A Better Way to Listen to the Brain

Picture eavesdropping on conversations happening in three rooms at once, but your only microphone is bolted to a single spot in the hallway. You would catch fragments and miss the rest.

That is roughly the challenge neuroscientists face when studying the brain. The regions involved in memory, movement, emotion and other functions communicate constantly, and a recording device placed in the wrong spot can miss much of that activity.

Most brain implants use tiny electrodes to record the electrical signals produced by nearby brain cells. But these devices often come in standard shapes that do not match the brain's complex anatomy, causing them to miss important activity. A team of USC researchers has flipped that approach, developing a way to design implants that match the shape and depth of the specific brain regions being studied.

For Dong Song, whose research focuses on memory functions, the new approach could provide a more complete picture of how memories form and are retrieved.

"We can record multiple regions simultaneously, which is quite useful," said Song, corresponding author of the paper, titled "Design of Conformal Planar Penetrating Multi-Electrode Arrays for Recording Multiple Brain Regions," and an associate professor of neurological surgery at the Keck School of Medicine of USC and of biomedical engineering at the USC Viterbi School of Engineering. "That's because memory formation involves multiple regions."

The research was published on August 12, 2026, in the journal Progress in Biomedical Engineering. The first author is Huijing Xu, a senior biomedical engineer at Bioventus and USC Viterbi alumna who earned her Ph.D. at the Alfred E. Mann Department of Biomedical Engineering from the USC Neural Modeling and Interface Lab directed by Song; Zihan Jin and Zhouxiao Lu, USC Viterbi BME doctoral students advised by Song; Yingyi Gao and Jingjing Liang, USC Viterbi BME students supervised by Professor Ellis Meng; Yan Gong, a postdoctoral scholar in the USC Biomedical Microsystems Laboratory; Kee Scholten, director of the USC Polymer Implantable Electrode Foundry (PIE); Xiling Li, a postdoctoral fellow at Stanford Medicine; Lu Chen, a professor of neurosurgery and of psychiatry and behavioral sciences at Stanford Medicine; and Meng, the Shelly and Ofer Nemirovsky Chair in Convergent Bioscience, a professor in biomedical engineering and the Ming Hsieh Department of Electrical and Computer Engineering.

Why Standard Devices Miss the Signal

The brain is not built like a filing cabinet, with information stored in neat, evenly spaced drawers. Its structures curve and fold, and the cells researchers want to study sit at irregular depths that vary by brain region, species and even individual animals.

Most standard devices space their recording points evenly in a grid. That often means many of those points land where few or no useful cells are located.

"It doesn't make sense to build a design that completely ignores anatomy," said Meng, who also serves as USC Viterbi's vice dean for technology innovation and entrepreneurship.

Building Devices That Bend to the Brain

Rather than focusing on a single device, the paper outlines a process for designing brain-recording devices tailored to specific research needs.

Neuroscientists first identify the brain regions and cells they want to study. Engineers then use detailed maps of the brain's anatomy, along with information about the safest and most practical way to implant the device, to determine how many slender arms it should have, how long they should be and where to place the recording points along each arm.

Meng's lab then builds a prototype; Song's lab tests it and examines recorded signals and tissue samples to confirm that it reaches the right regions. The design is revised and tested again until it works.

We want "to loop neuroscientists in at the very beginning," Meng said, "and then to build devices that take this into account, iterating with the neuroscientists as needed until we get it right."

Added Song: "This iterative co-design strategy is rarely feasible in commercial settings but is essential for the success of neuroscience research and neural engineering applications."

The devices are also made of soft, flexible polymers rather than rigid materials. A stiff implant placed in brain tissue that moves and shifts can trigger scarring, which blocks the very signals researchers want to capture. Softer materials tend to keep tissue healthier during recordings lasting days, weeks or months.

The paper documents 21 such designs, though Meng's lab and the USC PIE have produced hundreds of custom arrays over the years. Most of the 21 were built for rats and mice. Seven of the devices described in detail either recorded strong signals from multiple parts of the hippocampus, a brain region central to memory, or offered early evidence that the designs worked.

Customization first

"The best way is to customize the design," Song said. "Your recording electrodes conform to the brain region you're trying to record from."

The devices remain research tools for animals, not treatments for people. Before any human studies can begin, researchers must test the approach consistently in larger animals and build the safety record regulators will require.

"The hope is that some of these tools get adopted into regular research practice," Meng said.

For Song, these devices are one part of a much larger goal: developing a memory prosthesis. Such a device could someday read brain activity tied to memory and help restore it in people who have lost that ability because of Alzheimer's disease, traumatic brain injury or another disorder. Reaching that goal will require exactly what these devices could provide - long-term recordings from several memory-related brain regions at once.

Song has separately proposed pairing such interfaces with artificial intelligence to decode how everyday experiences become lasting memories, an effort he has described as central to his lab's mission.

"I feel like this is one step closer," Song said, "to building a real memory prosthesis for future human study."

Published on October 6th, 2026

Last updated on October 6th, 2026