Researchers have developed a tiny computer chip that can model brain activity at speeds close to those of the human brain itself.
The breakthrough could make future brain-computer interfaces, medical imaging, and surgical guidance systems much faster and more energy efficient.
The study, led by Professor Yang Yuchao of Peking University in collaboration with researchers from the Shanghai Institute of Microsystem and Information Technology of the Chinese Academy of Sciences, was published in the journal Science.
Many modern technologies rely on computer models that simulate how complex systems change over time.
These models, known as neural dynamical systems, combine artificial intelligence with mathematical equations to predict and describe changing processes.
They are widely used in areas such as brain imaging, three-dimensional brain reconstruction, and other scientific applications.
However, these calculations are extremely demanding. Traditional computers repeatedly move data between memory and the processor while constantly checking for errors and adjusting calculations.
This process takes time and consumes large amounts of energy, making it difficult to perform detailed brain modeling in real time.
The newly developed chip tackles this problem in a different way. Instead of constantly transferring data back and forth, it performs many of the calculations directly where the data are stored. This approach, called in-memory computing, greatly reduces delays and energy use.
At the heart of the new chip are tiny electronic components called phase-change memristors. These devices can both store information and perform calculations, helping eliminate one of the biggest bottlenecks in conventional computing.
Although the chip itself occupies just 0.28 square millimeters—smaller than a grain of rice—it delivers remarkable performance. Operating at 50 megahertz, it completed neural dynamics calculations between about four and 36 times faster than today’s leading application-specific computer chips while using roughly 12 to 25 times less power.
The researchers also tested the chip on one of the most demanding tasks in medical computing: reconstructing the surface of the human brain. This involves creating highly detailed three-dimensional models of the brain’s outer layers, including the intricate folds of the gray and white matter.
The chip successfully generated smooth, accurate 3D brain surface models in real time. Compared with one of the world’s most powerful graphics processors, the NVIDIA A100, it completed some reconstruction tasks up to 478 times faster while maintaining high accuracy.
The ability to build detailed brain models in less than 10 milliseconds could have important medical benefits. Doctors could eventually use the technology to support real-time surgical navigation, helping them visualize brain structures during delicate procedures. It could also improve brain-computer interfaces that translate brain signals into commands for external devices, making them faster and more responsive.
Beyond surgery and assistive technologies, the chip may also help scientists create digital models of the brain, sometimes called digital brain twins, which could improve research into neurological conditions such as Alzheimer’s disease and Parkinson’s disease.
While more development is needed before the technology reaches hospitals or commercial devices, the researchers believe this new chip represents an important step toward real-time brain modeling that is both fast and energy efficient, bringing advanced brain technologies closer to everyday clinical use.
