Imec ITF World 2026: Barun Dutta on Convergence of Neuroscience and Semiconductors
Why It Matters
These ultra‑high‑density probes turn the brain into a data source capable of training AI models, paving the way for personalized diagnostics and therapies that were previously impossible.
Key Takeaways
- •Semiconductor nanotech enables high‑density, minimally invasive neural probes.
- •Neuropixels 1.0 recorded up to 5,000 neurons simultaneously.
- •International Brain Lab mapped full‑brain activity across 139 mice.
- •Next‑gen probes aim for 100,000 neurons per session.
- •Data may power AI models for personalized brain‑disease therapies.
Summary
Baron Dutta explained how IMEC’s semiconductor expertise is reshaping neuroscience by creating ultra‑dense, minimally invasive neural probes. The effort began around 2013‑14 when neuroscientists needed system‑level recordings beyond the reach of two‑photon microscopes, prompting a partnership that leveraged sub‑nanometer CMOS fabrication to build the first Neuropixels probes.
The Neuropixels 1.0 platform, released in 2019 after a five‑year development cycle, expanded channel counts from 16‑32 to thousands, enabling simultaneous recording of 3,000‑5,000 neurons over a 10 mm span. Collaborative work with Nobel laureate John O’Keefe, Edward Moser, and the Allen Institute produced the International Brain Lab dataset—139 mice, 459 sessions, and over 600,000 neurons—demonstrating full‑brain functional maps for a single task.
Dutta highlighted that the probes integrate MEMS structures with 10 µm × 10 µm electrodes and on‑chip amplifiers operating at 30 kHz, delivering an order‑of‑magnitude noise reduction while staying under 100 µm in size to avoid heating brain tissue. Early applications include reconstructing spoken words from patients with aphasia, achieving sub‑1 % error rates, and generating rich behavioral datasets that can train transformer‑style AI models for disease‑specific simulations.
Looking ahead, IMEC aims to push density to 100,000 neurons per animal, opening pathways for real‑time brain‑computer interfaces, personalized neuro‑therapies, and new AI‑driven models of brain function that could accelerate drug discovery and clinical interventions.
Comments
Want to join the conversation?
Loading comments...