Cleveland Clinic is putting artificial intelligence to work in the neurological ICU, training a system to read brain waves and flag seizures in seconds. The project teams the health system with startup Piramidal and is designed to turn a full day of EEG recordings, a job that can keep specialists busy for hours, into near-instant alerts at the bedside.
According to Axios, the Piramidal-powered tool can scan a 24-hour EEG in just a few seconds and will roll into the Cleveland Clinic's Neurological Institute as part of a larger tech upgrade. Leaders say that kind of speed could help clinicians catch seizures, strokes and other acute brain emergencies earlier than current workflows typically allow.
Wired reports that Piramidal trained its model on a mix of public datasets and proprietary hospital records, pulling in nearly a million hours of EEG monitoring to teach the system what abnormal activity looks like. The company pitches it as a foundation model for the brain that can adapt to different patients' neural signatures, and representatives say formal, peer-reviewed accuracy data will follow later.
How the Clinic will test it
According to WKYC, Cleveland Clinic will first run the AI system in parallel with human monitoring, so clinicians can verify every alert before it influences patient care. During that trial phase, the tool is set up to push live seizure notifications to the medical team while engineers and neurologists fine-tune sensitivity settings and alert thresholds.
Sepsis pilot shows a path
Cleveland Clinic leaders point to a recent sepsis detection pilot as proof that well-integrated AI can actually move the needle. The Clinic's newsroom reports that a Fairview Hospital pilot covering more than 3,330 patients produced a tenfold decrease in false alerts, a 46 percent rise in identified sepsis cases and a sevenfold increase in alerts that fired before antibiotics were given, compared with legacy tools. Officials say those results helped shape plans for broader AI use across the system.
Neurologists call it a co-pilot
Physicians emphasize that the new system is meant to support human experts, not sideline them. "This type of thing is time-consuming. It is subjective, and it is experience- and expertise-dependent," Imad Najm told Wired, which is why the rollout will be gradual and human oversight will stay central to treatment decisions.
Privacy and ethical questions
Beyond raw accuracy, the work raises uncomfortable questions about who owns and controls brain data and how it should be stored, shared and reused. Ethicists argue that guardrails should be in place before brain-reading models become routine in hospitals. The American Hospital Association's market-scan analysis recommends bringing clinicians, ethicists and legal experts into the development process early to help hospitals avoid missteps as they adopt these tools.
Clinic officials told WKYC they hope to have the seizure-spotting AI ready by the end of summer, with validation continuing alongside human reads as the Neurological Institute moves into a new building early next year. If the tests hold up, the shift could expand access to real-time brain monitoring beyond specialized centers and shorten the time it takes to spot life threatening neurologic events.
