How Perfusion Imaging Extended the Stroke Treatment Window

The short answer#

Two clinical trials published in 2018, DAWN and DEFUSE 3, showed that some stroke patients still benefit from mechanical clot removal 6 to 24 hours after they were last known well, far outside the old treatment window. The reason the field could stretch that window was not a faster procedure but a better picture. Perfusion imaging lets clinicians see which brain tissue has already died and which is still starving but alive, and it is the amount of living, threatened tissue, not the hours on the clock, that predicts who has something to gain.

Key points#

• The traditional stroke window was a fixed clock: treat within roughly 6 hours for a large vessel blockage, or the door closed.
• Perfusion imaging maps two volumes: the dead core and the at-risk but salvageable penumbra. A large gap between them is called a mismatch.
• DAWN and DEFUSE 3 used mismatch to select late patients for thrombectomy, and both were stopped early because the benefit was so clear.
• The results apply only to patients who fit narrow entry criteria. They do not make everyone a late candidate.

Why brain tissue does not die all at once#

When a clot lodges in an artery feeding the brain, the tissue downstream loses its blood supply, but it fails in stages rather than all at once. At the center of the blocked territory is the infarct core, tissue that is already beyond rescue. Wrapped around it is the penumbra, brain that is underfed and functionally silent yet still structurally alive, kept going by small collateral vessels rerouting a trickle of blood. The penumbra is the prize. Reopen the artery while it survives and much of it can recover. Wait too long and it collapses, one region at a time, into permanent core.

Here is the part that matters for treatment: the speed of that collapse varies enormously between people. Someone with a dense network of collateral vessels may hold a large penumbra for many hours. Someone with poor collaterals can lose the same tissue in under an hour. Two patients at the identical 10-hour mark can be in completely different situations.

Why the old rule used a clock anyway#

For years, acute stroke care ran on time limits. Clot-dissolving drugs and, later, thrombectomy (threading a catheter up to the clot and physically pulling it out) were tested and approved inside tight windows, commonly within about 6 hours for a large vessel occlusion. On average the logic held, because the longer the delay, the more penumbra has usually converted to core.

The weakness of an average is that it flattens the people at the edges. A fixed clock treats the fast collapser and the slow collapser as the same case and, by doing so, turns away the slow collapser who still has a large volume of rescuable brain long after the standard cutoff. What was missing was a way to tell those two patients apart at the bedside.

How a perfusion scan draws the map#

Perfusion imaging supplies that distinction. Using either CT perfusion or diffusion and perfusion MRI, the scanner tracks how blood moves through brain tissue. Automated software, most prominently the RAPID program used in these trials, reads the images and gives you volume estimates: roughly how much tissue is already infarcted, and how much is underperfused but not yet dead. The difference between the two numbers is the mismatch. A large mismatch means a small amount of brain has been lost relative to a large amount still hanging on.

That single measurement reframes the whole decision. The question stops being "how many hours has it been" and becomes "how much brain can still be saved." The clock is replaced by a tissue map, and the map is what carries the predictive weight.

The two trials that proved it#

DAWN, reported in the New England Journal of Medicine in 2018 by Nogueira and colleagues, enrolled patients with a blockage in the intracranial internal carotid artery or the proximal middle cerebral artery who were last known well 6 to 24 hours earlier. It defined its target group by a clinical-imaging mismatch: a severe neurological deficit, scored on the NIH Stroke Scale, sitting alongside a small infarct core, with the exact thresholds tuned by age. At 90 days, 49 percent of the thrombectomy group had regained functional independence versus 13 percent with standard care. The trial was halted early for benefit.

DEFUSE 3, published the same year by Albers and colleagues, enrolled patients 6 to 16 hours out and drew its mismatch from imaging alone: an ischemic core under 70 mL, a mismatch ratio above 1.8, and an absolute mismatch of at least 15 mL. Functional independence at 90 days reached 45 percent with thrombectomy versus 17 percent with medical management. It, too, stopped early.

Two trials, two slightly different selection recipes, one conclusion. When the scan shows a small core and a large threatened penumbra, reopening the artery late still helps, and the size of the effect is unusual for stroke medicine. Roughly one additional person reaches an independent life for every few patients treated.

What the guidelines settled on#

The 2019 update to the American Heart Association and American Stroke Association guideline for acute ischemic stroke folded both trials into practice. For selected patients 6 to 16 hours from last known well with an anterior circulation large vessel occlusion who meet DAWN or DEFUSE 3 criteria, thrombectomy carries a Class I recommendation, the strongest tier. For selected patients in the 16 to 24 hour window who meet DAWN criteria, it is Class IIa, meaning reasonable. The guideline recommends obtaining CT perfusion or diffusion MRI to guide selection, but only for patients who otherwise fit the trial criteria. That final qualifier is where your careful reading earns its keep.

The fine print#

The power of these trials is inseparable from their narrowness. Benefit was shown in patients who cleared specific gates: particular vessels, a favorable mismatch profile, defined limits on core size, and deficits severe enough to qualify. A patient with a large established core, weak collaterals, or a blockage in a different location was never represented, and you should not stretch the findings to cover them. The trials proved that imaging can find late responders. They did not prove that imaging turns everyone into a candidate.

Two further cautions keep the interpretation honest. First, RAPID and comparable tools produce estimates, not ground truth, and calculated core volumes can shift with the scanner you have, the timing of the scan, and the processing thresholds. Second, the enrolled patients were a selected minority of everyone arriving late with a stroke, so the striking success rates belong to that enriched group rather than to the average person walking in 12 hours after symptoms began. Researchers are still testing whether simpler imaging or looser criteria can safely widen the pool. The fair summary is that perfusion-based selection is well validated for the populations studied and remains an open question at the margins.

What is not in doubt is the conceptual shift. DAWN and DEFUSE 3 swapped a single blunt question, how long has it been, for a sharper one, how much brain is still alive. That is why the stroke treatment window is now written in tissue as much as in time.