Summary: Researchers at Princeton University have developed a three-dimensional, programmable bio-hybrid device that integrates living brain cells with microfabricated electronics. Unlike earlier “brain-on-a-chip” efforts that cultured neurons on flat surfaces, this approach uses a flexible, microscopic metal mesh as a scaffold so tens of thousands of neurons can grow around and through the sensors. The resulting 3D biological neural network can be stimulated and recorded over long periods and trained to recognize complex electrical patterns, pointing toward much more energy-efficient forms of computation.
The study, published in Nature Electronics on Apr. 23, shows that integrating soft, flexible electronics inside a living neural culture lets researchers monitor and shape neural connectivity at a finer scale than before. The device enabled chronic recording and stimulation across multiple planes of the network for more than six months, and the team demonstrated pattern recognition across both spatial and temporal pulse patterns by combining neural training with algorithmic readout.
Key Facts
- From the inside out: The system uses a three-dimensional mesh of microscopic metal wires and electrodes coated with an ultra-thin epoxy. The thin coating preserves flexibility so the mesh can interface directly with delicate neural tissue as neurons grow around and through it.
- Long-term stability: The researchers tracked electrical activity and connectivity changes for over six months, observing how synaptic connections formed, weakened, and strengthened in response to stimulation.
- Pattern recognition: By applying targeted electrical stimulation to induce synaptic plasticity and using an algorithmic decoder to interpret resulting activity, the team distinguished distinct spatial and temporal electrical patterns.
- Energy efficiency: One motivation is closing the energy gap between biological and artificial neural systems. The authors note that the human brain performs similar computations with orders-of-magnitude less energy than current AI hardware, an advantage this bio-hybrid approach seeks to exploit.
The device was fabricated with microfabrication techniques to form a three-dimensional scaffold of electrodes that serve both as sensors and stimulators. Tens of thousands of neurons were cultured onto this mesh to create a dense, volumetric network capable of computation. Because the electronics are integrated into the network rather than only probing it from the outside, the team was able to stimulate and record from multiple depths and track evolving connectivity maps over months.
In experiments, researchers used electrical stimulation protocols to tune connectivity strengths—strengthening or weakening synapses between targeted neurons through synaptic plasticity—and then trained a computational decoder to read patterns of action potentials. The system successfully discriminated between pairs of distinct spatial stimulation patterns and between different temporal stimulation sequences, demonstrating that 3D cultured networks can serve as a substrate for reservoir-style biocomputing.
The research was led by Tian-Ming Fu (Assistant Professor of Electrical and Computer Engineering and Omenn-Darling Bioengineering Institute), James Sturm (Stephen R. Forrest Professor of Electrical and Computer Engineering), and Kumar Mritunjay (postdoctoral researcher in electrical and computer engineering). While the primary aims include studying neural development and disease, the team highlights potential applications in energy-efficient computing and in developing neural interfaces for medical therapies.
Key Questions Answered
Q: Is this a “living computer”?
A: In a practical sense, yes. The device forms a three-dimensional biological neural network (3D-BNN) in which living neurons act as the information-processing elements, and the micro-mesh provides sensing and stimulation points. Together, they form a hybrid system that can be programmed to perform specific tasks such as pattern recognition.
Q: How are the neurons trained within the mesh?
A: Training relies on electrical stimulation to alter synaptic strengths—synaptic plasticity—between neurons. After stimulation protocols change connectivity, decoding algorithms read the network’s electrical output to classify patterns, effectively tuning the network for target behaviors.
Q: Could this technology be used to treat brain diseases?
A: That is a long-term goal. Because the mesh is flexible and designed to mimic aspects of the brain’s three-dimensional architecture, it could inform the development of implants or interfaces that communicate with neural tissue in a more natural, spatially distributed way, with potential therapeutic applications for neurological disorders.
Editorial Notes
- This article was edited by a Neuroscience News editor.
- The journal paper was reviewed in full.
- Additional context was added by staff.
About this neurotech research news
Author: Scott Lyon
Source: Princeton University
Contact: Scott Lyon – Princeton University
Image credit: Neuroscience News
Original Research: Closed access. “A three-dimensional micro-instrumented neural network device” by Kumar Mritunjay, James C. Sturm & Tian-Ming Fu. Nature Electronics. DOI: 10.1038/s41928-026-01608-1
Abstract
A three-dimensional micro-instrumented neural network device
Three-dimensional (3D) cultured neural networks that reproduce aspects of the brain’s structure and computation could advance brain-inspired computing, artificial intelligence, and our understanding of neural development and disease. Creating stable interfaces between such networks and electronic devices is challenging, which has limited the potential of 3D neural cultures.
Here, the authors report a 3D micro-instrumented neural network device that integrates a flexible electronic sensor and stimulator array with a 3D cultured neural network. The device supports recording of action potentials across multiple planes for at least six months, enabling quantitative tracking of evolving connectivity maps and pharmacological responses. It also supports chronic electrical stimulation used to train networks by tuning connectivity strengths, creating a reservoir neural network useful for biocomputing applications.