Researchers at the University of California, Davis published a paper in Nature Medicine describing a new machine‑learning platform called BRAND that converts neural activity recorded during attempted speech into linguistic units. The system’s decoding pipeline maps brain signals to phonemes and then to words.
What is novel?
BRAND runs on commercially available Blackrock electrode arrays rather than requiring bespoke implant hardware. The platform is presented as a hardware‑agnostic decoding layer that can be slotted beneath different implant types, rather than a single device tied to one patient.
Where and who is using it?
According to the paper, BRAND has been deployed across sites in the BrainGate consortium spanning multiple universities. The researchers note real‑world use: trial participant Casey Harrell, who has ALS, has used the system to work full‑time as an environmental advocate for over a year.
Why it matters
In brain–computer interface research the technical challenge has moved from capturing neural signals to interpreting them reliably as language. BRAND aims to provide a general translation layer: once decoding is solved in a broadly applicable way, signals from varied electrode arrays can be interpreted through the same stack, rather than requiring bespoke decoders for each implant.
Practical implications
Because BRAND is hardware‑agnostic, it could allow different implant manufacturers to connect to the same decoding stack, potentially accelerating clinical deployment and wider adoption. The approach reframes implants primarily as signal collectors, while the universal decoding layer performs the language mapping.
Summary
BRAND represents an attempt to create a universal decoder that transforms attempted‑speech neural activity into phonemes and words using off‑the‑shelf Blackrock arrays. It is already used in the BrainGate consortium and has enabled an ALS trial participant, Casey Harrell, to maintain full‑time employment for over a year.
