UCSF

Loss of speech after paralysis is devastating, but circumventing motor pathway injury by directly decoding speech from intact cortical activity has the potential to restore natural communication and self-expression. Recent discoveries have defined how key features of speech production are facilitated by the coordinated activity of vocal-tract articulatory and motor-planning cortical representations. Building on these fundamental discoveries, we describe a series of studies performed in people with paralysis and anarthria (inability to articulate speech) to restore naturalistic communication through an electrocorticography-based neuroprosthesis. In the first study, we develop a neuroprosthesis capable of decoding cortical activity during attempted speech into sentences, audible speech, and animated movements in a virtual avatar, using a vocabulary of over 1024 words. Further, we interpret the underlying cortical signals that drive decoding performance, finding a residual articulatory and somatotopic organization of speech motor cortex, decades after paralysis. In the second study, we develop a bilingual speech neuroprosthesis, capable of decoding cortical activity into multiple languages, solely based on the user’s speech attempts. Finally, in this last study, we develop a speech neuroprosthesis that maintains specificity to attempted speech and accurate performance in real-world settings, such as when the participants listen to podcasts or read articles. Thus, overall, this work highlights the utility of speech neuroprostheses in restoring naturalistic communication for individuals with paralysis, taking key steps toward their ultimate clinical viability.