Roughly two million neurons die in the brain every minute a stroke goes untreated, a statistic stroke physicians cite constantly to explain why speed matters more in this condition than almost any other emergency diagnosis. At Catholic Health’s network of hospitals in Buffalo, New York, an AI platform called RapidAI now processes brain imaging and alerts entire stroke teams simultaneously, cutting analysis time that used to take upwards of 45 minutes down to just a few, according to a July 28, 2026 report from Spectrum Local News.
Why Minutes Have Always Decided Stroke Outcomes
Stroke treatment operates on some of the tightest clocks in medicine. Clot-busting drugs and mechanical clot-removal procedures are only effective within narrow windows after symptom onset, and every additional minute spent waiting for a radiologist to manually review a CT or MRI scan, then phone the on-call neurologist, then page the interventional team, is time the patient’s brain does not have. For decades that chain of phone calls and pages was simply how stroke care worked, and it left hospitals structurally dependent on how fast an individual radiologist happened to be reachable at 3 a.m.
How RapidAI Changes the Sequence
Rather than replacing that chain of communication, RapidAI is built to compress it. The software processes brain imaging data as soon as it is captured and pushes alerts to every relevant team member at once, according to Dr. Kalyan Shastri, Medical Director for Primary Stroke Centers at Catholic Health, rather than requiring a single radiologist to read the scan first and then manually notify each specialist in sequence. Dr. Arun Babu, Director of the Comprehensive Stroke Center at Mercy Hospital, part of the Catholic Health network, described the tool’s role as a prioritization aid rather than a diagnostic authority, telling Spectrum Local News, “We never trust the AI. We always do everything that we would normally do.”
A Deliberately Limited Role for the Algorithm
That framing matters. RapidAI is not being deployed at Catholic Health to make the call on whether a patient is having a stroke; it is deployed to make sure the humans who do make that call are looking at the right scan at the right moment, without waiting on a phone tree. That distinction reflects a broader pattern across emergency AI tools generally, where hospitals have been far more willing to adopt AI as a triage and alerting layer than as an independent diagnostic decision-maker, precisely because a missed or delayed stroke diagnosis carries such severe, irreversible consequences that clinicians are unwilling to fully delegate judgment to software, however fast it is.
A Crowded and Fast-Moving Field
Catholic Health’s use of RapidAI sits within a broader wave of investment in pre-hospital and in-hospital stroke AI. Separately, a company called AI-Stroke, founded in 2022, has raised a $4.6 million seed round to develop technology that lets paramedics record a 30-second video of a patient’s face, arm movement and speech, which its AI then analyzes for stroke signs before the patient even reaches a hospital CT scanner, built on what the company describes as the largest clinically annotated stroke-video dataset assembled to date, spanning 20,000 videos and six million images, according to a PR Newswire announcement of the funding round. Academic researchers have published parallel work in outlets including Scientific Reports showing that machine learning models fed by emergency medical services data can improve stroke triage decisions even before a patient arrives at the hospital door.
The Trust Gap Clinicians Won’t Close Yet
Even as adoption spreads, physicians using these tools remain explicit about their limits. Dr. Babu’s comment that his team never trusts the AI alone and still performs every standard diagnostic step reflects a broader caution among stroke specialists nationally: these tools have not been validated with the kind of large, multi-year randomized outcome data that would justify skipping conventional confirmation, and false alerts, while rare, could in theory divert scarce stroke team resources toward a patient who does not actually need urgent intervention. Separately, survey data from the American Medical Association found that 80% of physicians already use AI to summarize medical research in some form, suggesting clinical comfort with AI as a supporting tool is rising even as trust in it as a sole decision-maker remains firmly limited.
What Faster Triage Could Mean for Patients
If the time savings Catholic Health describes hold up across its full network and are replicated elsewhere, the practical effect for patients living near hospitals with similar systems could be measured in outcomes that matter most in stroke care: how much brain tissue is saved before treatment begins, and how many patients walk out of the hospital with less permanent disability. The next test for tools like RapidAI and AI-Stroke will be whether hospitals and researchers can produce large-scale outcome data, tracking actual disability and mortality rates rather than just diagnosis speed, that proves faster alerts translate into measurably better recoveries, not simply faster paperwork inside an already time-pressured emergency department.
