For decades, the intensive care unit has demanded a difficult bargain: to know a critically ill patient's blood pressure in real time, clinicians have had to pierce an artery, accepting the risks of infection, clotting, and immobility as the cost of vigilance. Researchers at Johns Hopkins University are now challenging that bargain with MOSAIC, a wearable two-sensor system guided by artificial intelligence that reconstructs continuous blood pressure waveforms without breaking the skin. In early trials, the technology matched the precision of arterial catheters closely enough to suggest that me
AI-powered wearable sensors offer safer alternative to invasive ICU blood pressure monitoring
We wanted to find a better way.
Mark
Why does an arterial catheter carry so much risk if it's been the standard for so long?
Mimi
Because it's a trade-off medicine has accepted. You're threading a tube into an artery to get perfect data. That invasion itself—the puncture, the foreign object sitting in the vessel—creates the possibility of bleeding, clotting, infection. It works, but the tool itself is a source of harm.
Mark
And the AI piece—how does it know what the blood pressure actually is from just electrical signals and blood flow?
Mimi
It's trained on data from patients where you have both the sensor readings and the arterial catheter readings side by side. The model learns the relationship between those signals and the actual pressure. Once it's learned that pattern, it can generate the waveform from the sensors alone.
Mark
Twenty-eight patients is a small sample. What happens if the larger trial doesn't replicate those results?
Mimi
That's the real test. Small initial trials often look promising. Larger, more diverse populations reveal where the model breaks down—different body types, different conditions, different medications. The team knows this. That's why they're moving to the next phase.
Mark
The hypertension angle seems like the bigger story than ICU monitoring. Is that where the real impact would be?
Mimi
Potentially, yes. The ICU application solves a problem for a small number of critically ill people. But hypertension affects hundreds of millions worldwide. If you could give someone continuous monitoring instead of a blood pressure check every six months, you change how they understand their own health. That's transformative.
Mark
What would continuous monitoring actually change about how people manage hypertension?
Mimi
Right now, most people don't know their blood pressure is high until a doctor tells them. With continuous data, you'd see how your pressure responds to stress, to exercise, to sleep, to food. You'd have real-time feedback. That's the glucose monitor model—it changed diabetes management because people could see cause and effect immediately.
Mark
And the researchers don't know what healthy blood pressure looks like across a normal day?
Mimi
Exactly. We monitor sick people intensively. But a healthy person's blood pressure throughout a typical day? That's largely unmeasured territory. The sensors could fill that gap.
Il Polso
- Every minute in an ICU, a patient's blood pressure can swing from dangerous heights toward organ-starving lows, and the only reliable tool to catch those swings has long been a catheter threaded into an artery — a device that bleeds, clots, and infects.
- A Johns Hopkins biomedical engineering team refused to accept that the instrument of safety should itself be a source of danger, and spent years building an alternative from chest sensors, finger sensors, and deep learning.
- Their MOSAIC system translates the body's electrical signals and blood flow patterns into continuous waveforms that, in a 28-patient trial, tracked closely enough with traditional arterial lines to suggest a genuine replacement may be within reach.
- The team is now scaling into larger ICU trials, racing to confirm whether early results hold under the full weight of clinical complexity.
- If they do, the horizon stretches far beyond hospital walls — toward a world where people with hypertension wear these sensors daily, gaining the same continuous self-knowledge that glucose monitors gave to people with diabetes.
For decades, the intensive care unit has demanded a difficult bargain: to know a critically ill patient's blood pressure in real time, clinicians have had to pierce an artery, accepting the risks of infection, clotting, and immobility as the cost of vigilance. Researchers at Johns Hopkins University are now challenging that bargain with MOSAIC, a wearable two-sensor system guided by artificial intelligence that reconstructs continuous blood pressure waveforms without breaking the skin. In early trials, the technology matched the precision of arterial catheters closely enough to suggest that medicine may be approaching a moment when the price of knowing need not be paid in harm — and when that knowledge might follow patients not just into the ICU, but through the ordinary hours of their lives.
Inside an intensive care unit, a thin catheter runs into a patient's artery — usually the arm or groin — delivering a continuous stream of blood pressure data to overhead monitors. It works. It also carries real danger: bleeding, infection, clotting, and the loss of any freedom to move. For decades, this has been the price of knowing whether a critically ill person's pressure is climbing toward stroke or falling toward organ failure.
Researchers at Johns Hopkins University spent years asking whether that price was truly necessary. Their answer is MOSAIC — two small sensors, one on the chest and one on a finger, paired with a deep learning model that reads the body's electrical signals and blood flow patterns and translates them into continuous blood pressure readings. In initial trials with 28 ICU patients, the waveforms matched those from traditional arterial catheters closely enough to suggest the technology might genuinely replace them.
The stakes are not abstract. Blood pressure swings in critically ill patients can be violent and fast. Too high, and stroke or heart attack looms. Too low, and the brain and organs begin to starve. Catching these shifts in real time is why arterial lines exist — but the catheters themselves cause harm, and patients pay that cost simply to be monitored.
MOSAIC sidesteps these complications entirely. Robert Stevens, chief of the Division of Informatics, Integration, and Innovation at Johns Hopkins Medicine, describes the goal plainly: to reconstruct waveform data in a way that is meaningful, accurate, and, above all, non-invasive. The team is now moving into larger trials, and if those results hold, the implications reach well beyond the ICU.
The researchers envision people living with hypertension wearing these sensors through ordinary days — at work, under stress, during exercise — the way people with diabetes wear continuous glucose monitors. Where medicine currently sees only occasional snapshots of blood pressure taken in a clinic, MOSAIC could offer a continuous, living portrait of how the body actually behaves across a lifetime.
Inside an intensive care unit, a patient lies tethered to machines. A thin catheter runs into an artery—usually in the arm or groin—delivering a continuous stream of blood pressure data to monitors overhead. It works. It also carries the weight of real danger: bleeding, infection, clotting. The patient cannot move freely. For decades, this has been the price of knowing, in real time, whether a critically ill person's blood pressure is climbing toward stroke or falling toward organ failure.
Researchers at Johns Hopkins University have spent the last several years asking a simpler question: what if you didn't have to pay that price?
Their answer is a system called MOSAIC—two small sensors, one placed on the chest and one on a finger, paired with an artificial intelligence model that reads the body's electrical signals and blood flow patterns and translates them into continuous blood pressure readings. In initial trials with 28 patients in Johns Hopkins' intensive care unit, the waveforms generated by the sensors matched those produced by traditional arterial catheters closely enough to suggest the technology might actually work as a replacement.
Carl Harris, a biomedical engineering PhD student who led the work, frames the problem plainly: patients in the ICU need constant blood pressure monitoring to catch dangerous swings before they cause harm. But the tool that provides that monitoring—the arterial line—is itself a source of harm. "We wanted to find a better way," he says. The research, published in Computers in Biology and Medicine, represents a step toward that alternative.
The stakes are not small. Blood pressure fluctuations in critically ill patients can swing wildly and dangerously. When pressure climbs too high, stroke and heart attack become real risks. Kidney damage follows. When it drops too low, the brain and vital organs begin to starve for blood. Catching these swings early, in real time, can mean the difference between recovery and catastrophe. This is why arterial catheters exist—they provide the continuous, granular data that allows clinicians to intervene before crisis becomes irreversible.
But the catheters themselves are not benign. The procedure to insert one carries infection risk. The line itself can cause clotting. Bleeding at the insertion site is common. Patients lose mobility, tethered to the monitoring equipment. The system works, but it works at a cost.
The MOSAIC approach sidesteps these complications entirely. The two sensors record what the heart is doing electrically and how blood is moving through the vessels. A deep learning model processes these signals and reconstructs them into a waveform—a continuous readout of blood pressure over time. Robert Stevens, chief of the Division of Informatics, Integration, and Innovation at Johns Hopkins Medicine, describes the achievement in measured terms: "We reconstruct waveform data in a way that's meaningful, accurate, reliable and, most importantly, non-invasive."
The initial validation was narrow—28 patients, one hospital—but the results were encouraging enough that the team is now moving into larger trials with more Johns Hopkins ICU patients. If those trials hold, the implications extend far beyond the intensive care unit. The researchers envision a future where people living with hypertension, one of the world's most common and deadliest conditions, could wear these sensors daily, the way people with diabetes wear continuous glucose monitors. The technology could transform how millions of people manage their health, moving from occasional blood pressure checks at a doctor's office to a continuous, real-time picture of how their bodies respond to work, stress, exercise, and rest. "We observe sick patients in the intensive care unit, but we have no idea what's going on with blood pressure in a healthy person who's just living their life," Stevens notes. That gap in knowledge—what happens to blood pressure day after day in ordinary life—could finally be filled.
Citazioni salienti
Patients in the ICU need continuous blood pressure monitoring to catch problems early, but it means an arterial line, which comes with a risk of bleeding, clotting, and infection.— Carl Harris, biomedical engineering PhD student, Johns Hopkins University
We observe sick patients in the intensive care unit, but we have no idea what's going on with blood pressure in a healthy person who's just living their life, going to work and being with their family.— Robert Stevens, chief of the Division of Informatics, Integration, and Innovation at Johns Hopkins Medicine
Vuoi la storia completa?
Leggi l'originale su Futurity ↗
Technology