High-performance brain-to-text communication via handwriting
Open accessPublished version of a preprint
Abstract
Brain-computer interfaces (BCIs) can restore communication to people who have lost the ability to move or speak. So far, a major focus of BCI research has been on restoring gross motor skills, such as reaching and grasping 1-5 or point-and-click typing with a computer cursor 6,7. However, rapid sequences of highly dexterous behaviours, such as handwriting or touch typing, might enable faster rates of communication. Here we developed an intracortical BCI that decodes attempted handwriting movements from neural activity in the motor cortex and translates it to text in real time, using a recurrent neural network decoding approach. With this BCI, our study participant, whose hand was paralysed from spinal cord injury, achieved typing speeds of 90 characters per minute with 94.1% raw accuracy online, and greater than 99% accuracy offline with a general-purpose autocorrect. To our knowledge, these typing speeds exceed those reported for any other BCI, and are comparable to typical smartphone typing speeds of individuals in the age group of our participant (115 characters per minute)8. Finally, theoretical considerations explain why temporally complex movements, such as handwriting, may be fundamentally easier to decode than point-to-point movements. Our results open a new approach for BCIs and demonstrate the feasibility of accurately decoding rapid, dexterous movements years after paralysis.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
32 of 39 cited sources checked · 18 supported · 1 flagged · accuracy not yet measured
- Premisecitation
In parameter optimization, cursor gain is raised to increase typing speed until decoding errors make the cursor uncontrollable.
“During parameter optimization, the cursor gain is increased so as to increase typing rate, until the cursor moves so quickly that it becomes uncontrollable due to decoding errors22.”
Quote in the paperSubstantial claim1 of 1 sources checked- cites [22] Willett, F. R. et al. Signal-independent noise in intracortical brain–computer interfaces causes movement time properties inconsistent with Fitts’ law. J. Neural Eng. 14, 026010 (2017). source
The Results section, “Effect of gain on movement time” (page unavailable), reports that increasing gain initially shortened target-acquisition time, but higher gains made cursor movements fast, inaccurate, and uncontrollable around targets. The paper relates this difficulty to decoding noise, but it studies cursor target acquisition, not typing rate, and does not describe the claimed typing-speed optimization procedure. (LLM, conf=0.85)
Experimental — accuracy not yet measured
- Premisecitation
Decoding accuracy is the main factor limiting the speed of point-and-click BCIs.
“The speed of point-and-click BCIs is limited primarily by decoding accuracy.”
Quote in the paperSubstantial claim1 of 1 sources checked- cites [22] Willett, F. R. et al. Signal-independent noise in intracortical brain–computer interfaces causes movement time properties inconsistent with Fitts’ law. J. Neural Eng. 14, 026010 (2017). source
The Results (Effect of gain on movement time; page unavailable) report that increasing cursor speed made movements fast but inaccurate, and the Introduction identifies signal-independent decoding noise as the root cause of the movement-time properties observed. This supports a speed–accuracy constraint in the intracortical cursor systems studied, but not the broader claim that decoding accuracy is the primary speed limit for point-and-click BCIs generally. (LLM, conf=0.84)
Experimental — accuracy not yet measured
- Premisecitation
EEG spellers based on oddball potentials or motor imagery usually achieve 1-5 characters per minute.
“EEG spellers based on oddball potentials or motor imagery typically achieve 1–5 characters per minute28–32.”
Quote in the paperSubstantial claim5 of 5 sources checked- cites [28] Vansteensel, M. J. et al. Fully implanted brain–computer interface in a locked-in patient with ALS. N. Engl. J. Med. 375, 2060–2066 (2016). source
The Results section, “Communication and Independent Home Use” (page unavailable), reports spelling times for one patient using a fully implanted interface. Our copy does not give a typical rate for EEG spellers based on oddball potentials or motor imagery; searches for those methods and their characters-per-minute rates found no such passage. (LLM, conf=0.76)
Experimental — accuracy not yet measured - cites [29] Nijboer, F. et al. A P300-based brain–computer interface for people with amyotrophic lateral sclerosis. Clin. Neurophysiol. 119, 1909–1916 (2008). source
The Results report 1.2 and 2.1 selections per minute for the study’s P300 speller, which uses an oddball paradigm (Introduction). This supports the stated range for this speller in people with ALS, but the paper does not establish that EEG spellers generally achieve that rate or provide a rate for motor-imagery spellers. Page numbers were not available in our copy. (LLM, conf=0.86)
Experimental — accuracy not yet measured - cites [30] Townsend, G. et al. A novel P300-based brain–computer interface stimulus presentation paradigm: moving beyond rows and columns. Clin. Neurophysiol. 121, 1109–1120 (2010). source
The Discussion (§4.2; page unavailable) reports 4.10 selections/min for a previously studied P300 speller and 4.36 selections/min for this paper’s checkerboard P300 paradigm, both within the claimed range. It does not establish that 1–5 characters/min is typical across EEG spellers or report a rate for motor-imagery spellers. (LLM, conf=0.84)
Experimental — accuracy not yet measured - cites [31] McCane, L. M. et al. P300-based brain–computer interface (BCI) event-related potentials (ERPs): people with amyotrophic lateral sclerosis (ALS) vs. age-matched controls. Clin. Neurophysiol. 126, 2124–2131 (2015). source
The Introduction (page unavailable) identifies the P300 speller as an oddball-response system that provides slow communication. The Methods describe a characters-per-minute measure, and the Results report communication rates for this study, but the passages do not state that EEG spellers typically achieve 1–5 characters per minute or address motor-imagery spellers. (LLM, conf=0.77)
Experimental — accuracy not yet measured - cites [32] Wolpaw, J. R. et al. Independent home use of a brain–computer interface by people with amyotrophic lateral sclerosis. Neurology 91, e258–e267 (2018). source
The Results report an average of 2.9 (±1.3 SD) selections per minute for BCI applications used by people with ALS, within the claimed numerical range. The paper does not establish a typical rate of 1–5 characters per minute across EEG spellers based on oddball potentials or motor imagery; selections are also not necessarily characters. (LLM, conf=0.86)
Experimental — accuracy not yet measured
- Premisecitation
Among intracortical BCIs, 2D cursor point-and-click typing has been the best performer, reaching at most 40 characters per minute.
“For intracortical BCIs, the highest performing method has been point-and-click typing with a 2D computer cursor, peaking at 40 characters per minute 7”
Quote in the paperSubstantial claim1 of 1 sources checked- cites [7] Pandarinath, C. et al. High performance communication by people with paralysis using an intracortical brain–computer interface. eLife 6, e18554 (2017). source
The Introduction describes two-dimensional cursor control with click selection, and the Results say this study outperformed previous BCIs tested with people with motor impairment. The Results report T5’s average typing rate as 39.2 correct characters per minute, approximately 40, but do not identify 40 as a peak or an upper limit. (LLM, conf=0.85)
Experimental — accuracy not yet measured
- Premisecitation
Before this work, the best free typing rate for an intracortical BCI was 24.4 correct characters per minute.
“The prior state-of-the-art for free typing in intracortical BCIs was 24.4 correct characters per minute 7.”
Quote in the paperSubstantial claim1 of 1 sources checked- cites [7] Pandarinath, C. et al. High performance communication by people with paralysis using an intracortical brain–computer interface. eLife 6, e18554 (2017). source
- Findingown data
After applying a time-alignment method to reduce temporal variability, the neural activity showed consistent patterns that were specific to each character.
“We used a time-alignment technique to remove temporal variability11, revealing remarkably consistent underlying patterns of neural activity that are unique to each character (Fig. 1c).”
Quote in the paperSubstantial claim1 of 1 sources checked- cites [11] Williams, A. H. et al. Discovering precise temporal patterns in large-scale neural recordings through robust and interpretable time warping. Neuron 105, 246–259 (2020). source
- Premisecitation
Microelectrode array technology has retained function for more than 1000 days after implantation, including in this study.
“Current microelectrode array technology has been shown to retain functionality for 1000+ days post implant37,38 (including here; see Extended Data Fig. 7)”
Quote in the paperNot a substantial claim1 of 2 sources checked- cites [37] Simeral, J. D., Kim, S.-P., Black, M. J., Donoghue, J. P. & Hochberg, L. R. Neural control of cursor trajectory and click by a human with tetraplegia 1000 days after implant of an intracortical microelectrode array. J. Neural Eng. 8, 025027 (2011). source
- cites [38] Bullard, A. J., Hutchison, B. C., Lee, J., Chestek, C. A. & Patil, P. G. Estimating risk for future intracranial, fully implanted, modular neuroprosthetic systems: a systematic review of hardware complications in clinical deep brain stimulation and experimental human intracortical arrays. Neuromodulation 23, 411–426 (2020). source
- Premisecitation
Compared with other recording technologies such as EEG and ECoG, microelectrode arrays have enabled the best BCI performance to date for communication, arm control, and general computer use.
“and has enabled the highest BCI performance to date compared to other recording technologies (EEG, ECoG) for restoring communication7, arm control2,5, and general-purpose computer use39.”
Quote in the paperNot a substantial claim4 of 4 sources checked- cites [7] Pandarinath, C. et al. High performance communication by people with paralysis using an intracortical brain–computer interface. eLife 6, e18554 (2017). source
- cites [2] Collinger, J. L. et al. High-performance neuroprosthetic control by an individual with tetraplegia. Lancet 381, 557–564 (2013). source
- cites [5] Ajiboye, A. B. et al. Restoration of reaching and grasping movements through brain-controlled muscle stimulation in a person with tetraplegia: a proof-of-concept demonstration. Lancet 389, 1821–1830 (2017). source
The Methods describe implanted microelectrode arrays controlling reaching and grasping, and the Discussion says intracortical BCIs offer greater functional performance than non-invasive systems, specifically contrasting robotic-arm control with EEG. The paper does not establish a performance comparison with ECoG. Communication and general-purpose computer use have separate citations in the citing sentence and are not judged against this source. (LLM, conf=0.84)
Experimental — accuracy not yet measured - cites [39] Nuyujukian, P. et al. Cortical control of a tablet computer by people with paralysis. PLoS One 13, e0204566 (2018). source
The Introduction (page unavailable) says prior EEG and ECoG work had not shown general control of a computing device such as a commercial tablet. The Abstract and Conclusion (pages unavailable) report high-performance control of an unmodified general-purpose tablet using implanted arrays. The paper does not establish that this achieved the highest BCI performance relative to EEG or ECoG; the separate communication and arm-control claims carry other citations in the citing sentence. (LLM, conf=0.84)
Experimental — accuracy not yet measured
- Findingown data
When sessions were separated by only 2-7 days, performance remained reasonable without decoder retraining: 11.1% raw error rate and 1.5% with a language model.
“We found that when only 2–7 days passed between sessions, performance was reasonable with no decoder retraining (11.1% raw error rate, 1.5% with a language model)”
Quote in the paperSubstantial claim - Findingown data
Typing speed reached a plateau of 90 characters per minute with a 5.4% character error rate.
“Importantly, typing rates were high, plateauing at 90 characters per minute with a 5.4% character error rate (Fig. 2c, average of red circles).”
Quote in the paperSubstantial claim - Findingown data
Simulations of classification accuracy across neural noise levels showed that characters are easier to classify than straight lines.
“To confirm this, we simulated the classification accuracy for each set of movements as a function of neural noise (Fig. 4d), demonstrating that characters are easier to classify than straight lines.”
Quote in the paperSubstantial claim - Findingreasoning
The authors designed an alphabet that is theoretically easier to classify than the Latin alphabet.
“We explored doing so and designed an alphabet that is theoretically easier to classify than the Latin alphabet (Extended Data Fig. 6).”
Quote in the paperNot a substantial claim - Findingown data
An offline k-nearest neighbor classifier distinguished the characters with 94.1% accuracy (95% CI 92.6-95.8).
“Using a k-nearest neighbor classifier applied offline to the neural activity, we could classify the characters with 94.1% accuracy (95% CI = [92.6, 95.8]).”
Quote in the paperSubstantial claim - Findingown data
Neural activity patterns for characters had nearest-neighbor distances 72% larger than those for straight lines (95% CI 60-86%).
“We found that the nearest-neighbor distances for each movement were 72% larger for characters as compared to straight lines (95% CI = [60%, 86%])”
Quote in the paperSubstantial claim - Findingown data
The neural dimensions that encoded pen tip velocity explained 30% of the total variance in neural activity.
“The neural dimensions that represented pen tip velocity accounted for 30% of the total neural variance.”
Quote in the paperNot a substantial claim - Findingown data
Unsupervised retraining yielded a 7.3% raw error rate (0.84% with a language model) when sessions were no more than 7 days apart.
“Encouragingly, unsupervised retraining achieved a 7.3% raw error rate (0.84% with a language model) when sessions were separated by 7 days or less.”
Quote in the paperSubstantial claim - Findingreasoning
The optimized alphabet avoids large groups of redundant, similarly written letters.
“The optimized alphabet avoids large clusters of redundant letters that are written similarly (most Latin letters begin with either a downstroke or a counter-clockwise curl).”
Quote in the paperSubstantial claim - Findingown data
The toy model's results also held when noise was correlated across time and neurons, not only for independent white noise.
“Although neural noise in the toy model was assumed to be independent white noise, we found that these results also hold for noise that is correlated across time and neurons (Extended Data Fig. 5 and Supplemental Note 1).”
Quote in the paperNot a substantial claim - Findingown data
Offline analyses indicated that more limited or even unsupervised decoder retraining may be promising.
“Here, offline analyses showed the potential promise of more limited, or even unsupervised, decoder retraining (Fig. 3).”
Quote in the paperSubstantial claim - Findingnone given
New decoding methods were developed to handle unlabeled neural sequences when data are limited.
“To achieve high performance, we developed new decoding methods to work effectively with unlabeled neural sequences in data-limited regimes.”
Quote in the paperSubstantial claim - Interpretationown data
The results indicate that even years after paralysis, motor cortex likely retains a handwriting representation strong enough for a useful BCI.
“Taken together, these results suggest that, even years after paralysis, the neural representation of handwriting in motor cortex is likely strong enough to be useful for a BCI.”
Quote in the paperSubstantial claim - Interpretationown data
The greater temporal variety in letter writing appears to drive the greater separability of the movements.
“suggesting that the increased variety of temporal patterns in letter writing drives the increased separability of each movement (Fig. 4e).”
Quote in the paperSubstantial claim - Interpretationnone given
To the authors' knowledge, 90 characters per minute is the fastest typing rate reported for any BCI.
“To our knowledge, 90 characters per minute is the highest typing rate yet reported for any type of BCI (see Discussion).”
Quote in the paperSubstantial claim - Study design and conductreasoning
The RNN was evaluated on sentences excluded from training, so it could not have overfit to those evaluation sentences.
“We evaluated the RNN’s performance over a series of 5 days, each day containing 4 evaluation blocks of 7–10 sentences that the RNN was never trained on (thus ensuring that the RNN could not have overfit to those sentences).”
Quote in the paperSubstantial claim - Interpretationown data
Time-varying movement patterns such as handwritten letters are fundamentally easier to decode than point-to-point movements.
“These results suggest that time-varying patterns of movement, such as handwritten letters, are fundamentally easier to decode than point-to-point movements.”
Quote in the paperSubstantial claim - Interpretationown data
These findings suggest that unsupervised decoder retraining plus limited supervised retraining after longer breaks may be enough to maintain high performance.
“The above results are promising for clinical viability, as they suggest that unsupervised decoder retraining, combined with more limited supervised retraining after longer periods of inactivity, may be sufficient to achieve high performance.”
Quote in the paperSubstantial claim - Interpretationown data
The introduced approach sets a new communication-rate benchmark of 90 characters per minute.
“Here, we introduced a novel approach for communication BCIs – decoding a rapid, dexterous motor behavior in a person with tetraplegia – that sets a new benchmark for communication rate at 90 characters per minute.”
Quote in the paperSubstantial claim
- May combine two claims. 20
- Classified as probably not a substantial claim. 8
May combine two claims.
Interpretationown dataThose autocorrected error rates are similar to state-of-the-art speech recognition systems, which have word error rates of 4-5%, and fall within a usable range.
“which is comparable to state-of-the-art speech recognition systems (e.g. word error rates of 4–5% 14,15), putting it well within the range of usability.”
Quote in the paperSubstantial claim1 of 2 sources checked- cites [14] Xiong, W. et al. The Microsoft 2017 Conversational Speech Recognition System. Preprint at https://arxiv.org/abs/1708.06073 (2017). source
The Conclusions (p. 4) identify 6.2% as the state-of-the-art word error rate on the Switchboard test data, not 4–5%. The paper does not assess whether an error rate is usable; the citing sentence presents usability as its own inference. (LLM, conf=0.88)
Checked by a reasoning model — may be wrong.
Experimental — accuracy not yet measured - cites [15] He, Y. et al. Streaming end-to-end speech recognition for mobile devices. In 2019 IEEE International Conference on Acoustics, Speech and Signal Processing 6381–6385 (2019). source
Classified as probably not a substantial claim.
PremisecitationEarlier intracortical studies found that after paralysis, motor cortex still encodes intended reaching, grasping, and cursor movements.
“Prior BCI studies have shown that the motor intention for gross motor skills, such as reaching, grasping or moving a computer cursor, remains neurally encoded in motor cortex after paralysis1–7.”
Quote in the paperSubstantial claim6 of 7 sources checked- cites [1] Hochberg, L. R. et al. Reach and grasp by people with tetraplegia using a neurally controlled robotic arm. Nature 485, 372–375 (2012). source
- cites [2] Collinger, J. L. et al. High-performance neuroprosthetic control by an individual with tetraplegia. Lancet 381, 557–564 (2013). source
- cites [3] Aflalo, T. et al. Neurophysiology. Decoding motor imagery from the posterior parietal cortex of a tetraplegic human. Science 348, 906–910 (2015). source
The passage comparing this study with previous human M1 recordings (indexed section labeled Abstract; page unavailable) reports that those studies decoded movement trajectories, supporting the motor-cortex portion in part. This paper’s own recordings of imagined reaching, grasping, and cursor control were from the posterior parietal cortex of a tetraplegic participant, not motor cortex. It does not establish that the earlier motor-cortex studies demonstrated all three movements named in the claim. (LLM, conf=0.88)
Experimental — accuracy not yet measured - cites [4] Bouton, C. E. et al. Restoring cortical control of functional movement in a human with quadriplegia. Nature 533, 247–250 (2016). source
- cites [5] Ajiboye, A. B. et al. Restoration of reaching and grasping movements through brain-controlled muscle stimulation in a person with tetraplegia: a proof-of-concept demonstration. Lancet 389, 1821–1830 (2017). source
- cites [6] Jarosiewicz, B. et al. Virtual typing by people with tetraplegia using a self-calibrating intracortical brain–computer interface. Sci. Transl. Med. 7, 313ra179 (2015). source
- cites [7] Pandarinath, C. et al. High performance communication by people with paralysis using an intracortical brain–computer interface. eLife 6, e18554 (2017). source
The Discussion (page unavailable) reports that people with no limb movement controlled a cursor using an intracortical BCI, and that monkeys given a paralytic agent retained grasp-related cortical activity. Our copy does not establish the claimed post-paralysis encoding of intended reaching; its mention of reaching is in a reference title about primates. (LLM, conf=0.82)
Experimental — accuracy not yet measured
Classified as probably not a substantial claim.
PremisecitationEarlier work has shown that neural recordings are stable over short periods.
“as might be expected from prior work showing short-term stability of neural recordings19–21.”
Quote in the paperSubstantial claim3 of 3 sources checked- cites [19] Dickey, A. S., Suminski, A., Amit, Y. & Hatsopoulos, N. G. Single-unit stability using chronically implanted multielectrode arrays. J. Neurophysiol. 102, 1331–1339 (2009). source
- cites [20] Eleryan, A. et al. Tracking single units in chronic, large scale, neural recordings for brain machine interface applications. Front. Neuroeng. 7, 23 (2014). source
- cites [21] Downey, J. E., Schwed, N., Chase, S. M., Schwartz, A. B. & Collinger, J. L. Intracortical recording stability in human brain–computer interface users. J. Neural Eng. 15, 046016 (2018). source
Probably not a substantial claim: may not pass S3 (load-bearing).
InterpretationcitationMaximizing nearest-neighbor distance between movements could be used to design movement sets that are easier to classify.
“More generally, using the principle of maximizing the nearest neighbor distance between movements, it should be possible to optimize a set of movements for ease of classification26.”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites [26] Cunningham, J. P., Yu, B. M., Gilja, V., Ryu, S. I. & Shenoy, K. V. Toward optimal target placement for neural prosthetic devices. J. Neurophysiol. 100, 3445–3457 (2008). source
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationPerformance may vary across participants; in an earlier study, T5 had the best performance among three participants.
“Variability in performance across participants is also a potential concern (in a prior study, T5 achieved the highest performance of 3 participants7).”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites [7] Pandarinath, C. et al. High performance communication by people with paralysis using an intracortical brain–computer interface. eLife 6, e18554 (2017). source
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationSpeech-decoding BCIs show promise for rapid communication, but their accuracy and vocabulary size still need improvement for general-purpose use.
“Recently, speech-decoding BCIs have shown exciting promise for restoring rapid communication (e.g. 34,16,17), but their accuracies and vocabulary sizes require further improvement to support general-purpose use.”
Quote in the paperNot a substantial claim3 of 3 sources checked- cites [34] Mugler, E. M. et al. Direct classification of all American English phonemes using signals from functional speech motor cortex. J. Neural Eng. 11, 035015 (2014). source
- cites [16] Anumanchipalli, G. K., Chartier, J. & Chang, E. F. Speech synthesis from neural decoding of spoken sentences. Nature 568, 493–498 (2019). source
- cites [17] Makin, J. G., Moses, D. A. & Chang, E. F. Machine translation of cortical activity to text with an encoder-decoder framework. Nat. Neurosci. 23, 575–582 (2020). source
May combine two claims.
PremisecitationEEG spellers using visually evoked potentials have reached 60 characters per minute but have usability drawbacks: they occupy the eyes, are usually not self-paced, and need flashing light panels.
“EEG spellers that use visually evoked potentials have achieved speeds of 60 characters per minute 18, but have important usability limitations, as they tie up the eyes, are not typically self-paced, and require panels of flashing lights on a screen.”
Quote in the paperSubstantial claim1 of 1 sources checked- cites [18] Chen, X. et al. High-speed spelling with a noninvasive brain–computer interface. Proc. Natl Acad. Sci. USA 112, E6058–E6067 (2015). source
May combine two claims.
PremisecitationIntracortical BCIs using 2D cursor movements allow more freedom but have not exceeded 40 correct characters per minute in people.
“Intracortical BCIs based on 2D cursor movements give the user more freedom to look around and set their own pace of communication, but have yet to exceed 40 correct characters per minute in people7.”
Quote in the paperSubstantial claim1 of 1 sources checked- cites [7] Pandarinath, C. et al. High performance communication by people with paralysis using an intracortical brain–computer interface. eLife 6, e18554 (2017). source
The Introduction describes continuous two-dimensional cursor control, and the Discussion describes the point-and-click interface as flexible. The Results report average copy-typing rates up to 39.2 correct characters per minute, but do not establish that no individual typing rate exceeded 40. The cited passages also do not establish the specific claim that users can look around while communicating. (LLM, conf=0.84)
Experimental — accuracy not yet measured
May combine two claims.
Findingown dataDuring attempted handwriting, the recorded neural activity looked strong and repeatable, but the timing of its peaks and valleys differed between trials, possibly because writing speed fluctuated.
“The neural activity appeared to be strong and repeatable, although the timing of its peaks and valleys varied across trials, potentially due to fluctuations in writing speed.”
Quote in the paperNot a substantial claimMay combine two claims.
Findingown dataThe neural patterns for individual characters were highly stable in the short term (mean correlation 0.85 for intervals of 7 days or less), and neural changes appeared to accumulate steadily and predictably.
“We assessed the stability of the neural patterns associated with each character and found high short-term stability (mean correlation = 0.85 when 7 days apart or less), and neural changes that seemed to accumulate at a steady and predictable rate (Extended Data Fig. 4).”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataA non-causal offline RNN trained on all available sentences achieved a 0.17% character error rate, suggesting a high performance ceiling, but it could not give letter-by-letter feedback.
“Accuracy was extremely high in this regime (0.17% character error rate), indicating a high potential ceiling of performance, although this decoder would not be able to provide letter-by-letter feedback to the user.”
Quote in the paperNot a substantial claimMay combine two claims.
Findingown dataLinear decoding of pen tip velocity from trial-averaged neural activity produced recognizable letter shapes, indicating that pen tip velocity is strongly represented.
“Readily recognizable letter shapes confirm that pen tip velocity is robustly encoded.”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataSpatial dimensionality was only 1.24 times larger for characters (95% CI 1.19-1.30), but temporal dimensionality was 2.65 times larger (95% CI 2.58-2.72).
“We found that spatial dimensionality was only modestly larger for characters (1.24 times larger; 95% CI = [1.19, 1.30]), but that the temporal dimensionality was much greater (2.65 times larger; 95% CI = [2.58, 2.72])”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataOffline autocorrection by a language model reduced the average character error rate to 0.89% and the average word error rate to 3.4% across all days.
“The character error rate fell to 0.89% and the word error rate fell to 3.4% averaged across all days”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataA t-SNE visualization showed tight clusters of neural activity for individual characters and a mainly motoric organization in which characters written similarly were closer together.
“The t-SNE visualization revealed tight clusters of neural activity for each character and a predominantly motoric encoding where characters that are written similarly are closer together.”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataParticipant T5 had a C4 AIS C spinal cord injury, was paralyzed below the neck, and had no functional hand movement beyond twitching and micromotion.
“T5 has a high-level spinal cord injury (C4 AIS C) and was paralyzed from the neck down; his hand movements were entirely non-functional and limited to twitching and micromotion.”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataIn a toy model, letting trajectories vary over time increased the nearest-neighbor distance between neural trajectories and improved classification accuracy when noise was high enough.
“Simply by allowing the trajectories to change in time (Fig. 4g), the nearest neighbor distance between the neural trajectories can be increased, resulting in an increase in classification accuracy when noise levels are large enough (Fig. 4h).”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataWhen T5 freely answered open-ended questions, real-time performance was 73.8 characters per minute with an 8.54% character error rate, and 2.25% with a language model.
“The results confirm that high performance can also be achieved when the user writes self-generated sentences as opposed to copying on-screen prompts (73.8 characters per minute with an 8.54% character error rate in real-time, 2.25% with a language model).”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataIn a retrospective analysis, 10 calibration sentences (8.7 minutes) sufficed for an 8.5% raw error rate (1.7% with a language model), while 30 sentences were required to match the 5.9% raw online error rate.
“We found that 10 sentences (8.7 minutes) were enough to achieve a raw error rate of 8.5% (1.7% with a language model), although 30 sentences were needed to match the raw online error rate of 5.9%.”
Quote in the paperSubstantial claimMay combine two claims.
InterpretationreasoningThis temporal variety is probably one important factor, though not necessarily the only one, that made the handwriting BCI faster than continuous-motion point-and-click BCIs.
“We think this is one, but not necessarily the only, important factor that enabled a handwriting BCI to go faster than continuous-motion point-and-click BCIs.”
Quote in the paperSubstantial claimClassified as probably not a substantial claim.
InterpretationreasoningDaily decoder retraining is intended to compensate for changes in neural recordings over time, which may arise from neural plasticity or electrode array micromotion.
“Retraining helps account for changes in neural recordings that accrue over time, which might be caused by neural plasticity or electrode array micromotion.”
Quote in the paperSubstantial claimMay combine two claims.
PremisecitationOther discrete classification-based BCIs have generally used directional movements with little temporal variety, which may have constrained their accuracy or the number of movements they can distinguish.
“Other discrete (classification-based) BCIs have also typically used directional movements with little temporal variety, which may have limited their accuracy and/or the size of the movement set24,25.”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [24] Musallam, S., Corneil, B. D., Greger, B., Scherberger, H. & Andersen, R. A. Cognitive control signals for neural prosthetics. Science 305, 258–262 (2004). source
- cites [25] Santhanam, G., Ryu, S. I., Yu, B. M., Afshar, A. & Shenoy, K. V. A high-performance brain–computer interface. Nature 442, 195–198 (2006). source
May combine two claims.
Interpretationown dataThe real-time system is general, self-paced, leaves the eyes free, and is accurate enough for real-world use: 94.1% raw accuracy and over 99% offline with a large-vocabulary language model.
“The demonstrated real-time system is general (the user can express any sentence), easy to use (entirely self-paced and the eyes are free to move), and accurate enough to be useful in the real-world (94.1% raw accuracy, and >99% accuracy offline with a large-vocabulary language model).”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).
InterpretationreasoningThese methods could be applied to other sequential behaviors that cannot be observed directly, such as decoding speech in someone who cannot speak.
“These methods could be applied more generally to any sequential behavior that cannot be observed directly (e.g., decoding speech from someone who can no longer speak).”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationIntracortical microelectrode array technology is still maturing and needs further evidence of longevity, safety, and efficacy before broad clinical adoption.
“More broadly, intracortical microelectrode array technology is still maturing, and requires further demonstrations of longevity, safety, and efficacy before widespread clinical adoption35,36.”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [35] Nurmikko, A. Challenges for large-scale cortical interfaces. Neuron 108, 259–269 (2020). source
- cites [36] Vázquez-Guardado, A., Yang, Y., Bandodkar, A. J. & Rogers, J. A. Recent advances in neurotechnologies with broad potential for neuroscience research. Nat. Neurosci. 23, 1522–1536 (2020). source
May combine two claims.
InterpretationreasoningThe system is a single-participant proof of concept that a high-performance handwriting BCI is possible, not yet a complete clinically viable system.
“Finally, it is important to recognize that the current system is a proof-of-concept that a high-performance handwriting BCI is possible (in a single participant); it is not yet a complete, clinically viable system.”
Quote in the paperSubstantial claim
The authors state this as an assumption.
It was not yet known whether the neural representation for a rapid, highly dexterous skill such as handwriting remains intact after paralysis.
“However, it is still unknown whether the neural representation for a rapid and highly-dexterous motor skill, such as handwriting, also remains intact.”
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The decoding speed is 3.4 times as fast as the previous record for any type of BCI.
“which is 3.4 times as fast as the previous record8”
From A high-performance speech neuroprosthesis (2023) · cites this paper as [8]
The Discussion (page unavailable) reports 90 characters per minute for this BCI and 60 characters per minute for an earlier EEG speller. That makes the reported speed 1.5 times the earlier rate, not 3.4 times the previous record for any type of BCI. (LLM, conf=0.88)
Checked by a reasoning model — may be wrong.
Experimental — accuracy not yet measuredDuring 14 days of data collection shortly after implantation, text decoding performance exceeded previous BCIs' communication speeds by a factor of 4 or more.
“During 14 days of data collection shortly after device implantation, we achieved high-performance text decoding, exceeding communication speeds of previous BCIs by a factor of 4 or more1,3,9”
From A high-performance neuroprosthesis for speech decoding and avatar control (2023) · cites this paper as [9]
The Discussion reports high-performance text decoding at 90 characters per minute, exceeding previously reported BCI speeds. But it also reports prior speeds of 40 characters per minute for intracortical point-and-click typing and 60 for a visually evoked EEG speller, so it does not support a fourfold advantage over previous BCIs generally. The Daily decoder retraining section says the copy-typing data spanned 28 days, while the Discussion refers to functionality more than 1,000 days after implantation, not data collection shortly after implantation. (LLM, conf=0.89)
Experimental — accuracy not yet measuredA text-decoding-specific approach uses large language models to automatically recalibrate MEA-based text neuroprostheses.
“A second approach, specific to text decoding, involves leveraging large language models to automatically recalibrate MEA-based text neuroprostheses.”
From The speech neuroprosthesis (2024) · cites this paper as [9]
The Daily decoder retraining section reports an offline test in which a language model corrected errors and enabled automatic, unsupervised recalibration of a handwriting decoder. It does not identify that model as a large language model, and it says the approach still needs online confirmation. (LLM, conf=0.88)
Experimental — accuracy not yet measuredA spelling framework using attempted speech or handwriting can facilitate a similar tradeoff, allowing users to spell arbitrary sentences with high accuracy but lower speeds.
“A similar tradeoff can be facilitated by having a spelling framework available on the neuroprosthesis, either using attempted speech120 or handwriting9, which would allow users to spell out arbitrary sentences with high accuracy (albeit with substantially lower speeds).”
From The speech neuroprosthesis (2024) · cites this paper as [9]
The Discussion (page unavailable) says the handwriting BCI lets its user express any sentence, and Decoding handwritten sentences reports high accuracy. The paper does not establish the claimed tradeoff of substantially lower speed; it describes 90 characters per minute as a new BCI speed benchmark. The attempted-speech alternative has a separate citation and is not assessed here. (LLM, conf=0.83)
Experimental — accuracy not yet measuredThe median real-time text-decoding rate exceeded the participant's typical assistive-device communication rate of 14.2 words per minute and was closer to naturalistic speaking rates than previously reported communication neuroprostheses.
“This decoding rate exceeds our participant’s typical communication rate using her assistive device (14.2 WPM; Supplementary Note 2) and is closer to naturalistic speaking rates than has been previously reported with communication neuroprostheses1,3,7–9.”
From A high-performance neuroprosthesis for speech decoding and avatar control (2023) · cites this paper as [9]
In the automatic recalibration approach, MEA activity is decoded into subword units, converted to sentences by language models, and the language model output is used as pseudo-labels to update the RNN decoder.
“In this approach, MEA activity is decoded into subword units and converted into sentences using language models; this language model output is used as a ‘pseudo-label’ to update the RNN decoding model9,16,105.”
From The speech neuroprosthesis (2024) · cites this paper as [9]
MEA-based decoders, including text decoders, have traditionally needed day-to-day recalibration because of signal instabilities and drift.
“MEA-based decoders, including text decoders9,15, have traditionally required day-to-day recalibration owing to signal instabilities and drift, requiring additional dedicated time from the user.”
From The speech neuroprosthesis (2024) · cites this paper as [9]
Several studies indicate that UEAs retain their functionality in humans for more than 1000 days.
“and several studies indicate that UEAs retain their functionality in humans for more than 1000 days (29, 31).”
From Visual percepts evoked with an intracortical 96-channel microelectrode array inserted in human occipital cortex (2021) · cites this paper as [29]
UEAs have been successfully implanted in human subjects in numerous studies and these devices have continued to work for many years.
“Although the duration of this study was limited to 6 months, UEAs have been successfully implanted in human subjects in numerous studies (25-29) and these devices have continued to work for many years.”
From Visual percepts evoked with an intracortical 96-channel microelectrode array inserted in human occipital cortex (2021) · cites this paper as [29]
BCIs based on hand movement activity have enabled typing speeds of between eight and 18 words per minute in people with paralysis.
“Recently, BCIs based on hand movement activity have enabled typing speeds of between eight and 18 words per minute in people with paralysis8,29.”
From A high-performance speech neuroprosthesis (2023) · cites this paper as [8]
Neurotechnologies can decode and project very specific and often rudimentary forms of thought, such as imagined handwriting, typing, and other intended and directed movements.
“Neurotechnologies can decode and project very specific and often rudimentary forms of thought, such as imagined handwriting, 40 typing, 27 and other kinds of intended and directed movements. 28”
From AI and Neurotechnology (2023) · cites this paper as [40]
Cites
Adam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946 · cited 3× · see citing works
“We used the gradient descent method “Adam” (Kingma and Ba, 2017) (beta1 = 0.”
MethodLearning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation
Cho, K., van Merrienboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., & Bengio, Y. (2014) · Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.3115/v1/d14-1179 · Cited by 25,545 · cited 1× · see citing works
“We used a two layer, gated recurrent unit RNN (Cho et al., 2014) to convert T5’s neural activity into a time series of character probabilities (see Extended Data Fig.”
MethodHinton, G., Deng, L., Yu, D., Dahl, G., Mohamed, A.-R., Jaitly, N., Senior, A., Vanhoucke, V., Nguyen, P., Sainath, T., et al. (2012) · IEEE Signal Processing Magazine · doi:10.1109/msp.2012.2205597 · Cited by 10,260 · see citing works
TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems
Abadi, M., Agarwal, A., Barham, P., Brevdo, E., Chen, Z., Citro, C., Corrado, G. S., Davis, A., Dean, J., Devin, M., et al. (2016) · arXiv · doi:10.48550/arxiv.1603.04467 · Cited by 9,408 · cited 2× · see citing works
“We used the “Adam” gradient descent method (Abadi et al., 2016; Kingma and Ba, 2017).”
MethodSpeech recognition with deep recurrent neural networks
Graves, A., Mohamed, A.-R., & Hinton, G. (2013) · 2013 IEEE International Conference on Acoustics, Speech and Signal Processing · doi:10.1109/icassp.2013.6638947 · Cited by 8,893 · cited 1× · see citing works
“Our methods are similar to neural network methods used in automatic speech recognition (Hinton et al., 2012; Graves et al., 2013; Zeyer et al., 2017; Xiong et al., 2017), but with some key changes made to achieve high performance on neural activity in a highly data-limited regime (1-10 hours of data, as opposed to 1,0…”
Method“Librispeech: an ASR corpus based on public domain audio books,”
Panayotov, V., Chen, G., Povey, D., & Khudanpur, S. (2015) · 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) · doi:10.1109/icassp.2015.7178964 · Cited by 6,072 · cited 1× · see citing works
“, 2017), but with some key changes made to achieve high performance on neural activity in a highly data-limited regime (1-10 hours of data, as opposed to 1,000+ hours (Cieri et al., 2004; Panayotov et al., 2015; He et al., 2019)).”
BackgroundGraves, A., Fernández, S., Gomez, F., & Schmidhuber, J. (2006) · Proceedings of the 23rd international conference on Machine learning - ICML '06 · doi:10.1145/1143844.1143891 · Cited by 5,370 · see citing works
Neuronal ensemble control of prosthetic devices by a human with tetraplegia
Hochberg, L. R., Serruya, M. D., Friehs, G. M., Mukand, J. A., Saleh, M., Caplan, A. H., Branner, A., Chen, D., Penn, R. D., & Donoghue, J. P. (2006) · Nature · doi:10.1038/nature04970 · Cited by 3,295 · see citing works
Self-citationReach and grasp by people with tetraplegia using a neurally controlled robotic arm
Hochberg, L. R., Bacher, D., Jarosiewicz, B., Masse, N. Y., Simeral, J. D., Vogel, J., Haddadin, S., Liu, J., Cash, S. S., van der Smagt, P., et al. (2012) · Nature · doi:10.1038/nature11076 · Cited by 2,719 · cited 1× · see citing works
“To train the RNN, we adapted neural network methods in automatic speech recognition to overcome two key challenges: (1) the time that each letter was written in the training data was unknown (as T5’s hand was paralysed), making it challenging to apply supervised learning Prepare: a Go Return”
MethodSelf-citationHigh-performance neuroprosthetic control by an individual with tetraplegia
Collinger, J. L., Wodlinger, B., Downey, J. E., Wang, W., Tyler-Kabara, E. C., Weber, D. J., McMorland, A. J., Velliste, M., Boninger, M. L., & Schwartz, A. B. (2013) · The Lancet · doi:10.1016/s0140-6736(12)61816-9 · Cited by 1,848 · cited 1× · see citing works
“…performance so far compared to other recording technologies (for example, EEG or electrocorticography) for restoring communication 7 , arm control 2,5 ) was similar for characters and straight lines, but the temporal dimensionality was more than twice as high for characters, suggesting that more…”
BackgroundLocalization of the motor hand area to a knob on the precentral gyrus. A new landmark
Yousry, T. A. (1997) · Brain · doi:10.1093/brain/120.1.141 · Cited by 1,840 · cited 1× · see citing works
“We tested this by recording neural activity from two microelectrode arrays in the hand ‘knob’ area of the precentral gyrus (a premotor area) 9,10 while our BrainGate study participant, T5, attempted to handwrite individual letters and symbols (Fig.”
MethodcuDNN: Efficient Primitives for Deep LearningPreprint
Chetlur, S., Woolley, C., Vandermersch, P., Cohen, J., Tran, J. J., Catanzaro, B., & Shelhamer, E. (2014) · arXiv (Cornell University) · doi:10.48550/arxiv.1410.0759 · Cited by 1,394 · cited 1× · see citing works
“For the RNN layers, we used the following variant of the gated recurrent unit RNN that is implemented by the cuDNN library (Chetlur et al., 2014):”
MethodAn Integrated Brain-Machine Interface Platform With Thousands of Channels
Musk, E., & Neuralink (2019) · Journal of Medical Internet Research · doi:10.2196/16194 · Cited by 1,206 · see citing works
High-speed spelling with a noninvasive brain–computer interface
Chen, X., Wang, Y., Nakanishi, M., Gao, X., Jung, T.-P., & Gao, S. (2015) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1508080112 · Cited by 969 · cited 2× · see citing works
“Commonly used BCIs for restoring communication are either flashing electroencephalogram (EEG) spellers 18,28–32 or intracortical point-and-click BCIs 6,7,33 .”
BackgroundMethodRestoring cortical control of functional movement in a human with quadriplegia
Bouton, C. E., Shaikhouni, A., Annetta, N. V., Bockbrader, M. A., Friedenberg, D. A., Nielson, D. M., Sharma, G., Sederberg, P. B., Glenn, B. C., Mysiw, W. J., et al. (2016) · Nature · doi:10.1038/nature17435 · Cited by 924 · see citing works
Speech synthesis from neural decoding of spoken sentences
Anumanchipalli, G. K., Chartier, J., & Chang, E. F. (2019) · Nature · doi:10.1038/s41586-019-1119-1 · Cited by 883 · cited 2× · see citing works
“Recently, speech-decoding BCIs have shown exciting promise for restoring rapid communication 16,17,34 , but their accuracies and vocabulary sizes require further improvement to support general-purpose use.”
BackgroundMethodAjiboye, A. B., Willett, F. R., Young, D. R., Memberg, W. D., Murphy, B. A., Miller, J. P., Walter, B. L., Sweet, J. A., Hoyen, H. A., Keith, M. W., et al. (2017) · The Lancet · doi:10.1016/s0140-6736(17)30601-3 · Cited by 862 · see citing works
Self-citationA high-performance brain–computer interface
Santhanam, G., Ryu, S. I., Yu, B. M., Afshar, A., & Shenoy, K. V. (2006) · Nature · doi:10.1038/nature04968 · Cited by 698 · cited 1× · see citing works
“Other discrete (classification-based) BCIs have also typically used directional movements with little temporal variety, which may have limited their accuracy and/or the size of the movement set 24,25 .”
BackgroundSelf-citationCognitive Control Signals for Neural Prosthetics
Musallam, S., Corneil, B. D., Greger, B., Scherberger, H., & Andersen, R. A. (2004) · Science · doi:10.1126/science.1097938 · Cited by 663 · cited 1× · see citing works
“Other discrete (classification-based) BCIs have also typically used directional movements with little temporal variety, which may have limited their accuracy and/or the size of the movement set 24,25 .”
BackgroundDecoding motor imagery from the posterior parietal cortex of a tetraplegic human
Aflalo, T., Kellis, S., Klaes, C., Lee, B., Shi, Y., Pejsa, K., Shanfield, K., Hayes-Jackson, S., Aisen, M., Heck, C., et al. (2015) · Science · doi:10.1126/science.aaa5417 · Cited by 649 · see citing works
A P300-based brain–computer interface for people with amyotrophic lateral sclerosis
Nijboer, F., Sellers, E. W., Mellinger, J., Jordan, M. A., Matuz, T., Furdea, A., Halder, S., Mochty, U., Krusienski, D. J., Vaughan, T. M., et al. (2008) · Clinical Neurophysiology · doi:10.1016/j.clinph.2008.03.034 · Cited by 633 · see citing works
2005 IEEE International Conference on Acoustics, Speech, and Signal Processing
(2005) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp.2005.1415922 · Cited by 629 · see citing works
Streaming End-to-end Speech Recognition for Mobile Devices
He, Y., Sainath, T. N., Prabhavalkar, R., McGraw, I., Alvarez, R., Zhao, D., Rybach, D., Kannan, A., Wu, Y., Pang, R., et al. (2019) · ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) · doi:10.1109/icassp.2019.8682336 · Cited by 601 · cited 1× · see citing works
“The character error rate decreased to 0.89% and the word error rate decreased to 3.4% averaged across all days, which is comparable to state-of-the-art speech recognition systems with word error rates of 4–5% 14,15 , putting it well within the range of usability.”
BackgroundInfluentialLocalization of the motor hand area to a knob on the precentral gyrus.
Yousry, T. A. (1997) · Medical Entomology and Zoology · Cited by 594 · see citing works
Fully Implanted Brain–Computer Interface in a Locked-In Patient with ALS
Vansteensel, M. J., Pels, E. G. M., Bleichner, M. G., Branco, M. P., Denison, T., Freudenburg, Z. V., Gosselaar, P., Leinders, S., Ottens, T. H., Van Den Boom, M. A., et al. (2016) · New England Journal of Medicine · doi:10.1056/nejmoa1608085 · Cited by 578 · cited 2× · see citing works
“Commonly used BCIs for restoring communication are either flashing electroencephalogram (EEG) spellers 18,28–32 or intracortical point-and-click BCIs 6,7,33 .”
BackgroundMethodTownsend, G., LaPallo, B. K., Boulay, C. B., Krusienski, D. J., Frye, G. E., Hauser, C. K., Schwartz, N. E., Vaughan, T. M., Wolpaw, J. R., & Sellers, E. W. (2010) · Clinical Neurophysiology · doi:10.1016/j.clinph.2010.01.030 · Cited by 570 · see citing works
Pandarinath, C., Nuyujukian, P., Blabe, C. H., Sorice, B. L., Saab, J., Willett, F. R., Hochberg, L. R., Shenoy, K. V., & Henderson, J. M. (2017) · eLife · doi:10.7554/elife.18554 · Cited by 566 · cited 7× · see citing works
“So far, a major focus of BCI research has been on restoring gross motor skills, such as reaching and grasping 1–5 or point-and-click typing with a computer cursor 6,7 .”
BackgroundMethodInfluentialSelf-citationSimeral, J. D., Kim, S.-P., Black, M. J., Donoghue, J. P., & Hochberg, L. R. (2011) · Journal of Neural Engineering · doi:10.1088/1741-2560/8/2/025027 · Cited by 496 · cited 1× · see citing works
“Current microelectrode array technology has been shown to retain functionality for more than 1,000 days after implant 37,38 (including here; see Extended Data Fig.”
Self-citationJarosiewicz, B., Sarma, A. A., Bacher, D., Masse, N. Y., Simeral, J. D., Sorice, B., Oakley, E. M., Blabe, C., Pandarinath, C., Gilja, V., et al. (2015) · Science Translational Medicine · doi:10.1126/scitranslmed.aac7328 · Cited by 340 · cited 16× · see citing works
“To train the RNN to account for drifts in the means of neural features that naturally accrue over time in neural recordings (Downey et al., 2018; Jarosiewicz et al., 2015), we added artificial perturbations to the feature means, similar to (Sussillo et al.”
BackgroundMethodResultInfluentialSelf-citationA theory of multineuronal dimensionality, dynamics and measurement
Gao, P., Trautmann, E., Yu, B., Santhanam, G., Ryu, S., Shenoy, K., & Ganguli, S. (2017) · doi:10.1101/214262 · Cited by 332 · cited 1× · see citing works
“Spatial and temporal dimensionality were estimated using the ‘participation ratio’ of the principal component analysis (PCA) eigenvalue spectrum, which quantifies approximately how many spatial or temporal dimensions are required to explain 80% of the variance in the patterns of neural activity 23 .”
MethodSelf-citationThe microsoft 2016 conversational speech recognition system
Xiong, W., Droppo, J., Huang, X., Seide, F., Seltzer, M., Stolcke, A., Yu, D., & Zweig, G. (2017) · 2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) · doi:10.1109/icassp.2017.7953159 · Cited by 326 · see citing works
Machine translation of cortical activity to text with an encoder–decoder framework
Makin, J. G., Moses, D. A., & Chang, E. F. (2020) · Nature Neuroscience · doi:10.1038/s41593-020-0608-8 · Cited by 321 · cited 2× · see citing works
“Recently, speech-decoding BCIs have shown exciting promise for restoring rapid communication 16,17,34 , but their accuracies and vocabulary sizes require further improvement to support general-purpose use.”
BackgroundMethodSpeech Recognition with Weighted Finite-State Transducers
Mohri, M., Pereira, F. M. B., & Riley, M. (2007) · Springer handbooks · doi:10.1007/978-3-540-49127-9_28 · Cited by 257 · cited 3× · see citing works
“Word frequencies are then encoded into a weighted finite state transducer (Mohri et al., 2008) that can be used to infer the most likely sequence of characters given the word frequency counts combined with the character probabilities output by the neural decoder.”
MethodInfluentialAccurate Estimation of Neural Population Dynamics without Spike Sorting
Trautmann, E. M., Stavisky, S. D., Lahiri, S., Ames, K. C., Kaufman, M. T., O’Shea, D. J., Vyas, S., Sun, X., Ryu, S. I., Ganguli, S., et al. (2019) · Neuron · doi:10.1016/j.neuron.2019.05.003 · Cited by 255 · see citing works
Self-citationSingle-Unit Stability Using Chronically Implanted Multielectrode Arrays
Dickey, A. S., Suminski, A., Amit, Y., & Hatsopoulos, N. G. (2009) · Journal of Neurophysiology · doi:10.1152/jn.90920.2008 · Cited by 234 · see citing works
Wav2Letter: an End-to-End ConvNet-based Speech Recognition SystemPreprint
Collobert, R., Puhrsch, C., & Synnaeve, G. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1609.03193 · Cited by 230 · see citing works
Making brain–machine interfaces robust to future neural variability
Sussillo, D., Stavisky, S. D., Kao, J. C., Ryu, S. I., & Shenoy, K. V. (2016) · Nature Communications · doi:10.1038/ncomms13749 · Cited by 214 · cited 1× · see citing works
“Finally, we envision that a combination of algorithmic innovations 42–44 and improvements to device stability will continue to reduce the need for daily decoder retraining.”
BackgroundSelf-citationDegenhart, A. D., Bishop, W. E., Oby, E. R., Tyler-Kabara, E. C., Chase, S. M., Batista, A. P., & Yu, B. M. (2020) · Nature Biomedical Engineering · doi:10.1038/s41551-020-0542-9 · Cited by 212 · cited 1× · see citing works
“Finally, we envision that a combination of algorithmic innovations 42–44 and improvements to device stability will continue to reduce the need for daily decoder retraining.”
BackgroundRecent advances in neurotechnologies with broad potential for neuroscience research
Vázquez-Guardado, A., Yang, Y., Bandodkar, A. J., & Rogers, J. A. (2020) · Nature Neuroscience · doi:10.1038/s41593-020-00739-8 · Cited by 202 · cited 2× · see citing works
“New developments are under way for implant designs that increase the electrode count by at least an order of magnitude, which will further improve performance and longevity 35,36,40,41 .”
BackgroundMugler, E. M., Patton, J. L., Flint, R. D., Wright, Z. A., Schuele, S. U., Rosenow, J., Shih, J. J., Krusienski, D. J., & Slutzky, M. W. (2014) · Journal of Neural Engineering · doi:10.1088/1741-2560/11/3/035015 · Cited by 192 · cited 1× · see citing works
“Recently, speech-decoding BCIs have shown exciting promise for restoring rapid communication 16,17,34 , but their accuracies and vocabulary sizes require further improvement to support general-purpose use.”
BackgroundHow do People Type on Mobile Devices?
Palin, K., Feit, A. M., Kim, S., Kristensson, P. O., & Oulasvirta, A. (2019) · International Conference on Human-Computer Interaction with Mobile Devices and Services (MobileHCI) · doi:10.1145/3338286.3340120 · Cited by 191 · cited 1× · see citing works
“To our knowledge, these typing speeds exceed those reported for any other BCI, and are comparable to typical smartphone typing speeds of individuals in the age group of our participant (115 characters per minute) 8 .”
MethodA comprehensive study of deep bidirectional LSTM RNNS for acoustic modeling in speech recognition
Zeyer, A., Doetsch, P., Voigtlaender, P., Schlüter, R., & Ney, H. (2017) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp.2017.7952599 · Cited by 176 · cited 1× · see citing works
“Our methods are similar to neural network methods used in automatic speech recognition (Hinton et al., 2012; Graves et al., 2013; Zeyer et al., 2017; Xiong et al., 2017), but with some key changes made to achieve high performance on neural activity in a highly data-limited regime (1-10 hours of data, as opposed to 1,0…”
MethodHand Knob Area of Premotor Cortex Represents the Whole Body in a Compositional Way
Willett, F. R., Deo, D. R., Avansino, D. T., Rezaii, P., Hochberg, L. R., Henderson, J. M., & Shenoy, K. V. (2020) · Cell · doi:10.1016/j.cell.2020.02.043 · Cited by 171 · cited 1× · see citing works
“We tested this by recording neural activity from two microelectrode arrays in the hand ‘knob’ area of the precentral gyrus (a premotor area) 9,10 while our BrainGate study participant, T5, attempted to handwrite individual letters and symbols (Fig.”
MethodSelf-citationIndependent home use of a brain-computer interface by people with amyotrophic lateral sclerosis
Wolpaw, J. R., Bedlack, R. S., Reda, D. J., Ringer, R. J., Banks, P. G., Vaughan, T. M., Heckman, S. M., McCane, L. M., Carmack, C. S., Winden, S., et al. (2018) · Neurology · doi:10.1212/wnl.0000000000005812 · Cited by 170 · cited 2× · see citing works
“Commonly used BCIs for restoring communication are either flashing electroencephalogram (EEG) spellers 18,28–32 or intracortical point-and-click BCIs 6,7,33 .”
BackgroundMethodCortical control of a tablet computer by people with paralysis
Nuyujukian, P., Albites Sanabria, J., Saab, J., Pandarinath, C., Jarosiewicz, B., Blabe, C. H., Franco, B., Mernoff, S. T., Eskandar, E. N., Simeral, J. D., et al. (2018) · PLOS ONE · doi:10.1371/journal.pone.0204566 · Cited by 170 · cited 1× · see citing works
“and general-purpose computer use 39 .”
Self-citationIntracortical recording stability in human brain–computer interface users
Downey, J. E., Schwed, N., Chase, S. M., Schwartz, A. B., & Collinger, J. L. (2018) · Journal of Neural Engineering · doi:10.1088/1741-2552/aab7a0 · Cited by 149 · cited 1× · see citing works
“We found that when only 2–7 days passed between sessions, performance was reasonable with no decoder retraining (11.1% raw error rate, 1.5% with a language model), as might be expected from previous work showing the short-term stability of neural recordings 19–21 .”
ResultInfluentialMcCane, L. M., Heckman, S. M., McFarland, D. J., Townsend, G., Mak, J. N., Sellers, E. W., Zeitlin, D., Tenteromano, L. M., Wolpaw, J. R., & Vaughan, T. M. (2015) · Clinical Neurophysiology · doi:10.1016/j.clinph.2015.01.013 · Cited by 141 · see citing works
The Argo: a high channel count recording system for neural recording <i>in vivo</i>
Sahasrabuddhe, K., Khan, A. A., Singh, A. P., Stern, T. M., Ng, Y., Tadić, A., Orel, P., LaReau, C., Pouzzner, D., Nishimura, K., et al. (2021) · Journal of Neural Engineering · doi:10.1088/1741-2552/abd0ce · Cited by 124 · see citing works
Neural Point-and-Click Communication by a Person With Incomplete Locked-In Syndrome
Bacher, D., Jarosiewicz, B., Masse, N. Y., Stavisky, S. D., Simeral, J. D., Newell, K., Oakley, E. M., Cash, S. S., Friehs, G. M., & Hochberg, L. R. (2014) · Neurorehabilitation and neural repair · doi:10.1177/1545968314554624 · Cited by 123 · cited 1× · see citing works
“Commonly used BCIs for restoring communication are either flashing electroencephalogram (EEG) spellers 18,28–32 or intracortical point-and-click BCIs 6,7,33 .”
MethodSelf-citationChristie, B. P., Tat, D. M., Irwin, Z. T., Gilja, V., Nuyujukian, P., Foster, J. D., Ryu, S. I., Shenoy, K. V., Thompson, D. E., & Chestek, C. A. (2015) · Journal of Neural Engineering · doi:10.1088/1741-2560/12/1/016009 · Cited by 103 · cited 1× · see citing works
“, 2019), and that neural decoding performance is comparable (within 5%) to using sorted units (Chestek et al., 2011; Christie et al., 2014) – although see also (Todorova et al.”
ResultSelf-citationWilliams, A. H., Poole, B., Maheswaranathan, N., Dhawale, A. K., Fisher, T., Wilson, C. D., Brann, D. H., Trautmann, E. M., Ryu, S., Shusterman, R., et al. (2020) · Neuron · doi:10.1016/j.neuron.2019.10.020 · Cited by 102 · see citing works
Self-citationA cryptography-based approach for movement decoding
Dyer, E. L., Gheshlaghi Azar, M., Perich, M. G., Fernandes, H. L., Naufel, S., Miller, L. E., & Körding, K. P. (2017) · Nature Biomedical Engineering · doi:10.1038/s41551-017-0169-7 · Cited by 70 · see citing works
Challenges for Large-Scale Cortical Interfaces
Nurmikko, A. (2020) · Neuron · doi:10.1016/j.neuron.2020.10.015 · Cited by 68 · cited 2× · see citing works
“New developments are under way for implant designs that increase the electrode count by at least an order of magnitude, which will further improve performance and longevity 35,36,40,41 .”
BackgroundBullard, A. J., Hutchison, B. C., Lee, J., Chestek, C. A., & Patil, P. G. (2020) · Neuromodulation: Technology at the Neural Interface · doi:10.1111/ner.13069 · Cited by 65 · see citing works
Pels, E. G. M., Aarnoutse, E. J., Ramsey, N. F., & Vansteensel, M. J. (2017) · Neurorehabilitation and Neural Repair · doi:10.1177/1545968317714577 · Cited by 33 · see citing works
Toward Optimal Target Placement for Neural Prosthetic Devices
Cunningham, J. P., Yu, B. M., Gilja, V., Ryu, S. I., & Shenoy, K. V. (2008) · Journal of Neurophysiology · doi:10.1152/jn.90833.2008 · Cited by 25 · see citing works
Self-citationEleryan, A., Vaidya, M., Southerland, J., Badreldin, I. S., Balasubramanian, K., Fagg, A. H., Hatsopoulos, N., & Oweiss, K. (2014) · Frontiers in Neuroengineering · doi:10.3389/fneng.2014.00023 · Cited by 20 · see citing works
Willett, F. R., Murphy, B. A., Memberg, W. D., Blabe, C. H., Pandarinath, C., Walter, B. L., Sweet, J. A., Miller, J. P., Henderson, J. M., Shenoy, K. V., et al. (2017) · Journal of Neural Engineering · doi:10.1088/1741-2552/aa5990 · Cited by 16 · see citing works
Self-citationWilliams, A. H., Poole, B., Maheswaranathan, N., et al. (2019) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/661165 · Cited by 15 · cited 1× · see citing works
“We used a time-alignment technique to remove temporal variability 11 , which revealed notably consistent underlying patterns of neural activity that are unique to each character (Fig.”
Method
Flagged first, then by citations
Cited by
Sorted by citations
Rajpurkar, P., Chen, E., Banerjee, O., & Topol, E. J. (2022) · Nature Medicine · doi:10.1038/s41591-021-01614-0 · Cited by 3,212 · see citing works
“Moreover, AI’s ability to directly transform brain waves to speech or text has remarkable potential value for patients with aphasia or locked-in syndrome who have had strokes 62 .”
BackgroundSpinal cord injury: molecular mechanisms and therapeutic interventions
Hu, X., Xu, W., Ren, Y., Wang, Z., He, X., Huang, R., Ma, B., Zhao, J., Zhu, R., & Cheng, L. (2023) · Signal Transduction and Targeted Therapy · doi:10.1038/s41392-023-01477-6 · Cited by 823 · see citing works
“Another clinical trial showed that the neural activities of the motor cortex can be decoded into handwritten actions through BCI, significantly improving the writing speed and accuracy of paralyzed hands in patients with SCI (NCT00912041).(383)”
BackgroundWalking naturally after spinal cord injury using a brain–spine interface
Lorach, H., Galvez, A., Spagnolo, V., Martel, F., Karakas, S., Intering, N., Vat, M., Faivre, O., Harte, C., Komi, S., et al. (2023) · Nature · doi:10.1038/s41586-023-06094-5 · Cited by 631 · see citing works
A high-performance speech neuroprosthesis
Willett, F. R., Kunz, E. M., Fan, C., Avansino, D. T., Wilson, G. H., Choi, E. Y., Kamdar, F., Glasser, M. F., Hochberg, L. R., Druckmann, S., et al. (2023) · Nature · doi:10.1038/s41586-023-06377-x · Cited by 581 · see citing works
Self-citationA high-performance neuroprosthesis for speech decoding and avatar control
Metzger, S. L., Littlejohn, K. T., Silva, A. B., Moses, D. A., Seaton, M. P., Wang, R., Dougherty, M. E., Liu, J. R., Wu, P., Berger, M. A., et al. (2023) · Nature · doi:10.1038/s41586-023-06443-4 · Cited by 522 · see citing works
Semantic reconstruction of continuous language from non-invasive brain recordings
Tang, J., LeBel, A., Jain, S., & Huth, A. G. (2023) · Nature Neuroscience · doi:10.1038/s41593-023-01304-9 · Cited by 442 · see citing works
“No statistical methods were used to predetermine sample sizes, but our sample sizes are similar to those reported in previous publications 1,3,4,18,19 .”
BackgroundMethodFlexible brain–computer interfaces
Tang, X., Shen, H., Zhao, S., Li, N., & Jia Liu, B. (2023) · Nature Electronics · doi:10.1038/s41928-022-00913-9 · Cited by 386 · see citing works
“Clinical BCI studies have used Utah arrays in patients to, for example, control robotic limbs 3 , restore sensations with microstimulation 9 and decode dexterous handwriting movement for text translation 10 .”
MethodLarge-scale neural recordings with single neuron resolution using Neuropixels probes in human cortex
Paulk, A. C., Kfir, Y., Khanna, A. R., Mustroph, M. L., Trautmann, E. M., Soper, D. J., Stavisky, S. D., Welkenhuysen, M., Dutta, B., Shenoy, K. V., et al. (2022) · Nature Neuroscience · doi:10.1038/s41593-021-00997-0 · Cited by 283 · see citing works
“3a) and Utah arrays 22,23 , as well as single lowered electrodes as done during DBS surgeries 18,44 , have shown that we can parse and decode complex cognitive processes and motor control even from a small area of cortex.”
BackgroundSelf-citationRecent advances in regenerative biomaterials
Cao, D., & Ding, J. (2022) · Regenerative Biomaterials · doi:10.1093/rb/rbac098 · Cited by 274 · see citing works
Fernández, E., Alfaro, A., Soto-Sánchez, C., Gonzalez-Lopez, P., Lozano, A. M., Peña, S., Grima, M. D., Rodil, A., Gómez, B., Chen, X., et al. (2021) · Journal of Clinical Investigation · doi:10.1172/jci151331 · Cited by 250 · see citing works
“A recent systematic review of human neuroprosthetic research using the UEA reported that the mean number of days of UEA implantation across all participants at the time of the study was at least 578 days (30), and several studies indicate that UEAs retain their functionality in humans for more than 1000 days (29, 31).”
An Accurate and Rapidly Calibrating Speech Neuroprosthesis
Card, N. S., Wairagkar, M., Iacobacci, C., Hou, X., Singer-Clark, T., Willett, F. R., Kunz, E. M., Fan, C., Vahdati Nia, M., Deo, D. R., et al. (2024) · New England Journal of Medicine · doi:10.1056/nejmoa2314132 · Cited by 238 · see citing works
Self-citationRecommendations for Responsible Development and Application of Neurotechnologies
Goering, S., Klein, E. P., Sullivan, L. S., Wexler, A., y Arcas, B. A., Bi, G., Carmena, J. M., Fins, J. J., Friesen, P., Gallant, J. L., et al. (2021) · Neuroethics · doi:10.1007/s12152-021-09468-6 · Cited by 236 · see citing works
Decoding speech perception from non-invasive brain recordings
Défossez, A., Caucheteux, C., Rapin, J., Kabeli, O., & King, J.-R. (2023) · Nature Machine Intelligence · doi:10.1038/s42256-023-00714-5 · Cited by 242 · see citing works
“Brain Computer Interface (BCI) has been raising high expectations to detect [13, 22, 49, 72] and restore communication abilities in such patients [17, 37, 48, 66, 88, 93].”
BackgroundMethodInfluentialEmerging 4D Printing Strategies for Next‐Generation Tissue Regeneration and Medical Devices
Wang, Y., Cui, H., Esworthy, T. J., Mei, D., Wang, Y., & Zhang, L. G. (2021) · Advanced Materials · doi:10.1002/adma.202109198 · Cited by 204 · see citing works
Hou, B., Yi, L., Li, C., Zhao, H., Zhang, R., Zhou, B., & Liu, X. (2022) · Nature Electronics · doi:10.1038/s41928-022-00841-8 · Cited by 209 · see citing works
Mitchell, P. J., Lee, S. C. M., Yoo, P. E., Morokoff, A. P., Sharma, R. P., Williams, D. L., MacIsaac, C., Howard, M. E., Irving, L., Vrljic, I., et al. (2023) · JAMA Neurology · doi:10.1001/jamaneurol.2022.4847 · Cited by 199 · see citing works
Recent progress and challenges in the treatment of spinal cord injury
Tian, T., Zhang, S., & Yang, M. (2023) · Protein & Cell · doi:10.1093/procel/pwad003 · Cited by 196 · see citing works
CTNet: a convolutional transformer network for EEG-based motor imagery classification
Zhao, W., Jiang, X., Zhang, B., Xiao, S., & Weng, S. (2024) · Scientific Reports · doi:10.1038/s41598-024-71118-7 · Cited by 191 · see citing works
Non-Invasive Brain-Computer Interfaces: State of the Art and Trends
Edelman, B., Zhang, S.-L., Schalk, G., Brunner, P., Müller-Putz, G., Guan, C.-T., & He, B. (2024) · IEEE Reviews in Biomedical Engineering · doi:10.1109/rbme.2024.3449790 · Cited by 188 · see citing works
“For example, rather than decoding an intended sentence from a word or letter bank presented on a computer screen, modern BCIs can now decode handwriting signals directly from the brain [19].”
Brain–computer interface: trend, challenges, and threats
Maiseli, B. J., Abdalla, A. T., Massawe, L. V., Mbise, M. E., Mkocha, K., Ally, N., Ismail, M., James, M., & Kimambo, S. (2023) · Brain Informatics · doi:10.1186/s40708-023-00199-3 · Cited by 187 · see citing works
“Studies demonstrate that BCI may significantly improve the quality of life for physically challenged people [77, 102].”
BackgroundInfluentialMetzger, S. L., Liu, J. R., Moses, D. A., Dougherty, M. E., Seaton, M. P., Littlejohn, K. T., Chartier, J., Anumanchipalli, G. K., Tu-Chan, A., Ganguly, K., et al. (2022) · Nature Communications · doi:10.1038/s41467-022-33611-3 · Cited by 153 · see citing works
Directly wireless communication of human minds via non-invasive brain-computer-metasurface platform
Ma, Q., Gao, W., Xiao, Q., et al. (2022) · eLight · doi:10.1186/s43593-022-00019-x · Cited by 148 · see citing works
“The P300-based BCIs, which identify the operator’s intention from the P300 potentials evoked by flashes of the corresponding buttons on the control panel, have been frequently used [5–7] to assist the handicapped [8] or medical applications [3].”
MethodState-of-the-Art on Brain-Computer Interface Technology
Pekša, J., & Mamchur, D. (2023) · Sensors · doi:10.3390/s23136001 · Cited by 148 · see citing works
“Other studies have looked into how BCI technology can be used for more complex forms of communication, such as typing on a computer keyboard or giving speech commands via voice recognition software [131–136].”
BackgroundChaudhary, U., VLACHOS, I. K., Zimmermann, J. B., Espinosa, A., Tonin, A., Jaramillo-Gonzalez, A., Khalili‐Ardali, M., Topka, H. R., Lehmberg, J., Friehs, G. M., et al. (2022) · Nature Communications · doi:10.1038/s41467-022-28859-8 · Cited by 141 · see citing works
The Argo: a high channel count recording system for neural recording <i>in vivo</i>
Sahasrabuddhe, K., Khan, A. A., Singh, A. P., Stern, T. M., Ng, Y., Tadić, A., Orel, P., LaReau, C., Pouzzner, D., Nishimura, K., et al. (2021) · Journal of Neural Engineering · doi:10.1088/1741-2552/abd0ce · Cited by 124 · see citing works
Bio-inspired electronics: Soft, biohybrid, and “living” neural interfaces
Boufidis, D., Garg, R., Angelopoulos, E., Cullen, D. K., & Vitale, F. (2025) · Nature Communications · doi:10.1038/s41467-025-57016-0 · Cited by 154 · see citing works
“In contrast, invasive technologies such as brain-computer (BCI) and brain-machine (BMI) interfaces allow high-bandwidth recordings from deeper brain structures, including both intracortical and subcortical targets 5,13,14 .”
Transfer learning for motor imagery based brain–computer interfaces: A tutorial
Wu, D., Jiang, X., & Peng, R. (2022) · Neural Networks · doi:10.1016/j.neunet.2022.06.008 · Cited by 137 · see citing works
Flexible Electrodes for Brain–Computer Interface System
Wang, J., Wang, T., Liu, H., Wang, K., Kumi, M., Feng, Z., Li, P., & Huang, W. (2023) · Advanced Materials · doi:10.1002/adma.202211012 · Cited by 138 · see citing works
Conformal in-ear bioelectronics for visual and auditory brain-computer interfaces
Wang, Z., Shi, N., Zhang, Y., Zheng, N., Li, H., Jiao, Y., Cheng, J., Wang, Y., Zhang, X., Chen, Y., et al. (2023) · Nature Communications · doi:10.1038/s41467-023-39814-6 · Cited by 141 · see citing works
Shin, J. H., Kwon, J., Kim, J. U., Ryu, H., Ok, J., Joon Kwon, S., Park, H., & Kim, T.-I. (2022) · npj Flexible Electronics · doi:10.1038/s41528-022-00164-w · Cited by 123 · see citing works
“…between the brain and machines 14 – 16 , can be established using electrophysiological sensors, such as brain-surface 17 , brain-penetrating 18,19 , and skin-surface electrodes 20,21 , to assess electrocorticography (ECoG), local fi eld potentials (LFPs), and electroencephalography (EEG),…”
BackgroundImplantable brain machine interfaces: first-in-human studies, technology challenges and trends
Rapeaux, A. B., & Constandinou, T. G. (2021) · Current Opinion in Biotechnology · doi:10.1016/j.copbio.2021.10.001 · Cited by 119 · see citing works
TC-Net: A Transformer Capsule Network for EEG-based emotion recognition
Wei, Y., Liu, Y., Li, C., Cheng, J., Song, R., & Chen, X. (2022) · Computers in Biology and Medicine · doi:10.1016/j.compbiomed.2022.106463 · Cited by 117 · see citing works
Preparatory activity and the expansive null-space
Churchland, M. M., & Shenoy, K. V. (2024) · Nature reviews. Neuroscience · doi:10.1038/s41583-024-00796-z · Cited by 126 · see citing works
“A final possibility is to abandon the idea of decoder-potent dimensions entirely, and use neural networks 213 or other highly non-linear means 169 to convert factor trajectories into decoded variables.”
Self-citationDecoding spoken English from intracortical electrode arrays in dorsal precentral gyrus
Wilson, G. H., Stavisky, S. D., Willett, F. R., Avansino, D. T., Kelemen, J. N., Hochberg, L. R., Henderson, J. M., Druckmann, S., & Shenoy, K. V. (2020) · Journal of Neural Engineering · doi:10.1088/1741-2552/abbfef · Cited by 103 · see citing works
Self-citationDeep Learning Enabled Neck Motion Detection Using a Triboelectric Nanogenerator
An, S., Pu, X., Zhou, S., Wu, Y., Li, G., Xing, P., Zhang, Y., & Hu, C. (2022) · ACS Nano · doi:10.1021/acsnano.2c02149 · Cited by 116 · see citing works
Brain-Computer Interface: Applications to Speech Decoding and Synthesis to Augment Communication
Luo, S., Rabbani, Q., & Crone, N. E. (2022) · Neurotherapeutics · doi:10.1007/s13311-022-01190-2 · Cited by 106 · see citing works
“Significant advances have been made to assist these individuals through the use of other types of BCIs, including those using P300 [3], motor imagery [4], handwriting [5], and steady-state visually evoked potential [6].”
BackgroundThe state of clinical trials of implantable brain–computer interfaces
Patrick-Krueger, K. M., Burkhart, I., & Contreras-Vidal, J. L. (2024) · Nature Reviews Bioengineering · doi:10.1038/s44222-024-00239-5 · Cited by 130 · see citing works
Willsey, M. S., Nason, S. R., Ensel, S., Temmar, H., Mender, M. J., Costello, J. T., Patil, P. G., & Chestek, C. A. (2022) · Nature Communications · doi:10.1038/s41467-022-34452-w · Cited by 104 · see citing works
Silva, A. B., Littlejohn, K. T., Liu, J. R., Moses, D. A., & Chang, E. F. (2024) · Nature Reviews Neuroscience · doi:10.1038/s41583-024-00819-9 · Cited by 103 · see citing works
“In this approach, MEA activity is decoded into subword units and converted into sentences using language models; this language model output is used as a ‘pseudo-label’ to update the RNN decoding model 9,16,105 .”
InfluentialBioelectronic medicine: Preclinical insights and clinical advances
Pavlov, V. A., & Tracey, K. J. (2022) · Neuron · doi:10.1016/j.neuron.2022.09.003 · Cited by 106 · see citing works
Proietti, T., O’Neill, C., Gerez, L. F., et al. (2023) · Science Translational Medicine · doi:10.1126/scitranslmed.add1504 · Cited by 102 · see citing works
“In the future, improved strategies of motion intention detection will be investigated to overcome this requirement for the user, for example, by using a brain-computer interface (38, 39).”
MethodNeuroprosthetics: from sensorimotor to cognitive disorders
Gupta, A., Vardalakis, N., & Wagner, F. B. (2023) · Communications Biology · doi:10.1038/s42003-022-04390-w · Cited by 108 · see citing works
Kim, S., Baek, S., Sluyter, R., Konstantinov, K., Kim, J. H., Kim, S., & Kim, J. H. (2023) · EcoMat · doi:10.1002/eom2.12356 · Cited by 102 · see citing works
Artificial intelligence in neurology: opportunities, challenges, and policy implications
Voigtlaender, S., Pawelczyk, J., Geiger, M., Vaios, E., Karschnia, P., Cudkowicz, M., Dietrich, J., Haraldsen, I., Feigin, V. L., Owolabi, M. O., et al. (2024) · Journal of Neurology · doi:10.1007/s00415-024-12220-8 · Cited by 106 · see citing works
“…and restoration of neurological function difficult or impossible Outcome prediction, resource triaging BCI-based neurorehabilitation [96–98] Effective prevention and rehabilitation, leading to greater societal participation Lack of comprehensive, unbiased, accessible datasets for…”
Angrick, M., Luo, S., Rabbani, Q., Candrea, D. N., Shah, S., Milsap, G., Anderson, W. S., Gordon, C. R., Rosenblatt, K. R., Clawson, L., et al. (2024) · Scientific Reports · doi:10.1038/s41598-024-60277-2 · Cited by 96 · see citing works
A memristor-based adaptive neuromorphic decoder for brain–computer interfaces
Liu, Z., Mei, J., Tang, J., Xu, M., Gao, B., Wang, K., Ding, S., Liu, Q., Qin, Q., Chen, W., et al. (2025) · Nature Electronics · doi:10.1038/s41928-025-01340-2 · Cited by 98 · see citing works
“These trials were filtered in [1, 10] Hz using a Chebyshev type-I infinite impulse response filter.”
Transformer-based Spatial-Temporal Feature Learning for EEG DecodingPreprint
Song, Y., Jia, X., Yang, L., & Xie, L. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2106.11170 · Cited by 109 · see citing works
“B RAIN-computer interface (BCI) is a technology that provides the possibility of establishing direct connections between the brain and external devices [1].”
BackgroundLuo, S., Angrick, M., Coogan, C., Candrea, D. N., Wyse‐Sookoo, K., Shah, S., Rabbani, Q., Milsap, G. W., Weiss, A. R., Anderson, W. S., et al. (2023) · Advanced Science · doi:10.1002/advs.202304853 · Cited by 90 · see citing works
The neuroethics of disorders of consciousness: a brief history of evolving ideas
Young, M. J., Bodien, Y. G., Giacino, J. T., Fins, J. J., Truog, R. D., Hochberg, L. R., & Edlow, B. L. (2021) · Brain · doi:10.1093/brain/awab290 · Cited by 92 · see citing works
“One exception to this is the field of implanted brain–computer interfaces, where there is potential to enable rich efferent communication by people otherwise completely locked-in.(38) These same technologies may, in the future, similarly be used diagnostically, as the restoration of reliable and verifiable communicati…”
BackgroundSelf-citationInterim Safety Profile From the Feasibility Study of the BrainGate Neural Interface System
Rubin, D. B., Ajiboye, A. B., Barefoot, L., Bowker, M., Cash, S. S., Chen, D., Donoghue, J. P., Eskandar, E. N., Friehs, G., Grant, C., et al. (2023) · Neurology · doi:10.1212/wnl.0000000000201707 · Cited by 81 · see citing works
Self-citationClawson, W. P., & Levin, M. G. (2022) · Biological Journal of the Linnean Society · doi:10.1093/biolinnean/blac073 · Cited by 77 · see citing works
The science and engineering behind sensitized brain-controlled bionic hands
Pandarinath, C., & Bensmaı̈a, S. J. (2021) · Physiological Reviews · doi:10.1152/physrev.00034.2020 · Cited by 76 · see citing works
“…Nuyujukian* et al. 2017 [38] Intracortical (Utah array) Point-and - click (grid) T6, T7, T5 ALS (2), SCI (1) 2.4 2.4 “ “ “ T5 SCI 4.2* 4.2* Willett et al. 2021 [4] Intracortical (Utah array) Handwriting T5 SCI ~6.18** - Silversmith et al. 2021 [245] ECoG Point-and - click (grid) B1…”
InfluentialWillsey, M. S., Shah, N. P., Avansino, D. T., Hahn, N. V., Jamiolkowski, R. M., Kamdar, F. B., Hochberg, L. R., Willett, F. R., & Henderson, J. M. (2025) · Nature Medicine · doi:10.1038/s41591-024-03341-8 · Cited by 88 · see citing works
“27.”
Self-citationModulating Brain Activity with Invasive Brain–Computer Interface: A Narrative Review
Zhao, Z., Nie, C., Jiang, C.-T., Cao, S.-H., Tian, K.-X., Yu, S., & Gu, J. (2023) · Brain Sciences · doi:10.3390/brainsci13010134 · Cited by 86 · see citing works
“Using this invasive BCI, patients achieved a typing speed of 90 characters/min, with an online original accuracy of 94.1% and offline automatic correction accuracy of more than 99% [50], while the typing speed on typical smartphones is about 115 characters/min (see Figure 2).”
MethodInfluentialClassification of EEG signals using Transformer based deep learning and ensemble models
Zeynali, M., Seyedarabi, H., & Afrouzian, R. (2023) · Biomedical Signal Processing and Control · doi:10.1016/j.bspc.2023.105130 · Cited by 79 · see citing works
In‐Sensor Computing with Visual‐Tactile Perception Enabled by Mechano‐Optical Artificial Synapse
Guo, J., Guo, F., Zhao, H., Yang, H., Du, X., Fan, F., Liu, W., Zhang, Y., Tu, D., & Hao, J. (2025) · Advanced Materials · doi:10.1002/adma.202419405 · Cited by 79 · see citing works
Deep learning approaches for neural decoding across architectures and recording modalities
Livezey, J. A., & Glaser, J. I. (2020) · Briefings in Bioinformatics · doi:10.1093/bib/bbaa355 · Cited by 84 · see citing works
Vorontsova, D., Menshikov, I., Zubov, A., et al. (2021) · Sensors · doi:10.3390/s21206744 · Cited by 76 · see citing works
Hu, J., Small, H., Kean, H., Takahashi, A., Zekelman, L., Kleinman, D., Ryan, E., Nieto-Castañón, A., Ferreira, V., & Fedorenko, E. (2023) · Cerebral Cortex · doi:10.1093/cercor/bhac350 · Cited by 73 · see citing works
A Human-Centric Metaverse Enabled by Brain-Computer Interface: A Survey
Zhu, H. Y., Hieu, N. Q., Hoang, D. T., Nguyen, D. N., & Lin, C. (2024) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2024.3387124 · Cited by 75 · see citing works
“A recent innovation is the research by [128], which explores decoding brain activity into text.”
BackgroundEEG-based brain-computer interface enables real-time robotic hand control at individual finger level
Ding, Y., Udompanyawit, C., Zhang, Y., & He, B. (2025) · Nature Communications · doi:10.1038/s41467-025-61064-x · Cited by 98 · see citing works
“Both invasive and noninvasive BCI systemshavebeendevelopedforbiologicallimbmanipulation 3 – 6 ,robotic device control 7 – 11 , and linguistic communication 12 – 15 .”
Going beyond primary motor cortex to improve brain–computer interfaces
Gallego, J. Á., Makin, T. R., & McDougle, S. D. (2022) · Trends in Neurosciences · doi:10.1016/j.tins.2021.12.006 · Cited by 70 · see citing works
“’ Yet despite the impressive advances in BCI technology over the past several years [2,6,7], even seemingly simple tasks such as controlling a computer cursor with a BCI turn out to be challenging, often more so than performing analogous tasks with a limb [8].”
BackgroundSignal acquisition of brain–computer interfaces: A medical-engineering crossover perspective review
Sun, Y., Chen, X., Liu, B., Liang, L., Wang, Y., Gao, S., & Gao, X. (2024) · Fundamental Research · doi:10.1016/j.fmre.2024.04.011 · Cited by 72 · see citing works
“In 2021, Willett et al. employed the BrainGate system to achieve a brain-to-text communication scheme by interpreting handwriting [146] .”
Recent applications of EEG-based brain-computer-interface in the medical field
Liu, X., Wang, W.-L., Liu, M., Chen, M., Pereira, T., Doda, D. Y., Ke, Y., Wang, S., Dong, W., Tong, X., et al. (2025) · Military Medical Research · doi:10.1186/s40779-025-00598-z · Cited by 84 · see citing works
“Figure 1 illustrates the overall process of BCIs, including rehabilitation applications [92], 3 daily communication methods [93], epilepsy monitoring and prediction [94, 95], sleep monitoring, 4 brain stimulation techniques for treat-ment, and other related applications.”
Stabilizing brain-computer interfaces through alignment of latent dynamics
Karpowicz, B. M., Ali, Y. H., Wimalasena, L. N., Sedler, A. R., Keshtkaran, M. R., Bodkin, K. L., Ma, X., Rubin, D. B., Williams, Z. M., Cash, S. S., et al. (2025) · Nature Communications · doi:10.1038/s41467-025-59652-y · Cited by 73 · see citing works
“In recent years, iBCIs have attained impressive performance in a range of applications, including the control of anthropomorphic robotic arms, stimulation of paralyzed muscles to enable reaching and grasping, and even rapid decoding of handwriting 3–6 .”
Self-citationBrain control of bimanual movement enabled by recurrent neural networks
Deo, D. R., Willett, F. R., Avansino, D. T., Hochberg, L. R., Henderson, J. M., & Shenoy, K. V. (2024) · Scientific Reports · doi:10.1038/s41598-024-51617-3 · Cited by 73 · see citing works
Self-citationClosed-loop stimulation using a multiregion brain-machine interface has analgesic effects in rodents
Sun, G., Zeng, F., McCartin, M., Zhang, Q., Xu, H., Liu, Y., Chen, Z. S., & Wang, J. (2022) · Science Translational Medicine · doi:10.1126/scitranslmed.abm5868 · Cited by 61 · see citing works
Volitional activation of remote place representations with a hippocampal brain–machine interface
Lai, C., Tanaka, S., Harris, T. D., & Lee, A. K. (2023) · Science · doi:10.1126/science.adh5206 · Cited by 67 · see citing works
Gallego-Carracedo, C., Perich, M. G., Chowdhury, R. H., Miller, L. E., & Gallego, J. Á. (2022) · eLife · doi:10.7554/elife.73155 · Cited by 59 · see citing works
Decoding motor plans using a closed-loop ultrasonic brain–machine interface
Griggs, W. S., Norman, S. L., Deffieux, T., Segura, F., Osmanski, B.-F., Chau, G., Christopoulos, V., Liu, C. Y., Tanter, M., Shapiro, M. G., et al. (2023) · Nature Neuroscience · doi:10.1038/s41593-023-01500-7 · Cited by 70 · see citing works
Kosnoff, J., Yu, K., Liu, C., & He, B. (2024) · Nature Communications · doi:10.1038/s41467-024-48576-8 · Cited by 58 · see citing works
“Invasive BCIs have demonstrated the ability to control a robotic arm 27 – 30 , decode speech 29,31,32 , and interpret handwriting 33 .”
Cortico-basal ganglia plasticity in motor learning
Roth, R. H., & Ding, J. (2024) · Neuron · doi:10.1016/j.neuron.2024.06.014 · Cited by 66 · see citing works
“146,147 Overall, systems and behavioral studies of motor control and motor learning in human and non-human primates have yielded a rich and high-level understanding of the population-level encoding of movement parameters that can be exploited to control BCIs in patients with lost movement or speech…”
Vansteensel, M. J., Klein, E. P., van Thiel, G. J. M. W., Gaytant, M. A., Simmons, Z., Wolpaw, J. R., & Vaughan, T. M. (2022) · Journal of Neurology · doi:10.1007/s00415-022-11464-6 · Cited by 61 · see citing works
Ma, H., Liang, J., Qiu, J. R., Li, J., Ma, L., Sheng, H., Shao, M., Wang, Q., Li, F., Fu, Y., et al. (2023) · Advanced Materials · doi:10.1002/adma.202301218 · Cited by 61 · see citing works
Advocating for neurodata privacy and neurotechnology regulation
Yuste, R. (2023) · Nature Protocols · doi:10.1038/s41596-023-00873-0 · Cited by 60 · see citing works
Shared Control of Bimanual Robotic Limbs With a Brain-Machine Interface for Self-Feeding
Handelman, D. A., Osborn, L. E., Thomas, T. M., Badger, A. R., Thompson, M. C., Nickl, R. W., Anaya, M. A., Wormley, J. M., Cantarero, G. L., McMullen, D. P., et al. (2022) · Frontiers in Neurorobotics · doi:10.3389/fnbot.2022.918001 · Cited by 58 · see citing works
“Cortical BMIs can be used to restore function by decoding neural signals for a variety of applications including handwriting (Willett et al., 2021), restoring speech (Moses et al.”
BackgroundThe unbearable slowness of being: Why do we live at 10 bits/s?
Zheng, J., & Meister, M. (2024) · Neuron · doi:10.1016/j.neuron.2024.11.008 · Cited by 64 · see citing works
“These systems have demonstrated some interesting capabilities, for example, a recent benchmark-setting BCI can decode intended handwriting at 90 English characters per minute, or 1.5 bits/s [56].”
InfluentialZhang, J., Li, J., Huang, Z., Huang, D., Yu, H., & Li, Z. (2023) · Health Data Science · doi:10.34133/hds.0096 · Cited by 58 · see citing works
Artificial Intelligence for Metaverse: A Framework
Guo, Y., Yu, T., Wu, J., Wang, Y., Wan, S., Zheng, J., Fang, L., & Dai, Q. (2022) · CAAI Artificial Intelligence Research · doi:10.26599/air.2022.9150004 · Cited by 60 · see citing works
The Effect of Time Window Length on EEG-Based Emotion Recognition
Ouyang, D., Yuan, Y., Li, G., & Guo, Z. (2022) · Sensors · doi:10.3390/s22134939 · Cited by 53 · see citing works
“In addition, the characteristics of EEG signals related to some specific events will change over time, making EEG-based recognition more difficult and requiring calibration before application [27].”
BackgroundA Complete Scheme for Multi-Character Classification Using EEG Signals From Speech Imagery
Pan, H., Wang, Y., Li, Z., Chu, X., Teng, B., & Gao, H. (2024) · IEEE Transactions on Biomedical Engineering · doi:10.1109/tbme.2024.3376603 · Cited by 55 · see citing works
“Brain– computer interface (BCI) technology has the potential to restore communication abilities for individuals who have lost their ability to speak [2].”
BackgroundPelkonen, A., Pistono, C., Klecki, P., Gómez‐Budia, M., Dougalis, A., Konttinen, H., Stanová, I., Fagerlund, I., Leinonen, V., Korhonen, P., et al. (2021) · Cells · doi:10.3390/cells11010106 · Cited by 50 · see citing works
“This may skew our knowledge of brains unaffected by seizures, even though some data also originate from BMIs implanted in spinal cord injury patients [11,56].”
BackgroundAdvanced Electrode Technologies for Noninvasive Brain–Computer Interfaces
Lin, S., Jiang, J., Huang, K., Li, L., He, X., Du, P. Y., Wu, Y., Liu, J., Li, X., Huang, Z., et al. (2023) · ACS Nano · doi:10.1021/acsnano.3c06781 · Cited by 52 · see citing works
Translation of neurotechnologies
Schalk, G., Brunner, P., Allison, B. Z., Soekadar, S. R., Guan, C., Denison, T., Rickert, J., & Miller, K. J. (2024) · Nature Reviews Bioengineering · doi:10.1038/s44222-024-00185-2 · Cited by 62 · see citing works
Formento, E., Botros, P., & Carmena, J. M. (2021) · Journal of Neural Engineering · doi:10.1088/1741-2552/ac35ac · Cited by 47 · see citing works
Decoding Depression Severity From Intracranial Neural Activity
Xiao, J., Provenza, N. R., Asfouri, J. J., Myers, J. C., Mathura, R. K., Metzger, B. A., Adkinson, J. A., Allawala, A. B., Pirtle, V., Oswalt, D. N., et al. (2023) · Biological Psychiatry · doi:10.1016/j.biopsych.2023.01.020 · Cited by 52 · see citing works
Nonlinear manifolds underlie neural population activity during behaviourPreprint
Fortunato, C., Bennasar-Vázquez, J., Park, J., Chang, J. C., Miller, L. E., Dudman, J. T., Perich, M. G., & Gallego, J. Á. (2023) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2023.07.18.549575 · Cited by 51 · see citing works
Interpretable Passive Multi-Modal Sensor Fusion for Human Identification and Activity Recognition
Yuan, L., Andrews, J., Mu, H., Vakil, A., Ewing, R. L., Blasch, E., & Li, J. (2022) · Sensors · doi:10.3390/s22155787 · Cited by 54 · see citing works
“RNNs have shown success in a variety of additional human recognition applications such as gesture recognition [26], gait analysis [27], and translation [28].”
BackgroundNeuron devices: emerging prospects in neural interfaces and recognition
Wang, Y., Liu, S., Wang, H., Zhao, Y., & Zhang, X. (2022) · Microsystems & Nanoengineering · doi:10.1038/s41378-022-00453-4 · Cited by 52 · see citing works
Zhang, S., Suresh, L., Yang, J., Zhang, X., & Tan, S. C. (2022) · Advanced Intelligent Systems · doi:10.1002/aisy.202100194 · Cited by 55 · see citing works
Inner speech in motor cortex and implications for speech neuroprostheses
Kunz, E. M., Abramovich Krasa, B., Kamdar, F., Avansino, D. T., Hahn, N., Yoon, S., Singh, A., Nason-Tomaszewski, S. R., Card, N. S., Jude, J. J., et al. (2025) · Cell · doi:10.1016/j.cell.2025.06.015 · Cited by 55 · see citing works
“Such an approach could be preferred by some BCI users with partial paralysis since attempted speech requires activation of orofacial muscles and breath control that can be fatiguing and produces distracting vocalizations.”
Self-citationTen-Hour Stable Noninvasive Brain-Computer Interface Realized by Semidry Hydrogel-Based Electrodes
Liu, J., Lin, S., Li, W., Zhao, Y., Liu, D., He, Z., Wang, D., Lei, M., Hong, B., & Wu, H. (2022) · Research · doi:10.34133/2022/9830457 · Cited by 48 · see citing works
Soroush, P. Z., Herff, C., Ries, S. K., Shih, J. J., Schultz, T., & Krusienski, D. J. (2023) · NeuroImage · doi:10.1016/j.neuroimage.2023.119913 · Cited by 48 · see citing works
Bergeron, D., Iorio‐Morin, C., Bonizzato, M., Lajoie, G., Gaucher, N. O., Racine, É., & Weil, A. G. (2023) · Journal of Child Neurology · doi:10.1177/08830738231167736 · Cited by 48 · see citing works
“To improve the typing rate of language braincomputer interface, many researchers have restricted the number of possible words and relied on the use of some kind of “predictions” or “autocorrect” based on syntactic structure.(28,34) This may restrain the way patients express themselves, with limited access to nuanced,…”
MethodBrain–computer interface digital prescription for neurological disorders
Chai, X., Cao, T., He, Q., Wang, N., Zhang, X., Shan, X., Lv, Z., Tu, W., Yang, Y., & Zhao, J. (2024) · CNS Neuroscience & Therapeutics · doi:10.1111/cns.14615 · Cited by 48 · see citing works
Electrocorticogram (ECoG): Engineering Approaches and Clinical Challenges for Translational Medicine
Moon, H., Kwon, J., Eun, J., Chung, C. K., Kim, J. S., Chou, N., & Kim, S. (2024) · Advanced Materials Technologies · doi:10.1002/admt.202301692 · Cited by 51 · see citing works
Using adversarial networks to extend brain computer interface decoding accuracy over time
Ma, X., Rizzoglio, F., Bodkin, K. L., Perreault, E. J., Miller, L. E., & Kennedy, A. (2023) · eLife · doi:10.7554/elife.84296 · Cited by 47 · see citing works
An end-to-end CNN with attentional mechanism applied to raw EEG in a BCI classification task
Lashgari, E., Ott, J., Connelly, A., Baldi, P., & Maoz, U. (2021) · Journal of Neural Engineering · doi:10.1088/1741-2552/ac1ade · Cited by 45 · see citing works
Mechanoluminescent optical fiber sensors for human–computer interaction
Ren, B., Chen, B., Zhang, X., Wu, H., Fu, Y., & Peng, D. (2023) · 中国科学通报：英文版 · doi:10.1016/j.scib.2023.02.030 · Cited by 46 · see citing works
Integrated Memristor Network for Physiological Signal Processing
Cai, L., Yu, L., Yue, W., Zhu, Y., Yang, Z., Li, Y., Tao, Y., & Yang, Y. (2023) · Advanced Electronic Materials · doi:10.1002/aelm.202300021 · Cited by 46 · see citing works
- Authors
- Willett, F. R., Avansino, D. T., Hochberg, L. R., Henderson, J. M., & Shenoy, K. V.
- Affiliations
- Howard Hughes Medical Institute (US)
- Harvard University (US)
- Stanford Medicine (US)
- Corresponding author
- Willett, Francis R.
- Venue
- Nature
- Published
- 2021-05-12
- Volume
- 593
- Issue
- 7858
- Pages
- 249-254
- ISSN
- 0028-0836 (ISSN-L)
- 1476-4687
- PMC
- PMC8163299
- OpenAlex
- W3161343553
- Semantic Scholar
- bebaab40b42e63cd7cef7b334e5f6bacdcccfc79
- Cited by
- 969
- References
- 50
- Citations by year
- 2026: 143 · 2025: 214 · 2024: 203 · 2023: 201 · 2022: 134 · 2021: 68 · 2020: 4 · 2019: 2
- FWCI
- 72.98
- Type
- journal article
- Language
- en
- Open access
- green
- doi.org
- Licence
- public-domain
- www.springer.com/tdm (tdm)
- www.springer.com/tdm (vor)
- Locations
- Nature (publishedVersion)
- bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, public-domain)
- PubMed (publishedVersion)
- Europe PMC (PubMed Central) (submittedVersion)
- PubMed Central (submittedVersion)
- RePEc: Research Papers in Economics (submittedVersion)
- Funders
- Howard Hughes Medical Institute
- U.S. Department of Veterans Affairs (I01 RX002295, I50RX002864, B6453-R, N-2864-C)
- Office of Research and Development (N2864C, B6453R)
- National Institutes of Health (U01NS098968, R01DC014034)
- National Institute on Deafness and Other Communication Disorders (R01DC009899, R01-DC014034, DC017844, DC014034, U01- DC017844)
- National Institute of Neurological Disorders and Stroke (U01NS098968, UH2-NS095548, UO1-NS098968, NS098968, NS095548)
- Topics
- Life Sciences › Neuroscience › Cognitive Neuroscience › EEG and Brain-Computer Interfaces
- Life Sciences › Neuroscience › Cellular and Molecular Neuroscience › Neuroscience and Neural Engineering
- Physical Sciences › Computer Science › Human-Computer Interaction › Gaze Tracking and Assistive Technology
- Keywords
- brain-computer interface, handwriting, text entry, motor cortex, intracortical recording, spinal cord injury, paralysis, real-time communication, typing speed
- MeSH terms
- Brain, Communication, Handwriting, Humans, Spinal Cord Injuries, Time Factors, Neural Networks, Computer, Brain-Computer Interfaces
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-09-28.
Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-09-28.
Cited sources are checked by gpt-6-sol.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
