Welcome back to Healthy Innovations! 👋
This week’s deep dive is on stroke – and how AI is beginning to help at two critical moments: spotting a clot sooner and, eventually, reaching it faster. We look at new work from King’s College London on autonomous catheter navigation, alongside real-world NHS evidence showing that AI-read CT scans can flag “silent” strokes that might otherwise be missed.
Globally, an estimated 12 million people have a stroke each year, and roughly 6.5–7 million die – so the stakes could hardly be higher.
Let’s dive in!
Join over 1,000 healthcare leaders who get these insights delivered straight to their inbox! Healthy Innovations is read weekly by executives from AstraZeneca, GSK, Vertex, Roche, leading academics, startup founders, agency teams, and investors. Subscribe now to stay ahead of industry trends!
The stroke no one was looking for
Jean, 83, collapsed at home on a weekend in August, fracturing her collarbone in the fall, according to an NHS England press release. There was no slurred speech and no obvious loss of movement, just a broken bone and a trip to the Royal Berkshire Hospital. It was the routine CT scan on arrival, read within minutes by an AI imaging tool, that caught what nobody had suspected: a major ischemic stroke unfolding quietly underneath the fracture.
Within 25 minutes of reaching the emergency department, Jean was in a blue-light ambulance to the John Radcliffe Hospital in Oxford for a mechanical thrombectomy, a procedure that physically retrieves the clot blocking blood flow to the brain. It worked. "I feel incredibly lucky," she said. "Everything happened so quickly, and I know that made all the difference."
Her stroke was never going to announce itself. It was caught because a machine happened to be looking.
Speed decides stroke outcomes more than almost anything else clinicians can control. Each minute a large vessel occlusion goes untreated, roughly 1.9 million neurons die, according to a widely cited quantification of the "time is brain" principle. For every 20 minutes thrombectomy is delayed, the chance of full recovery drops by about a percentage point. Mechanical thrombectomy itself has been standard of care since 2015. What has lagged behind is the speed of working out who needs it.
Why diagnosis still takes too long
Identifying a large vessel occlusion precisely enough to trigger urgent transfer and a specialist team has traditionally depended on a radiologist reading a scan within a workflow built for routine cases, not emergencies. In many hospitals, that reading arrived 30 to 90 minutes after the scan was taken. AI was set a narrow task: read the scan and alert the team before a radiologist gets there. It is now doing that at scale.
Brainomix, an Oxford company whose 360 Stroke platform analyzes CT scans to flag occlusions, has the largest real-world evidence base of any stroke AI tool. A study published in The Lancet Digital Health in December 2025 found:
452,952 patients tracked across 107 NHS England hospitals over five years
Thrombectomy rates at evaluation sites doubled, from 2.3% to 4.6%
Around 15,000 patients are estimated to have directly benefited from AI-reviewed scans.
Jean was one of them.
Brainomix has since expanded into the US, earning its own FDA clearances and opening offices alongside its UK and Ireland operations.
Image source: Brainomix
Viz.ai, deployed across more than 1,800 hospitals in the US and Europe, showed similar gains in a 474-patient study presented at the International Stroke Conference in 2025:
A 44.13% reduction in time from arrival to large vein occlusion (LVO) diagnosis
Treatment times cut by an average of 31 minutes.
In stroke, 31 minutes is measured in neurons, not convenience. These tools work alongside radiologists, not instead of them, closing the gap between the scan and the alert.
AI takes the wheel
Reading the scan faster only helps if a specialist can act on it, and specialists able to perform thrombectomy remain scarce almost everywhere. The next experiment tries to remove that bottleneck at its source.
March 2026: Researchers at King's College London published the first demonstration that AI can autonomously steer a catheter through the full route from the leg to the brain in a physical lab setting, not a simulation, guiding a wire through a 3D-printed model of the blood vessels involved in thrombectomy.
April 2026: The same team brought together neurointerventionists and roboticists, working with patient representatives from the Stroke Association, to publish the first international consensus standards for testing robotic thrombectomy systems.
Senior author Dr. Thomas Booth was clear about where things actually stand: "While AI-assisted robotic MT is not yet ready for routine clinical use, the speed of innovation in both robotics and AI means transformative advances are likely to be just around the corner."
If autonomous or remote navigation eventually clears clinical trials, the implication is significant. A patient in a hospital without a neurointerventional team on site could, in principle, be treated by a specialist operating remotely, or by a system trained to complete part of the procedure itself. That is a distant prospect. But it is the first time the AI conversation in stroke has moved from watching to doing.
Image source: Brainomix
The access gap AI can't close alone
None of this touches the more basic global problem: most eligible patients never get a thrombectomy at all.
A worldwide survey of 75 countries, the MT-GLASS study, found a global median access rate of 2.79%
The gap between high-income and low-income countries runs to almost 90%
In India, where stroke cases exceed a million a year, thrombectomy-capable centers remain rare outside major cities
India's Technology Development Board is now backing S3V Vascular Technologies, a Mysuru-based manufacturer, to build the country's first domestically produced thrombectomy devices, aiming to bring costs down enough to widen access. It is slow, unglamorous work, and it matters more than any single algorithm.
The minutes that matter
Jean went home, and has thought since about the ward she briefly shared with people who arrived later, or at a hospital without the imaging pipeline to route them quickly enough. "It really hit me," she said, "how life-changing a stroke can be."
The tools that caught her stroke and the robot that steered a wire through a lab model of her arteries are solving two different halves of the same problem: seeing the clot fast enough, and eventually, reaching it without waiting for a specialist to be in the building.
Neither is finished work. Together, they're the clearest sign yet that the industry chasing stroke's shrinking window is starting to close it from both ends.
Innovation highlights
🐭 Old gut, foggy brain. Researchers found that aging alters gut bacteria in mice, weakening signals sent to the hippocampus via the vagus nerve. Young mice housed with older mice picked up their microbiome and started forgetting things too. Stimulating the vagus nerve, or clearing the bacteria with antibiotics, reversed it. One bacterium, Parabacteroides goldsteinii, was directly implicated. Vagus nerve stimulation is already FDA-approved for depression and epilepsy, so if the pathway holds in humans, a delivery device may already exist.
⚡ No more needles for MND. Researchers are using ultrasound and computer vision to detect motor neuron disease without the needle-based electromyography tests patients dread. Their software analyzes ultrasound videos for fasciculations, involuntary muscle twitches that can signal MND, and classifies them using machine learning. Needle EMG remains a key diagnostic tool, but patients often need repeated invasive tests before a diagnosis is reached. This approach could speed that up and make monitoring far more comfortable.
🧠 MS lesions, finally visible. A team used AI to reveal cortical brain lesions in multiple sclerosis that standard MRI has never been able to detect, despite gray matter damage being linked to disease progression and cognitive decline for over a century. Applying the method to scans from a 700-patient Ocrelizumab trial, they found 15 to 20 lesions per patient, more than 11,000 in total, all invisible on the original images. The tool spots subtle differences across multiple scan contrasts rather than reading a single image.
Company to watch
🏥 Courier Health. Getting a prescription is the easy part. For patients with chronic or rare diseases, the real fight often starts after that: benefits verification, prior authorization, insurance hurdles, before a single dose gets taken.
Courier Health, a New York-based startup, built an AI platform that replaces the patchwork of generic CRMs biopharma companies have long relied on for this, coordinating intake, approvals, and ongoing adherence support in one place, and flagging patients at risk of falling through the cracks. Some deployments have cut time-to-treatment by roughly two-thirds.
The company recently closed a large new funding round, on the back of quadrupled therapy coverage and more than doubled headcount in 2025.
Image source: Courier Health
Weird and wonderful
🏺 Ancient pots, modern medicine. A Korean designer has reimagined a 3,000-year-old cooling trick as a medical fridge. EVAPOT uses evaporating water to pull heat away from its contents, no compressor, no refrigerant, no electricity grid required. The idea borrows directly from the zeer pot, a pot-in-pot cooler that ancient Egyptians and Persians used long before refrigeration existed, and that a Nigerian teacher named Mohammed Bah Abba revived in the 1990s to help rural families store food, earning him a Rolex Award for Enterprise along the way.
The target use case is cold chain storage for vaccines, insulin, and blood products in clinics that lose power or never reliably had it. It is still just a concept, and evaporative cooling has a real weakness: it struggles in humid climates, which happen to be where a lot of that need exists. But it is a good reminder that not every medical breakthrough needs a battery, and that some of the best ideas in healthcare are just old ideas finally pointed at the right problem.
Image source: Yanko Design
Thank you for reading the Healthy Innovations newsletter!
Keep an eye out for next week’s issue, where I will highlight the healthcare innovations you need to know about.
Have a great week!
Alison ✨
P.S. Healthcare is evolving at an unprecedented pace, and your unique insights could be invaluable to others in the field. If you're considering starting your own newsletter to share your expertise and build a community around your healthcare niche, check out beehiiv (affiliate link). There's never been a better time to start sharing your knowledge with the world!
