Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Multiplexed representations for attempted speech, listening, and reading on the motor cortex and their effect on speech neuroprostheses
Poster Session D, Saturday, October 26, 10:30 am - 12:00 pm, Great Hall 3 and 4
Alexander Silva1,2, Jessie Liu1,2, Vanessa Anderson2, Cady Kurtz-Miott2, Irina Hallinan2, Kaylo Littlejohn2,3, Samantha Brosler2, David Moses2, Edward Chang2; 1equal contribution, 2University of California at San Francisco, 3University of California at Berkeley
The motor cortex is most well known for encoding low-level motor representations, such as those of the tongue or the lips, which are the basis for complex motor behaviors, such as speech. The motor cortex has also been shown to contain representations of both multi-effector movements, where neural populations are not tuned to any one particular muscle group, as well as representations relating to perceiving visual or auditory stimuli during reading and listening, respectively. Some studies have shown a direct overlap between these functions in the motor cortex, most commonly comparing reading to listening or listening to speech, but fewer have investigated the overlap between the three. Additionally, the nature of these multiplexed representations and whether they reflect identical or shared processing remains unclear. These questions are also particularly relevant for the development of speech neuroprostheses. Speech neuroprostheses have the potential to restore naturalistic communication to people with paralysis by decoding intended speech from persistent representations of low-level vocal-tract movements in the motor cortex. The existence of multiplexed representations in the motor cortex for attempted speech, reading, and listening then raises the concern of whether reading and listening, essential daily functions, may unintentionally engage or interfere with speech decoding. Indeed, one of the most commonly cited desires of potential users of speech neuroprostheses is that they remain specific to decoding only attempted speech and not other functions. To address these questions, we first developed a speech-decoding system in two participants with vocal-tract paralysis that maintained online performance and specificity to volitional speech attempts, regardless of whether they were also participating in listening, reading, and non-speech mental imagery tasks. Offline, we observed multiplexed neural populations that responded to attempted speech, listening, and reading, but we found that they leveraged different neural representations, partially facilitated by their unique spectrotemporal response patterns across tasks. Strikingly, these neural populations localized to the middle precentral gyrus (midPrCG). Though multiplexed representations were found in the midPrCG, it was also an important area for discriminating between attempted speech, reading, and listening and for decoding attempted speech just prior to execution. Together, these results suggest that the midPrCG may have a distinct role in speech-motor planning, which is supported by multiplexed neural activity. These results further our understanding of multiplexed speech activity in the motor cortex and demonstrate a maximally-reliable decoding framework for speech neuroprostheses.
Topic Areas: Speech Motor Control, Speech Perception