Columbia unveils flexible high-density BCI chip

Columbia Engineering has detailed a brain-computer interface platform that compresses much of the traditional implant electronics chain into a single, mechanically flexible silicon chip. The device, known as BISC, or Biological Interface System to Cortex, is designed to sit below the dura on the brain surface and communicate wirelessly with an external relay station, combining sensing, stimulation, conversion, telemetry, and power functions on one implantable platform.

The chip integrates a 256 x 256 micro-electrocorticography array with 65,536 recording electrodes on a flexible CMOS substrate around 50 micrometres thick. From that matrix, the system can acquire up to 1,024 channels simultaneously and includes 16,384 stimulation channels. Wireless power transfer and a bidirectional data link remove the need for bulkier wired implant architectures that have long constrained brain-computer interface form factors.

The design was published in Nature Electronics and demonstrated in animal studies as well as intraoperative human recordings. The team reported chronic recordings for up to two weeks in pigs and up to two months in behaving non-human primates across sensory, motor, and visual cortical activity. That gives the platform a more substantial footing than a fabrication demonstration alone, even though it remains in the research domain rather than clinical deployment.

What stands out is the degree of integration. Conventional high-performance BCI systems have often depended on multiple modules, separate electronics stages, or larger implanted hardware assemblies to manage amplification, conversion, and communications. BISC pulls those functions close to the sensing interface itself. The reported wireless link, operating at around 100 Mbit/s, is intended to support a denser stream of neural data than lower-bandwidth implant links can comfortably sustain.

The architecture reflects a broader change in implantable electronics. Progress is no longer being driven by electrode count alone. Mixed-signal integration, power efficiency, packaging, wireless throughput, and mechanical compliance are being treated as one design problem. As implantable systems aim for more channels and more stable long-term recording, the cost of separating those functions across multiple hardware blocks rises quickly in size, complexity, and thermal burden.

That is pushing neural interfaces closer to application-specific system-on-chip design. Analogue front ends, stimulation drivers, data converters, digital control, telemetry, and power transfer need to coexist on a substrate that is thin enough for implantation and stable enough for chronic use. The requirements are unusually severe. Electrical performance has to be maintained in a structure that also has to tolerate tissue contact, motion, heat limits, and long operating periods without a conventional service path.

There is still a substantial engineering gap between an advanced laboratory platform and a broadly deployable medical product. Chronic biocompatibility, encapsulation, long-term drift, surgical repeatability, thermal management, and external communication security all remain active challenges. None of them are peripheral. They sit directly in the route from promising semiconductor design to usable clinical hardware.

Even so, the direction is becoming unmistakable. Implantable neurotechnology is moving toward thinner, denser, and more integrated electronics with fewer compromises between scale and form factor. BISC shows how far that shift has already progressed. In medical electronics, the centre of gravity is moving away from loosely assembled implant subsystems and toward highly integrated silicon platforms built to survive inside the body while handling workloads that once belonged to racks of external instrumentation.