England’s National Health Service has completed the world’s first national rollout of AI-assisted stroke diagnosis, equipping every stroke center in the country with decision-support technology. The results from over 111,000 patients show the system triples recovery rates.
The Numbers
Before the AI deployment, 16% of stroke patients recovered with no or minimal disability. After: 48%. That’s a threefold increase.
Hospitals using the Brainomix 360 Stroke system saw thrombectomy rates - the mechanical removal of blood clots from the brain - double from 2.3% to 4.6%. Hospitals without the technology saw smaller gains, from 1.6% to 2.6%.
Perhaps more significant than the treatment rates: time. The AI reduced “door-in-door-out” time from 140 minutes to 79 minutes, cutting 61 minutes from the assessment window. In stroke treatment, this matters enormously - brain tissue dies at a rate of roughly 1.9 million neurons per minute during a stroke.
How It Works
The Brainomix e-Stroke system analyzes CT scans to detect large vessel occlusions (LVOs) - blockages in major arteries that supply the brain. These are among medicine’s most time-sensitive diagnoses. The AI flags suspected LVOs immediately, helping clinicians decide whether to transfer patients to specialized thrombectomy centers.
The system doesn’t replace radiologists. It processes scans faster than humans can, providing decision support while definitive reads follow. When the AI detects something concerning, clinicians get alerted within minutes of the scan completing.
Across the 11 stroke networks using the technology, the system detected over 4,500 large vessel occlusions. Some regions saw more than double their previous access to mechanical thrombectomy.
Scale of Deployment
The rollout covers more than 70 NHS hospitals and all regularly admitting stroke services in England. This happened through the NHS’s AI in Health and Care Award program, backed by £123 million in funding to test and scale health technologies.
Five stroke networks received direct funding through the program. The rest adopted the technology after seeing results from the funded sites. This makes it the largest prospective evaluation of AI in acute stroke care anywhere in the world.
What This Demonstrates
Medical AI typically struggles to match its performance in research settings when deployed at scale. This data suggests otherwise - at least for this specific application.
The results come from a prospective observational study published in The Lancet Digital Health, comparing outcomes at sites using the AI against those that weren’t. The design allows for real-world assessment rather than controlled trial conditions.
One patient’s story illustrates the human impact. Carol Wilson, a teaching assistant, had a brain blood clot detected rapidly using the AI system. She received thrombectomy treatment and returned to work within days.
The Fine Print
The technology works well for what it’s designed to do: detect large vessel occlusions quickly. This represents a subset of all strokes. Ischemic strokes from smaller vessel blockages and hemorrhagic strokes (bleeding rather than clots) require different approaches.
The threefold improvement in recovery rates reflects both faster detection and increased access to thrombectomy - a procedure that’s highly effective when delivered quickly but limited to certain stroke types. Not every patient benefits from these specific interventions.
Still, the scale and consistency of these results across the English NHS provides some of the strongest evidence yet that AI can meaningfully improve patient outcomes when deployed nationally rather than just tested in research settings.