This AI-powered chip can offer a deeper look at faulty brain activity

Source: Northeastern Global News, by Katya Poltorak (September 30, 2026). AI-generated summary by biochip.com, published October 1, 2026. Not independently reviewed.

Northeastern University electrical and computer engineering professor Aatmesh Shrivastava is developing a low-power chip that records brain activity beneath the scalp and analyzes it on the device with artificial intelligence. Supported by the National Institutes of Health, the system is designed for continuous monitoring over days or months rather than the brief recording windows common in standard electroencephalogram tests. Its immediate goal is to identify personalized electrical signatures linked to seizures and possible triggers such as missed medication, stress, food, or sleep disruption. If it works as intended, the platform could give clinicians a more detailed view of how abnormal brain activity changes over time.

Getting Closer to Brain Signals

An electroencephalogram, or EEG, records the brain's electrical activity through electrodes placed on the scalp. These measurements capture the combined activity of large populations of nerve cells, but signals can overlap, much like several people talking at once over a phone line.

Shrivastava's proposed device would measure signals under the scalp, closer to their source. That placement is intended to improve signal quality while avoiding the need for conventional scalp EEG sessions, which typically last 20 to 90 minutes.

Why Continuous Recording Changes the Picture

Many neurological events do not appear on command during a clinical test. A device that can collect EEG data over days or months could reveal transitions into and out of abnormal activity, along with connections to sleep, behavior, and events that happen before a seizure.

Long-term monitoring also creates a practical engineering problem: power. An implanted or wearable system cannot rely on frequent charging, so the chip builds on Shrivastava's earlier work in low-power and self-powering electronics. The aim is to make extended recording feasible without treating the brain monitor like a device that must be plugged in each night.

AI Built Into the Chip

Recording continuously produces a huge volume of data. The chip's machine-learning system is meant to sort that stream in real time by breaking each EEG recording into measurable features, such as the height, speed, and shape of its waves.

It can then connect combinations of those features with recorded events, including seizures. Over time, the on-chip network would store and refine a set of markers, meaning telltale signal patterns that may differ by person, seizure type, and location in the brain.

This local analysis distinguishes the planned device from continuous subscalp systems that primarily collect sensor data. Shrivastava compares its role to a glucose tracker: it would inform users and physicians about patterns rather than intervene directly to stop seizures.

Why This Matters

Epilepsy can follow individual rhythms across days or months, and seizure symptoms can vary substantially. A personalized record could help clinicians determine whether someone is actually seizure-free, identify periods of potentially higher risk, and investigate whether controllable factors line up with changes in brain activity.

The same approach may eventually help researchers study disrupted signaling in conditions including Alzheimer's disease and amyotrophic lateral sclerosis, also called ALS. Those possibilities remain future applications, not demonstrated clinical uses.

Next Steps

The device is still under development. Once the chip operates as intended, the team plans to test the prototype in rats or macaques at the University of Florida, an early step toward determining whether long-term, AI-assisted subscalp monitoring can reliably identify meaningful neurological patterns.

Topics: Brain & neurological health

Start here: What is a biochip?