Computation Through Neural Population Dynamics

Annual Review of Neuroscience, Год журнала: 2020, Номер 43(1), С. 249 - 275

Опубликована: Июль 8, 2020

Significant experimental, computational, and theoretical work has identified rich structure within the coordinated activity of interconnected neural populations. An emerging challenge now is to uncover nature associated computations, how they are implemented, what role play in driving behavior. We term this computation through population dynamics. If successful, framework will reveal general motifs quantitatively describe dynamics implement computations necessary for goal-directed Here, we start with a mathematical primer on dynamical systems theory analytical tools apply perspective experimental data. Next, highlight some recent discoveries resulting from successful application systems. focus studies spanning motor control, timing, decision-making, working memory. Finally, briefly discuss promising lines investigation future directions framework.

Язык: Английский

---

Large-scale neural recordings with single neuron resolution using Neuropixels probes in human cortex

Nature Neuroscience, Год журнала: 2022, Номер 25(2), С. 252 - 263

Опубликована: Янв. 31, 2022

Язык: Английский

---

Home Use of a Percutaneous Wireless Intracortical Brain-Computer Interface by Individuals With Tetraplegia

IEEE Transactions on Biomedical Engineering, Год журнала: 2021, Номер 68(7), С. 2313 - 2325

Опубликована: Март 30, 2021

Individuals with neurological disease or injury such as amyotrophic lateral sclerosis, spinal cord stroke may become tetraplegic, unable to speak even locked-in. For people these conditions, current assistive technologies are often ineffective. Brain-computer interfaces being developed enhance independence and restore communication in the absence of physical movement. Over past decade, individuals tetraplegia have achieved rapid on-screen typing point-and-click control tablet apps using intracortical brain-computer (iBCIs) that decode intended arm hand movements from neural signals recorded by implanted microelectrode arrays. However, cables used convey brain tether participants amplifiers decoding computers require expert oversight, severely limiting when where iBCIs could be available for use. Here, we demonstrate first human use a wireless broadband iBCI. Based on prototype system previously pre-clinical research, replaced external 192-electrode iBCI transmitters high-resolution recording field potentials spiking activity paralysis. Two an ongoing pilot clinical trial completed item selection tasks assess iBCI-enabled cursor control. Communication bitrates were equivalent between cabled configurations. Participants also standard commercial computer browse web several mobile applications. Within-day comparison evaluated bit error rate, packet loss, recovery spike rates waveforms signals. In representative case, two arrays one participant continuously through 24-hour period at home. Wireless multi-electrode over extended periods introduces valuable tool neuroscience research is important step toward practical deployment technology independent On-demand access high-performance home promises mobility severe motor impairment.

Язык: Английский

---

An Accurate and Rapidly Calibrating Speech Neuroprosthesis

New England Journal of Medicine, Год журнала: 2024, Номер 391(7), С. 609 - 618

Опубликована: Авг. 14, 2024

BackgroundBrain–computer interfaces can enable communication for people with paralysis by transforming cortical activity associated attempted speech into text on a computer screen. Communication brain–computer has been restricted extensive training requirements and limited accuracy.MethodsA 45-year-old man amyotrophic lateral sclerosis (ALS) tetraparesis severe dysarthria underwent surgical implantation of four microelectrode arrays his left ventral precentral gyrus 5 years after the onset illness; these recorded neural from 256 intracortical electrodes. We report results decoding as he to speak in both prompted unstructured conversational contexts. Decoded words were displayed screen then vocalized use text-to-speech software designed sound like pre-ALS voice.ResultsOn first day (25 days surgery), neuroprosthesis achieved 99.6% accuracy 50-word vocabulary. Calibration required 30 minutes recordings while participant speak, followed subsequent processing. On second day, 1.4 additional hours system training, 90.2% using 125,000-word With further data, sustained 97.5% over period 8.4 months implantation, used it communicate self-paced conversations at rate approximately 32 per minute more than 248 cumulative hours.ConclusionsIn person ALS dysarthria, an reached level performance suitable restore brief training. (Funded Office Assistant Secretary Defense Health Affairs others; BrainGate2 ClinicalTrials.gov number, NCT00912041.)

Язык: Английский

---

Single-neuronal elements of speech production in humans

Nature, Год журнала: 2024, Номер 626(7999), С. 603 - 610

Опубликована: Янв. 31, 2024

Abstract Humans are capable of generating extraordinarily diverse articulatory movement combinations to produce meaningful speech. This ability orchestrate specific phonetic sequences, and their syllabification inflection over subsecond timescales allows us thousands word sounds is a core component language 1,2 . The fundamental cellular units constructs by which we plan words during speech, however, remain largely unknown. Here, using acute ultrahigh-density Neuropixels recordings sampling across the cortical column in humans, discover neurons language-dominant prefrontal cortex that encoded detailed information about arrangement composition planned production natural These represented order structure events before utterance reflected segmentation sequences into distinct syllables. They also accurately predicted phonetic, syllabic morphological components upcoming showed temporally ordered dynamic. Collectively, show how these mixtures cells broadly organized along activity patterns transition from articulation planning production. We demonstrate reliably track consonant vowel perception they distinguish processes specifically related speaking those listening. Together, findings reveal remarkably structured organization encoding cascade representations humans process can support

Язык: Английский

---

Brain control of bimanual movement enabled by recurrent neural networks

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Янв. 18, 2024

Brain-computer interfaces have so far focused largely on enabling the control of a single effector, for example computer cursor or robotic arm. Restoring multi-effector motion could unlock greater functionality people with paralysis (e.g., bimanual movement). However, it may prove challenging to decode simultaneous multiple effectors, as we recently found that compositional neural code links movements across all limbs and tuning changes nonlinearly during dual-effector motion. Here, demonstrate feasibility high-quality two cursors via network (NN) decoders. Through simulations, show NNs leverage 'laterality' dimension distinguish between left right-hand both hands become increasingly correlated. In training recurrent networks (RNNs) two-cursor control, developed method alters temporal structure data by dilating/compressing in time re-ordering it, which helps RNNs successfully generalize online setting. With this method, person can simultaneously. Our results suggest decoders be advantageous decoding, provided they are designed transfer

Язык: Английский

---

The speech neuroprosthesis

Nature reviews. Neuroscience, Год журнала: 2024, Номер 25(7), С. 473 - 492

Опубликована: Май 14, 2024

Язык: Английский

---

Generating Natural, Intelligible Speech From Brain Activity in Motor, Premotor, and Inferior Frontal Cortices

Frontiers in Neuroscience, Год журнала: 2019, Номер 13

Опубликована: Ноя. 22, 2019

Neural interfaces that directly produce intelligible speech from brain activity would allow people with severe impairment neurological disorders to communicate more naturally. Here, we record neural population in motor, premotor and inferior frontal cortices during production using electrocorticography (ECoG) show ECoG signals alone can be used generate output preserve conversational cues. To data, adapted a method the field of synthesis called unit selection, which units are concatenated form audible output. In our approach, call \emph{Brain-To-Speech}, chose subsequent based on measured audio waveforms recordings. \emph{Brain-To-Speech} employed user's own voice sounded very natural included features such as prosody accentuation. By investigating areas involved separately, found motor cortex provided information for reconstruction process than other cortical areas.

Язык: Английский

---

Decoding spoken English from intracortical electrode arrays in dorsal precentral gyrus

Journal of Neural Engineering, Год журнала: 2020, Номер 17(6), С. 066007 - 066007

Опубликована: Ноя. 25, 2020

Objective. To evaluate the potential of intracortical electrode array signals for brain-computer interfaces (BCIs) to restore lost speech, we measured performance decoders trained discriminate a comprehensive basis set 39 English phonemes and synthesize speech sounds via neural pattern matching method. We decoded correlates spoken-out-loud words in 'hand knob' area precentral gyrus, step toward eventual goal decoding attempted from ventral areas patients who are unable speak. Approach. Neural audio data were recorded while two BrainGate2 pilot clinical trial participants, each with chronically-implanted 96-electrode arrays, spoke 420 different that broadly sampled phonemes. Phoneme onsets identified recordings, their identities then classified features consisting electrode's binned action counts or high-frequency local field power. Speech synthesis was performed using 'Brain-to-Speech' also examined confounds specific overt speech: acoustic contamination systematic differences labeling phonemes' onset times. Main results. A linear decoder achieved up 29.3% classification accuracy (chance = 6%) across phonemes, an RNN classifier 33.9% accuracy. Parameter sweeps indicated did not saturate when adding more electrodes training data, improved utilizing time-varying structure data. Microphonic phoneme modestly increased accuracy, but could be mitigated by artifact subtraction marker, respectively. r 0.523 correlation between true reconstructed audio. Significance. The ability decode nontraditional suggests placing arrays is promising direction BCIs.

Язык: Английский

---

The FACTS model of speech motor control: Fusing state estimation and task-based control

PLoS Computational Biology, Год журнала: 2019, Номер 15(9), С. e1007321 - e1007321

Опубликована: Сен. 3, 2019

We present a new computational model of speech motor control: the Feedback-Aware Control Tasks in Speech or FACTS model. employs hierarchical state feedback control architecture to simulated vocal tract and produce intelligible speech. The includes higher-level tasks lower-level articulators. task controller is modeled as dynamical system governing creation desired constrictions tract, after Task Dynamics. Both articulatory controllers rely on an internal estimate current generate commands. This derived, based efference copy applied controls, from forward that predicts both next well expected auditory somatosensory feedback. A comparison between predicted actual then used update prediction. able qualitatively replicate many characteristics human system: robust noise sensory pathways, relatively unaffected by loss but more significantly impacted feedback, responds appropriately externally-imposed alterations also replicates previously hypothesized trade-offs reliance shows for first time how this relationship may be mediated acuity each domain. These results have important implications our understanding humans.

Язык: Английский

---