BrainGate2 researchers decode inner speech from motor cortex in real time; Cell journal publication

BrainGate consortium/BrainGate2 clinical trial (speech neuroprosthesis)

A multi-site BrainGate2 team led by Stanford and Emory researchers published findings in Cell demonstrating real-time decoding of silent inner speech from motor cortex neural activity. Four trial participants with speech impairments from ALS or stroke imagined speaking English words; the system decoded their inner monologue with up to 74% accuracy and identified a distinct 'motor-intent' neural dimension separating imagined from attempted speech.

• Inner speech decoding achieved up to 74% accuracy across four BrainGate2 trial participants.
• First evidence that imagined (silent) speech is robustly represented in motor cortex.
• Motor-intent signal could enable BCIs to distinguish deliberate communication attempts from unrelated thoughts.
• Findings expand BCI applications beyond decoded arm/hand movement to natural, thought-based speech communication.