A semiconductor-based technology capable of reading electrical signals from thousands of neurons simultaneously is bringing scientists closer to understanding how the human brain processes information.
Researchers at Harvard University have developed a new platform that combines neuroscience and semiconductor engineering, a breakthrough that could accelerate advances in next-generation artificial intelligence (AI) and neuromorphic computing systems.
The development was highlighted during a special lecture hosted by the Chey Institute for Advanced Studies in Seoul on May 28, where Professor Donhee Ham of Harvard University presented the latest progress in brain-inspired computing technologies.
At the center of the research is a system known as the Intracellular Microelectrode Array (iMEA), a semiconductor-based platform designed to record electrical activity directly from inside living neurons on a large scale.
For decades, neuroscience researchers have faced a trade-off between precision and scalability. Patch-clamp technology can measure intracellular signals with exceptional accuracy but is limited to only a small number of neurons at a time. Conventional Microelectrode Arrays (MEAs), meanwhile, can monitor thousands of neurons simultaneously but are generally restricted to recording extracellular activity.
“Traditional approaches forced researchers to choose between seeing more neurons and seeing them in greater detail,” Ham said. “Our goal was to achieve both single-cell resolution and large-scale neural network analysis simultaneously.”
To accomplish this, the research team integrated microscopic hole-shaped electrodes onto a semiconductor chip using advanced fabrication techniques. When neurons grow on the chip, they form a tight seal around the electrodes. By applying a brief electrical pulse, researchers can temporarily open the cell membrane and directly access intracellular electrical signals.
A key challenge involved optimizing the electrical interaction between neurons and electrodes through sophisticated circuit design. The resulting system allows large-scale intracellular recordings that were previously difficult or impossible using conventional electrode array technologies.
Experimental results demonstrated the platform’s capabilities. Using cultured mouse neurons, the researchers successfully obtained intracellular recordings from more than 3,600 of approximately 4,000 electrodes on average. The system achieved a maximum recording success rate of about 97%.
The platform also enabled researchers to reconstruct a functional synaptic connectivity map containing roughly 70,000 neural connections, allowing near real-time analysis of communication pathways and signal flow within neural networks.
Ham emphasized the importance of synapses—the connections between neurons—in understanding brain function.
“Learning and memory are determined not by individual neurons themselves, but by synapses and the way neurons are connected,” he said. “Understanding how synaptic strength changes and how those connections evolve is essential to uncovering the principles of information processing in the brain.”
Researchers and industry experts believe the implications extend well beyond neuroscience.
Major technology companies worldwide are investing heavily in neuromorphic computing, a field that seeks to replicate the brain’s structure and learning mechanisms in semiconductor hardware. Unlike conventional AI processors that focus primarily on maximizing computational power, neuromorphic chips aim to perform learning and inference tasks using dramatically lower energy consumption.
Because the iMEA platform can analyze biological neural networks at extremely high resolution, it may provide critical insights for the design of future brain-inspired semiconductor architectures.
“Ultimately, our objective is to understand how the brain processes information and translate those principles into engineering systems,” Ham said. “The convergence of neuroscience and semiconductor technology has the potential to reshape the future direction of artificial intelligence.”
As global competition intensifies in AI hardware and next-generation computing technologies, researchers view advances such as iMEA as an important step toward bridging the gap between biological intelligence and machine intelligence.
