waves that tax the penumbra's meager budget, and over days by inflammation and blood-brain-barrier breakdown (the road to edema and hemorrhagic transformation). Two honest clinical corollaries: this cascade is why speed beats any drug we have — and why the graveyard of neuroprotectant trials is so large: blocking one step of a web, late, in humans, has not yet reproduced what reperfusion does.3

Reperfusion's fine print: injury, no-reflow, futile recanalization

Restoring flow is not free. Reperfusion injury — oxidative stress, inflammatory influx, barrier damage on re-oxygenation — contributes to edema and hemorrhagic transformation in badly injured tissue; it is one reason large-core reperfusion demands careful post-procedure care rather than celebration. And recanalizing the artery does not guarantee reperfusing the tissue: no-reflow — microvascular obstruction from pericyte constriction, swollen endfeet, plugged capillaries — can leave downstream beds ischemic behind a beautifully open vessel. Together these phenomena help explain futile recanalization: the TICI 3 result with the unchanged exam. The practical reading is neither nihilism nor hurry-less-ness — it is that the physiology asks two things of a system: open the artery fast, and protect the microcirculation and the tissue after (pressure management, glucose, temperature — the unglamorous half of the post-reperfusion targets).

The bottom line

Hold the model and the decisions order themselves. Two thresholds make a penumbra; collaterals decide how long it lasts; time is a population average wearing a stopwatch; imaging exists to catch the individual truth; and reperfusion is the beginning of tissue rescue, not the end. Every eligibility page on this site is an application of one sentence: find the brain that is dark but not dead, and turn its lights back on before the definition changes.