REVIEW 3 major objections 5 minor 18 references
Early Assessment of Artificial Lower Extremity Sensory Response Times and Proprioceptive Acuity via Sensory Cortex Electrical Stimulation
T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash
Pith's one-line read Direct brain-surface stimulation can encode leg swing speed into an artificial sensation, but perception lags by about one second.
desk verdict First objective test of lower-limb S1 DCES proprioceptive acuity, but the rate-encoding claim is confounded by charge differences and the statistics overreach the sample. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The central mechanism is rate encoding of leg swing velocity into DCES pulse trains. Three synthetic bell-shaped velocity profiles (fast, medium, slow) are passed through an Izhikevich spiking-neuron model to produce stimulation trains with maximum rates near $417$, $213$, and $141$ Hz, and these trains drive a bipolar S1 electrode. The proprioceptive acuity task asks whether the second of two percepts feels faster, slower, or the same, and the sensory response task compares leg DCES against hand DCES, visual, and auditory stimuli. Because the intended information-carrying variable is the temporal rate pattern, the observation that discrimination accuracy scales with the maximum-frequency difference is what ties the results to rate encoding.
What would settle it
A matched-charge discrimination experiment would settle the issue: construct fast, medium, and slow trains with identical total pulse count and total charge, differing only in temporal pulse pattern, and if accuracy falls to chance, the rate-encoding explanation is falsified. Recruiting additional subjects and showing that fast/slow discrimination replicates above chance would also directly test the feasibility claim.
Extended reading notes
Core claim
The authors' central claim is that rate-modulated direct cortical electrical stimulation (DCES) of the primary sensory cortex can carry leg swing velocity-like information: a single subject discriminated fast/slow, fast/medium, medium/slow, and same-speed artificial leg percepts with 80%, 70%, 60%, and 53% accuracy, respectively, against a 50% chance level (empirical $p = 1.9\times10^{-5}$), and accuracy rose with the size of the maximum-frequency difference. They further found that perceiving an artificial leg sensation took $1007 \pm 413$ ms, significantly longer than visual ($528 \pm 137$ ms) or auditory ($393 \pm 106$ ms) cues to the same location and longer than DCES of the hand ($599 \pm 171$ ms). The authors take these results as early feasibility evidence that S1 DCES can provide proprioceptive feedback for a bidirectional brain-computer interface, while acknowledging that the delay must be reduced for timely gait responses.
Load-bearing premise
The central claim stands or falls on whether the single subject's above-chance discrimination was caused by the rate-coded velocity difference rather than by incidental differences such as total pulse count, charge, or perceived intensity.
Editorial extensions
If this is right
- If the single-subject result generalizes, rate-modulated S1 stimulation can deliver graded proprioceptive feedback about leg swing speed without intact peripheral sensation, giving people with spinal cord injury a route to closed-loop gait feedback.
- The roughly one-second artificial leg percept latency means a walking brain-computer interface cannot rely on the current stimulation parameters for reactive balance corrections; either faster stimulation protocols or predictive movement strategies will be needed.
- Discrimination accuracy rose with maximum-frequency difference (80% for fast/slow versus 60% for medium/slow), so coarse velocity gradations are the realistic near-term target rather than fine proprioceptive resolution.
- The authors' stated next steps—sensory training and pulse trains derived from real recorded leg trajectories—follow directly from their interpretation, since both could plausibly reduce discrimination errors and response times.
Reading between the lines
- The paper's synthetic velocity profiles and hand-picked rate mapping leave open whether the subject solved the task by rate or by incidental cues such as total pulse count or charge; a matched-charge control with identical pulse counts but different temporal patterns would isolate the rate cue.
- If the rate cue is what matters, the same discrimination should survive rescaling all stimulation rates by a common factor; if performance instead tracks total charge, the proprioceptive interpretation would need revision.
- The ~1 s delay, if it originates from cortical stimulation dynamics rather than medication, implies that any S1-based feedback loop carries an intrinsic latency floor; comparing the same patients on and off sedating medications would separate the two contributions.
- Feeding actual motion-captured leg trajectories through the same Izhikevich pipeline would test whether the bell-shaped synthetic profiles limited acuity; improved discrimination with real trajectories would strengthen the proprioceptive claim.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper reports an early feasibility study with three epilepsy surgery patients implanted with subdural ECoG electrodes over S1. In a proprioceptive acuity task, one subject (SJ1) received pairs of DCES pulse trains generated by rate-encoding synthetic leg-swing velocity profiles (fast, medium, slow) and reported whether the second stimulus was faster, slower, or the same as the first; overall accuracy across 45 trials was 64%, with an empirical p-value of 1.9×10^-5 against random guessing. In a sensory response time task, three subjects responded to DCES-induced leg/hand sensations and to visual and auditory controls; mean response times were substantially longer for DCES-leg (1007±413 ms) than for visual-leg (528±137 ms) and auditory-leg (393±106 ms). The authors conclude that proprioceptive information can be artificially conveyed via S1 DCES, but that the perception delay is significant for real-time BD-BCI operation.
Significance. The response-time measurements are among the first quantitative data on lower-extremity DCES latency and address a genuine gap in BD-BCI sensory feedback. The use of an empirical permutation test against guessing and the careful reporting of clinical stimulation mapping are strengths. If the proprioceptive-encoding claim were uniquely supported, the paper would be a valuable early contribution. However, the central claim that rate-encoded leg swing velocity is conveyed is currently undermined by a charge/rate confound and by reliance on a single subject; the response-time comparisons also contain a statistical level-of-analysis error. The significance is therefore preliminary: the results merit publication only as a pilot with substantially more cautious claims or after additional controls.
major comments (3)
- [III-C, II-F, Tables III and IV] The proprioceptive acuity task cannot distinguish rate encoding from total-charge or intensity discrimination. The three bell-shaped velocity profiles have similar durations, and all pulses use fixed current (6.2 mA) and pulse width (250 µs/phase), so the total number of pulses (and thus injected charge) is proportional to the integrated rate. The reported peak rates (fast ~417 Hz, medium ~213 Hz, slow ~141 Hz) thus generate exactly the same ordering in total charge, and the accuracy ordering (80%, 70%, 60%) tracks that ordering. The empirical p=1.9×10^-5 test only rejects uniform guessing; it does not test whether the subject used perceived speed rather than perceived intensity/tingling strength. Consequently, the Discussion's inference that 'rate encoding may be a viable strategy' is not uniquely supported. Please add a charge-matched control condition (or an intensity-rating task) or substantially soften the claim.
- [III-D, Fig. 3] The response-time comparisons pool individual trials across subjects and treat them as independent in t-tests, while the number of subjects contributing to each condition is only 1–3. Specifically, auditory-leg data come from a single subject (SJ3), DCES-leg from SJ1 and SJ3, and DCES-hand from SJ1 and SJ2. With n=1–2 subjects per condition, pairwise trial-level t-tests do not support population-level inference and the reported p-values are inflated by pseudoreplication. The conclusion that artificial leg perception is 'significantly delayed' relative to visual and auditory conditions should be based on subject-level statistics (e.g., mixed-effects models or paired subject means) or presented as descriptive pilot findings without inferential claims.
- [III-C, Table IV] The overall empirical p-value is driven primarily by the two coarsest contrasts. Per-condition performance is 6/10 for medium/slow (binomial one-sided p≈0.08) and 8/15 for the 'same' condition (p≈0.09), neither of which is clearly above chance at conventional levels. Thus the data do not demonstrate that the subject could discriminate all three velocity levels; the significant result rests on the fast/slow and fast/medium contrasts, which also carry the largest charge differences. Please report per-condition significance tests and explicitly limit the conclusion to contrasts with adequate evidence.
minor comments (5)
- [II-F] The sentence 'There is an 50%, 200%, and 100% difference in maximum velocity between medium and fast, slow and fast, and slow and medium speeds, respectively' is inconsistent with the values in Section III-C (~196%, ~96%, and ~51% differences in maximum stimulation rate); please reconcile the percentages and state the reference value used for each comparison.
- [II-G, Fig. 3] In the sentence 'All pairwise comparisons (t-test) were significant anα threshold of 0.01', the word 'at' is missing; also, use 'an' before 'α' consistently (e.g., 'at a significance threshold of 0.01').
- [Fig. 2 caption] The caption reads 'Representations of slow leg swing velocity time series were translated in stimulation pulse trains'; the word 'in' should be 'into'.
- [III-C] The phrase 'There was ∼196%, ∼96%, and ∼51% difference between maximum stimulation rate between slow and fast, medium and fast, and slow and medium speeds, respectively' is confusing; please list the pairs in a consistent order (e.g., fast/slow, fast/medium, medium/slow).
- [II-F, Table IV] The text says 'For each 4 stimulus combination, at least 10 trials were administered', but Table IV shows 15 trials for the 'Same' condition only; please clarify how the 15 'Same' trials were allocated (e.g., fast-fast, medium-medium, slow-slow) and whether the counts are balanced.
Circularity Check
No significant circularity; the acuity result is tested against random-guess trials and response times are measured directly.
full rationale
The paper's central claims are empirical measurements, not derived quantities. The proprioceptive acuity p-value (p = 1.9e-05) is computed by comparing the subject's accuracy to 10^6 random-guess trials, an external benchmark, and the response-time comparisons are direct t-tests across modalities. The rate-encoding scheme is an input design choice taken from prior literature (Izhikevich model and cortical rate coding), not a fitted output, and the authors explicitly acknowledge in the Discussion that the synthetic velocity profiles and small sample size are limitations rather than assumed results. Self-citations to the group's prior hardware validation [5] and earlier BD-BCI studies [2], [3] support the experimental apparatus and context, but the target findings (discrimination accuracy and response times) do not reduce to those citations; the cited device validation is independent benchtop and bedside work. The unresolved charge and intensity confound in the acuity task is a validity risk that would weaken the proprioceptive interpretation, but it is not a circular reduction: the subject's responses are not constructed from the stimulus parameters. Therefore no circular step is present.
Assumptions & free parameters
free parameters (1)
- Velocity-to-rate encoding scale =
fast ~417 Hz, medium ~213 Hz, slow ~141 Hz maximum stimulation frequency
assumptions (4)
- domain assumption Rate encoding in S1 conveys perceived limb velocity
- domain assumption The Izhikevich neuron model gives a perceptually valid mapping from velocity profiles to stimulation trains
- domain assumption Clinically mapped electrode channels correctly localize S1 leg and hand representations
- standard math Random-guess simulation models chance performance on the 3-alternative task
Cite this review
Pith. "Pith review of Early Assessment of Artificial Lower Extremity Sensory Response Times and Proprioceptive Acuity via Sensory Cortex Electrical Stimulation." pith.science (2026). https://pith.science/paper/OGA64EYD
@misc{pith2026250522691,
author = {Pith},
title = {Pith review of: Early Assessment of Artificial Lower Extremity Sensory Response Times and Proprioceptive Acuity via Sensory Cortex Electrical Stimulation},
year = {2026},
howpublished = {\url{https://pith.science/paper/OGA64EYD}},
note = {Machine review of arXiv:2505.22691}
}
abstract
Bi-directional brain computer interfaces (BD-BCIs) may restore brain-controlled walking and artificial leg sensation after spinal cord injury. Current BD-BCIs provide only simplistic "tingling" feedback, which lacks proprioceptive information to perceive critical gait events (leg swing, double support). This information must also be perceived adequately fast to facilitate timely motor responses. Here, we investigated utilizing primary sensory cortex (S1) direct cortical electrical stimulation (DCES) to deliver leg proprioceptive information and measured response times to artificial leg sensations. Subjects with subdural electrocorticogram electrodes over S1 leg areas participated in two tasks: (1) Proprioceptive acuity: subjects identified the difference between DCES-induced percepts emulating various leg swing speeds; (2) Sensory response: measuring subjects' reaction time to DCES-induced leg sensations, with DCES-hand, visual and auditory control conditions. Three subjects were recruited. Only one completed the proprioceptive assessment, achieving 80%, 70%, 60%, and 53% accuracy in discriminating between fast/slow, fast/medium, medium/slow, and same speeds, respectively (p-value=1.9x10$^{-5}$). Response times for leg/hand percepts were 1007$\pm$413/599$\pm$171 ms, visual leg/hand responses were 528$\pm$137/384$\pm$84 ms, and auditory leg/hand responses were 393$\pm$106/352$\pm$93 ms, respectively. These results suggest proprioceptive information can be delivered artificially, but perception may be significantly delayed. Future work should address improving acuity, reducing response times, and expanding sensory modalities.
Figures
Reference graph
Works this paper leans on
-
[1]
A brain-computer interface that evokes tactile sensations improves robotic arm control,
S. N. Flesher, J. E. Downey, J. M. Weiss, C. L. Hughes, A. J. Herrera, E. C. Tyler-Kabara, M. L. Boninger, J. L. Collinger, and R. A. Gaunt, “A brain-computer interface that evokes tactile sensations improves robotic arm control,”Science, vol. 372, no. 6544, pp. 831–836, 2021
work page 2021
-
[2]
Early feasibility of an embedded bi-directional brain-computer interface for ambulation,
J. Lim, P. T. Wang, W. J. Sohn, C. Serrano-Amenos, M. Ibrahim, D. Lin, S. Thaploo, S. J. Shaw, M. Armacost, H. Gong,et al., “Early feasibility of an embedded bi-directional brain-computer interface for ambulation,” in2024 46th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), pp. 1–5, IEEE, 2024
work page 2024
-
[3]
Real-time brain- computer interface control of walking exoskeleton with bilateral sensory feedback,
J. Lim, P. T. Wang, W. J. Sohn, D. Lin, S. Thaploo, L. Bashford, D. Bjanes, A. Nguyen, H. Gong, M. Armacost,et al., “Real-time brain- computer interface control of walking exoskeleton with bilateral sensory feedback,”arXiv preprint arXiv:2505.00219, 2025
arXiv 2025
-
[4]
D. J. Caldwell, J. A. Cronin, J. Wu, K. E. Weaver, A. L. Ko, R. P. Rao, and J. G. Ojemann, “Direct stimulation of somatosensory cortex results in slower reaction times compared to peripheral touch in humans,” Scientific reports, vol. 9, no. 1, p. 3292, 2019
work page 2019
-
[5]
W. J. Sohn, J. Lim, P. T. Wang, H. Pu, O. Malekzadeh-Arasteh, S. J. Shaw, M. Armacost, H. Gong, S. Kellis, R. A. Andersen, and others, “Benchtop and bedside validation of a low-cost programmable cortical stimulator in a testbed for bi-directional brain-computer-interface re- search,”Frontiers in Neuroscience, vol. 16, 2022. Publisher: Frontiers Media SA
work page 2022
-
[6]
Velocity perception and propriocep- tion,
G. K. Kerr and C. J. Worringham, “Velocity perception and propriocep- tion,”Sensorimotor control of movement and posture, pp. 79–86, 2002
work page 2002
-
[7]
The measurement of proprioceptive accuracy: A systematic literature review,
´A. Horv ´ath, E. Ferentzi, K. Schwartz, N. Jacobs, P. Meyns, and F. K¨oteles, “The measurement of proprioceptive accuracy: A systematic literature review,”Journal of sport and health science, vol. 12, no. 2, pp. 219–225, 2023
work page 2023
-
[8]
Simple model of spiking neurons,
E. M. Izhikevich, “Simple model of spiking neurons,”IEEE Transactions on neural networks, vol. 14, no. 6, pp. 1569–1572, 2003
work page 2003
Show all 18 references
-
[9]
Emergence of an invariant repre- sentation of texture in primate somatosensory cortex,
J. D. Lieber and S. J. Bensmaia, “Emergence of an invariant repre- sentation of texture in primate somatosensory cortex,”Cerebral Cortex, vol. 30, no. 5, pp. 3228–3239, 2020
2020
-
[10]
Tactile spatial resolution. i. two-point discrimination, gap detection, grating resolution, and letter recognition,
K. O. Johnson and J. R. Phillips, “Tactile spatial resolution. i. two-point discrimination, gap detection, grating resolution, and letter recognition,” Journal of neurophysiology, vol. 46, no. 6, pp. 1177–1192, 1981
1981
-
[11]
Detailed analysis of proprioception in normal and acl-deficient knees,
G. Pap, A. Machner, W. Nebelung, and F. Awiszus, “Detailed analysis of proprioception in normal and acl-deficient knees,”The Journal of Bone & Joint Surgery British Volume, vol. 81, no. 5, pp. 764–768, 1999
1999
-
[12]
The effects of age and amplitude on wrist proprioceptive acuity,
F. Marini, C. M. Hughes, P. Morasso, and L. Masia, “The effects of age and amplitude on wrist proprioceptive acuity,” in2017 International Conference on Rehabilitation Robotics (ICORR), pp. 609–614, IEEE, 2017
2017
-
[13]
Neuronal correlates of tactile speed in primary somatosensory cortex,
A. D ´epeault, E.-M. Meftah, and C. E. Chapman, “Neuronal correlates of tactile speed in primary somatosensory cortex,”Journal of neurophysi- ology, vol. 110, no. 7, pp. 1554–1566, 2013
2013
-
[14]
Human perception of electrical stimulation on the surface of somatosensory cortex,
S. V . Hiremath, E. C. Tyler-Kabara, J. J. Wheeler, D. W. Moran, R. A. Gaunt, J. L. Collinger, S. T. Foldes, D. J. Weber, W. Chen, and e. a. Boninger, M L, “Human perception of electrical stimulation on the surface of somatosensory cortex,”PloS One, vol. 12, no. 5, p. e0176020,
-
[15]
Re- sponse time to a vibrotactile stimulus presented on the foot at rest and during walking on different surfaces,
L. D. Chapwouo Tchakout ´e, L. Tremblay, and B.-A. J. Menelas, “Re- sponse time to a vibrotactile stimulus presented on the foot at rest and during walking on different surfaces,”Sensors, vol. 18, no. 7, p. 2088, 2018
2018
-
[16]
The effect of sensory discrimination training on sensori- motor performance in individuals with central neurological conditions: A systematic review,
S. C. Taylor, C. McKenna, J. Kent, M.-A. Jess, J. R. Robinson, J. Dixon, and C. Ryan, “The effect of sensory discrimination training on sensori- motor performance in individuals with central neurological conditions: A systematic review,”British Journal of Occupational Therapy,...
2021
-
[17]
Somatic motor and sensory representation in the cerebral cortex of man as studied by electrical stimulation,
W. Penfield and E. Boldrey, “Somatic motor and sensory representation in the cerebral cortex of man as studied by electrical stimulation,”Brain, vol. 60, no. 4, pp. 389–443, 1937
1937
-
[2017]
Publisher: Public Library of Science
Reviewed August 7, 2026 · model on record in the stance chip above.
Discussion (0). Sign in to comment.