Researchers affiliated with ETH Zurich, UC Santa Barbara, and Miguel Hernández University studied one participant temporarily implanted with a 96-electrode visual cortex array. They trained a deep neural network to predict electrically evoked neural activity using pre-stimulation brain activity and stimulation parameters. In this proof of concept, model-selected stimulation more accurately evoked targeted neural activity while using less electrical current. Whether the approach improves safety or generalizes to other patients remains open.
Individual Post Research Log • ID: RL-2026-1115-D3-LI3 • Logged: 2026-11-16
Scope & Ethics: Reports exact 96-electrode proof-of-concept metrics in one participant.
Ref: Liu, Soo, López-Peco & Soto-Sánchez (2026, Neuron); Mendes (2026, EBRAINS).

