Every NICU Can See a Heartbeat. Almost None Can See a Seizure.

AI seizure detection is turning neonatal brain monitoring from a specialist service into a bedside product, and three companies are racing to own the new layer.

Walk into any NICU in America and you will see continuous readouts of heart rate, respiration, and oxygen saturation on every isolette. The organ the NICU exists to protect, the newborn brain, is the one most units still cannot watch in real time.

That gap is now the most active startup lane in neonatal care. In the past ten months, Ceribell won FDA clearance for the first neonatal AI seizure-detection algorithm, CergenX carried an FDA Breakthrough Device Designation into the agency's priority-review track, and NeuroBell raised $5.5 million with a US health system in the round to bring a portable neonatal EEG monitor to market. Three companies, three different wedges, one shared bet: the constraint in neonatal brain monitoring was never the hardware. It was the interpretation. And interpretation is exactly what AI is commoditizing.

PHD's thesis: neonatal neuromonitoring is following the same arc that pulse oximetry followed a generation ago, from specialist procedure to standard-of-care vital sign. The companies that win will be the ones that make brain monitoring usable by ordinary NICU staff at 2 a.m. without a neurophysiologist awake anywhere in the time zone. The next 18 months of FDA decisions and hospital pilots will decide whether this becomes a platform-extension story, a screening story, or a new-device story.

The Bottom Line

• Neonatal seizures occur in roughly 1 to 5.5 per 1,000 live births in high-income countries, and the majority are subclinical: the baby shows nothing outward while the brain seizes. In infants with moderate-to-severe hypoxic-ischemic encephalopathy undergoing therapeutic hypothermia, 62% of those with seizures have events detectable only electrographically.
• Clinical observation fails at this task. In one study of 526 electrographic seizures, experienced clinicians watching the infant identified just 9%.
• The gold standard, continuous video EEG, exists mostly on paper. Guidelines from the American Clinical Neurophysiology Society call for it in high-risk neonates, but many units have no access to continuous EEG at all, and many that do have no access to real-time interpretation. The common fallback, amplitude-integrated EEG, has reported seizure-detection sensitivity as low as 30% on a single channel.
• Ceribell, the public comp, put numbers on the opportunity: a $400 million neonatal and pediatric market layered onto its existing $2 billion addressable market, roughly 200 NICU hospitals already in its installed base, and about 280 children's hospitals in its target universe. Nine-month 2025 revenue was $64.3 million, up 37%.

The Clinical Gap Is an Economics Gap

The reason neonatal brain monitoring stalled for decades is not that anyone doubted its value. Untreated seizures compound brain injury in exactly the infants, cooled HIE babies and extremely preterm infants, whose neurological outcomes drive lifetime cost. The problem is the workflow. Conventional continuous EEG requires a technologist to apply a full electrode montage, a reading service to watch the trace, and a neurophysiologist to interpret it. That stack exists at large academic children's hospitals and almost nowhere else.

So most NICUs bought the affordable compromise: amplitude-integrated EEG, a simplified trend monitor that bedside staff can apply, with Natus's Olympic Brainz Monitor as the long-standing incumbent product. The compromise is real. Reported aEEG seizure-detection sensitivity ranges from about 31% to 90% depending on configuration and reader, and single-channel setups sit at the bottom of that range. A monitoring standard that misses a third to two-thirds of seizures in the highest-risk patients is not a solved market. It is an incumbent position waiting to be attacked.

The attack vector is the same one reshaping radiology and cardiology: move the interpretation into the algorithm, and the scarce specialist stops being the bottleneck.

Three Companies, Three Wedges

The Irish cluster is not a coincidence. UCC's INFANT Centre spent two decades building the neonatal EEG datasets and algorithm research that this entire category now runs on, and it has now seeded two venture-backed challengers from the same lab bench.

What This Means for the Incumbents

Natus built the aEEG installed base, and the Natus story of 2026 has been consolidation: after the ARCHIMED take-private and the split into Natus Neuro and Natus Sensory, the Sensory side spent the first quarter acquiring its way across the neonatal workflow stack (Keriton, TheraB Medical). The neuromonitoring lane sits on the Neuro side of that split, and it now faces a choice familiar to every incumbent whose product is the compromise: defend aEEG, build or buy an AI interpretation layer, or watch the category get redefined around it. Given how this year has gone in neonatal M&A, do not be surprised if the answer is an acquisition. At least one of the three companies above is a plausible target before it becomes a plausible threat.

For NICU operators and investors, the sequencing matters. Ceribell is already selling. NeuroBell's clearance decision lands within months. CergenX's TAP pathway is a slower, deeper regulatory bet. By mid-2027 we will know whether neonatal brain monitoring is one market or three: continuous seizure surveillance in the level III/IV NICU, universal screening at the delivery hospital, and the portable bedside layer connecting them.

What We're Watching

• NeuroBell's FDA decision and first US commercial deployments through Parkview Health
• Ceribell's neonatal ramp: whether NICU adoption shows up in 2026 revenue and whether the children's-hospital pipeline converts
• CergenX's first pivotal-study milestones under the TAP program, and whether new financing follows the regulatory momentum
• Any move by Natus Neuro to answer the AI interpretation gap, organically or by acquisition
• Guideline evolution: ACNS's continuous-EEG recommendations gave this category its clinical mandate; payer and quality-measure recognition would give it a reimbursement engine

The newborn brain has been the NICU's blind spot for fifty years. The technology to close it now exists, the FDA is clearing it, and capital is arriving. What happens next is a market-structure question, and we will keep mapping it.

Sources

• MedTech Dive: Ceribell wins FDA clearance for seizure-detection algorithm in neonates
• Medical Device Network: CergenX wins FDA breakthrough designation for neonatal EEG assessment tool
• MassDevice: NeuroBell raises $5.5M for AI-powered neonatal brain monitor
• Pediatric Research: Electrographic monitoring for seizure detection in the neonatal unit
• Citeline Medtech Insight: Ceribell's Neonatal US Clearance Sets Up 2026 Growth