Harvard Professor Donhee Ham Delivers Special Lecture at Choi Jonghyun Academy: "New Possibilities for Neuromorphic Computing"

Overcoming the Precision-Scale Dilemma
with Semiconductor-Based Neural Signal Measurement Technology
Large-Scale Parallel Analysis of Synaptic Signals
Opens New Possibilities for Implementing Neuromorphic Semiconductors

A technology that can simultaneously measure the electrical signals inside thousands of neurons using a semiconductor chip has been developed. This achievement, which secures both single-cell precision and large-scale neural network analysis capabilities, is regarded as opening up new possibilities not only for neuroscience research aimed at uncovering the brain's information processing principles, but also for the development of next-generation neuromorphic computing technology.

The Choi Jonghyun Academy announced on June 1 that it held a special lecture on May 28 at the Korea Higher Education Foundation building in Gangnam-gu, Seoul, inviting Donhee Ham, the TDK Professor of Electrical Engineering and Applied Physics at Harvard University.

At the Korea Foundation for Advanced Studies building in Gangnam, Seoul, on the 28th of last month, Donhee Ham, Shelby C. Davis Professor of Engineering and Applied Sciences at Harvard University, gave a special lecture hosted by the Cho Jonghyun Academy, speaking on the topic of "Reconfiguration of the Brain." Cho Jonghyun Academy

In the lecture, Professor Ham introduced the latest research achievements aimed at uncovering the information processing principles of the human brain through the convergence of neuroscience and semiconductor engineering and implementing them into next-generation computing technology. The core of the lecture was the 'iMEA (Intracellular Microelectrode Array),' a semiconductor-based neural signal measurement platform developed by Professor Ham's research team. This system is capable of simultaneously measuring electrical signals generated inside living neurons on a large scale, and is attracting attention as a new approach that can overcome the long-standing "dilemma of precision and scale" in neuroscience.

While conventional technologies required a trade-off between precise measurement of a small number of neurons and large-scale neural network observation, the research team took on the challenge of achieving both by designing a circuit that optimizes the structure and electrical interaction between cells and electrodes. As a result, they developed a technology that can simultaneously measure not only action potentials (AP), but also postsynaptic potentials (PSP), which have been difficult to capture with traditional electrode arrays, on a large scale.

The key lies in the microhole-structured electrodes fabricated on top of the semiconductor chip. When a cell settles over a hole, the area around the electrode is sealed, and by applying a minute current, the cell membrane is temporarily opened, allowing measurement of intracellular electrical signals. This technology maintains single-cell precision while enabling observation of large-scale neural networks at the same time.

Professor Ham stated, "Whereas previously, one could only observe with precision or at large scale, our goal is to achieve both simultaneously," adding, "It is a new approach that allows analysis of large-scale neural networks while maintaining information at the single-cell level."

During the lecture, Professor Ham emphasized that to understand the brain's information processing principles, it is necessary to look at synapses—the connections between neurons—rather than just the neurons themselves. He explained, "Memory and learning in the brain occur through synapses," and, "Understanding the strength and changes of synaptic connections is key to elucidating the information processing principles of the brain."

The iMEA system developed by Professor Ham's research team integrates approximately 4,000 electrodes on a single chip. In experiments with cultured mouse neurons, intracellular signals were measured simultaneously from an average of 3,600 electrodes (about 90%), with a maximum of up to 3,900 electrodes (97%) successfully capturing signals. Based on this, the team succeeded in reconstructing a functional synaptic connectivity map of about 70,000 connections. This achievement significantly expands the limitations of previous technologies in terms of both the scale of synaptic networks and the precision of single-neuron information. The team also confirmed quantized patterns of PSP amplitude, proving that the measured signals indeed reflect actual synaptic activity. Professor Ham stressed that this research not only advances brain science but also provides important implications for next-generation computing technology.

Professor Ham revealed that his team is currently advancing from in-vitro neural cell culture stages to in-vivo studies that measure neural signals in the brains of living animals. However, he noted that challenges such as micro-vibrations caused by brain tissue movement and immune responses during electrode insertion remain unresolved. He explained that the team is focusing its research efforts on overcoming these technical challenges to establish a platform capable of stably measuring neural signals over long periods.

In the subsequent discussion, Changhwan Shin, professor at the Department of Electrical and Electronic Engineering at Korea University, served as moderator, joining Professor Ham in discussing the technological and industrial potential of neuromorphic semiconductors. The discussion addressed not only the implications that the learning and memory mechanisms of the human brain may offer for next-generation semiconductor design, but also new computing architectures in the age of artificial intelligence as a major topic.

Professor Ham assessed that neuromorphic semiconductors currently being discussed in industry have not yet fully implemented the operating principles of the human brain. However, he stressed that basic research aimed at uncovering the brain's information processing principles will serve as the foundation for future innovations in computing technology. He stated, "The brain is structured such that memory and computation occur simultaneously within a single network, which is fundamentally different from the current Von Neumann architecture," and added, "If we can build even more precise maps of synaptic connections, it will become possible to design next-generation neuromorphic chips based on that." He continued, "The ultimate goal is to understand how the brain learns, remembers, and makes judgments," and concluded, "We will continue research to uncover the brain's information processing principles through the convergence of neuroscience and semiconductor engineering, and to connect this understanding to future computing technology."

This content was produced with the assistance of AI translation services.