Artificial neurons interface with brain cells

- Northwestern engineers reported in April 2026 that printed artificial neurons directly activated living mouse brain cells, showing a real bioelectronic interface in tissue. - The devices use printed graphene and MoS2 nanosheet memristors, match physiological spike timing, run past 10^6 cycles, and can fire up to 20 kHz. - It matters because brain-like hardware could cut AI power use and make softer, more natural neural implants possible.

Artificial neurons are usually good at one thing — imitating a cartoon version of how brain cells fire. The hard part is talking to actual living neurons in a way the tissue will recognize. That gap matters for two big reasons: brain implants still struggle to feel natural inside the body, and AI hardware still burns absurd amounts of power. What changed in April 2026 is that a Northwestern team showed printed artificial neurons can send realistic electrical spikes into mouse brain tissue and trigger responses from real cells. ### What did they actually build? They built flexible, fully printed devices on polymer substrates using graphene electrodes and networks of molybdenum disulfide — MoS2 — nanosheets. Those printed components act like memristors, which are circuit elements whose behavior depends on their recent electrical history. That matters because real neurons are not simple on-off switches. They have timing, thresholds, bursts, pauses, and all kinds of messy dynamics. The Northwestern group was trying to reproduce more of that mess instead of flattening it into a toy model. (news.northwestern.edu) ### Why is “printed” a big deal? Most neuromorphic hardware still lives on rigid silicon. Brains do not. Brain tissue is soft, curved, wet, and constantly moving a little. A printed device can be made flexible, lower cost, and easier to scale over larger areas. Basically, if you want electronics to sit near nerves or brain tissue without behaving like tiny shards of glass, soft printable materials are a much better starting point. Northwestern framed the whole thing as a route toward flexible brain-machine interfaces, not just another lab demo chip. (nature.com) ### What makes these spikes more realistic? The key is that the circuits do more than simple integrate-and-fire behavior. The paper says the printed neuristors reproduced first-, second-, and third-order spiking complexity, including spike latency, tonic firing, class 1 excitability, tonic bursting, and phasic dynamics. In plain English, the waveform has richer timing structure — closer to how biological neurons actually signal. That richer behavior is why the output could land inside physiological timescales instead of just blasting tissue with crude pulses. (news.northwestern.edu) ### Did they really talk to living brain cells? Yes — but in a limited, preclinical way. The test was done on mouse cerebellar slices, not in a living animal and definitely not in humans. The artificial neurons generated spike waveforms that stimulated Purkinje neurons in that tissue. Northwestern described this as real neurons responding to the printed devices, which is the important threshold here: the biological cells treated the synthetic signal as something meaningful enough to answer. (nature.com) ### Why does this matter for implants? Current neural implants can stimulate tissue, but they often do it with electronics that are mechanically mismatched and electrically blunt. The dream is an interface that “speaks neuron” more fluently — softer materials, lower voltages, more natural timing. If that gets good enough, you can imagine better prosthetics for hearing, vision, or movement, because the device would be sending signals in a form the nervous system is more willing to interpret. (news.northwestern.edu) Northwestern explicitly points to neuroprosthetics and brain-machine interfaces as targets. ### Why bring AI into this? Because the brain is insanely efficient. Northwestern highlights that brains are about five orders of magnitude more energy efficient than digital computers for comparable kinds of adaptive processing. So this is not just a medical-device story. It is also a computing story — one where richer neuron-like hardware could do useful work without the giant power bill that comes with today’s data-hungry AI systems. (news.northwestern.edu) ### What’s the catch? The catch is scale and durability in the real world. The paper reports stable operation over more than 10^6 cycles and tunable frequencies up to 20 kHz, which is solid for a device paper. But a useful implant has to survive far longer, inside living tissue, without inflammation, drift, or signal loss. And a useful computing platform has to connect lots of these units into systems that do something competitive. This result clears an important scientific hurdle, but it does not mean biohybrid brain implants are around the corner. (news.northwestern.edu) ### Bottom line This is one of those advances that sounds sci-fi but is actually pretty concrete. A printed artificial neuron fired a biologically realistic signal, and living mouse brain cells responded. That does not give us synthetic brains. But it does move the field from “brain-inspired” electronics toward electronics that can genuinely interface with the brain. (news.northwestern.edu) (nature.com)

Get your own daily briefing

Scout delivers personalized news, insights, and conversations tailored to your role and industry.

Download on the App Store

Shared from Scout - Be the smartest in the room.