Towards model-based design of causal manipulations of brain circuits with high spatiotemporal precision

Anandita De; Luca Mazzucato; Roozbeh Kiani

REVIEW 4 major objections 5 minor 36 references

Towards model-based design of causal manipulations of brain circuits with high spatiotemporal precision

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A two-step model-based workflow can design brain-stimulation patterns: find hubs from spontaneous activity, then compute minimum-energy multi-site inputs from a linearized delay-space model.

desk verdict Clear roadmap, unsupported feasibility claim: the RNN demonstration lacks the equations, parameters, and error metrics needed to back the paper's central claim. read the letter →

arxiv 2505.24790 v1 pith:MNKTMOQH submitted 2025-05-30 q-bio.NC

Anandita De , Roozbeh Kiani , Luca Mazzucato This is my paper

classification q-bio.NC

keywords neuralstimulationdesignhubidentificationCausalFlowKoopmanoperatordynamicmodedecompositionwithcontroldelayembeddingminimum-energyrecurrentnetworks

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper claims that causal manipulation of neural circuits can be planned computationally instead of by trial and error. The proposal is a two-step workflow: identify high-efficacy hub sites from short recordings of spontaneous activity using delay-embedding Causal Flow, then design multi-site stimulation inputs by treating the nonlinear circuit as an approximately linear input-driven system in a delay space and solving a minimum-energy control problem. The authors demonstrate the workflow in recurrent network models, where the stimulation computed from the fitted linear model drives the network close to a desired final state. If the approach carries over to real circuits, it would let experimenters and clinicians design stimulation protocols from resting-state data rather than screening electrodes one by one.

What carries the argument

The machinery is delay-space linearization coupled with causal hub selection. A Hankel matrix stacks time-delayed observations, and the Koopman operator is approximated by the matrix $A$ that maps each column to its neighbor; with inputs present, the pair $(A,B)$ is fit by dynamic mode decomposition with control, yielding the linear controlled system $z(t+1)=Az(t)+Bu(t)$. Causal Flow—based on the asymmetry of cross-embedding reconstructability—identifies which nodes are hubs, so the control input is restricted to high-efficacy sites. The linear form makes the minimum-energy control problem tractable in the delay space, and the final activity is read out through $x=Cz$.

What would settle it

In the recurrent network models, compute the final-state error $\|x_f - \hat{x}_f\|$ while increasing the control horizon $T$ beyond the prediction horizon at which the fitted model's forecasts degrade; if the error grows without bound, or if refitting $(A,B)$ on a different spontaneous-activity segment changes the evoked final state, the central feasibility claim would be refuted.

Watch

Extended reading notes

Core claim

The paper's central claim is that the problem of where and how to stimulate a neural circuit can be solved as a model-based control problem rather than by trial and error. After delay-embedding recorded activity, the nonlinear network is treated as an approximately linear, input-driven system $z(t+1)=Az(t)+Bu(t)$, with $A$ and $B$ fitted from a short training period in which currents are injected into hubs previously identified from spontaneous activity via Causal Flow. The paper then uses linear minimum-energy control to compute a multi-site input that steers the delay-space trajectory from $z_0$ to a target $z_f$, and shows in recurrent network models that applying that input to the true nonlinear system evokes a final state close to the desired target. The claim is that both steps—hub selection and stimulation design—are feasible from data alone.

Load-bearing premise

The whole method depends on the assumption that the simple linear model fitted from a short training period remains accurate enough over the control period that the stimulation computed from it actually guides the real nonlinear circuit to the target, even though the model's predictions drift after a few steps.

Editorial extensions

If this is right

Stimulation targets can be chosen from spontaneous recordings alone, replacing exhaustive single-electrode trial-and-error screening with a data-driven ranking.
Multi-site stimulation protocols can be optimized for minimum total injected current, directly addressing energy and tissue-safety constraints.
A single fitted linear model in delay space can be reused to design inputs for many different target states, without retraining on each target.
The tradeoff between prediction horizon and control horizon becomes an explicit design parameter, making the reliability of stimulation protocols quantifiable.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A natural extension, not explored in the paper, is closed-loop operation: re-fit $A$ and $B$ online as new observations arrive, which could compensate for the short prediction horizon and let the controller reach targets over longer timescales.
Because the hub-identification step relies only on spontaneous activity, the workflow may scale to multi-region or whole-brain recordings, although the paper demonstrates it only on single recurrent network models.
The paper's observation that predicted and evoked trajectories diverge en route yet reconverge at the target hints that the minimum-energy controller has an implicit robustness margin; quantifying that margin as a function of model error would turn the feasibility demonstration into a design principle.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The thing to know about arXiv:2505.24790 is that it's a clearly written perspective proposing a two-step workflow — Causal Flow hub identification from spontaneous activity, then Koopman/DMDc-based minimum-energy control — and it claims an RNN demonstration that is effectively absent from the page. The assembled pipeline is genuinely new as a package, and the authors are honest about how far this is from real spiking data. But the central feasibility claim rests on a Figure 4 with no network equations, no parameters, no error metrics, and no code. The text says the evoked xf closely approximates the target, and then says predicted and evoked trajectories diverge beyond short horizons but "converge at the desired target state." That convergence is exactly the load-bearing fact, and it is asserted rather than supported.

What the paper does well: it gives a clear high-level account of why trial-and-error stimulation design is a bottleneck, why delay-embedding might help, and how the components fit together. The theory it invokes — Takens embedding, Koopman linearization, DMD with control — is described correctly, and the citation pattern is honest, including explicit statements about where the method has not yet been applied. For a reader wanting a map of this research program, it's a useful resource.

The soft spots are real but fixable. The missing simulation details are the big one: without equations or parameters, nobody can check whether the control works in general or only in a favorable regime. The paper's own admission of divergence beyond short horizons makes the claimed convergence at target a surprise that needs justification; a short control horizon could explain it, but the horizon isn't given. There's also no comparison with non-hub or random stimulation, so we don't know if hub choice matters in the RNN, which is central to the first step. These omissions prevent the paper from supporting its headline claim. If it's meant as a perspective, the tone should be softened; if it's meant as a demonstration, the evidence has to be provided.

I'd send this to serious reviewers if the authors are willing to substantiate the simulation or reframe the claim as an illustration of the workflow. As is, I would not cite it as a feasibility demonstration. Recommendation: deserves peer review only with the expectation of major revision; otherwise it reads as a useful review, not a research result.

Referee Report

4 major / 5 minor

Summary. The paper proposes a two-step workflow for designing multi-site brain stimulation: first identify high-efficacy hub sites from spontaneous activity using Causal Flow, then design minimum-energy stimulation sequences using a delay-embedding/Koopman linear model with control (DMDc), i.e., z(t+1) = A z(t) + B u(t). The authors argue that nonlinear neural dynamics become approximately linear in a high-dimensional delay space, making linear control theory applicable. They claim to demonstrate feasibility using recurrent network models, with hub identification via Causal Flow and control simulations shown schematically in Figures 3 and 4. The manuscript also candidly states that the DMDc/Koopman pipeline has not yet been applied to spiking neural data or brain-computer interfaces.

Significance. If the central feasibility claim were quantitatively established, the framework would be a valuable conceptual contribution: it combines Causal Flow hub detection, Koopman linearization, and linear minimum-energy control into a practical recipe for designing multi-site stimulation protocols. The paper correctly describes standard tools (Takens embedding, Koopman/DMD, DMDc) and cites relevant literature. However, the only direct evidence for the central claim is an unreported simulation: no network equations, parameters, error metrics, code, or quantitative convergence results are provided, and the one quantitative figure (Fig. 3d) shows that prediction accuracy degrades rapidly with horizon. The contribution is therefore conditional: the proposed workflow is plausible, but the demonstration of feasibility is asserted rather than established in the submitted text.

major comments (4)

[Section 5, Figure 4] The central feasibility claim, 'We find that the evoked xf closely approximates the desired target xf', is not supported by any quantitative evidence in the manuscript. The recurrent network used for the control simulations is never specified: there are no network equations, no parameter values, no number of neurons, no details of the delay embedding, and no description of how the training inputs were chosen. The paper also refers to 'Figure 4d' in the text, but the figure caption lists only panels a-c, and the missing panel appears to be where the convergence evidence would be shown. At minimum, the authors must provide the full model and training details, the final-state error ||x_evoked(T) - x_f||, and the distribution of this error over initial conditions and control horizons.
[Section 5, Eq. (1)] The control input is designed from matrices A and B fitted to training data, so the success of the controller against the true nonlinear RNN is the load-bearing result. The paper itself states that predicted and evoked trajectories diverge beyond short prediction horizons, and then asserts that they nevertheless 'converge at the desired target state' without giving any mechanism or quantitative support. The authors should report the final-state error as a function of the control horizon T, compare the model-based minimum-energy input against baselines such as random stimulation or stimulation of non-hub sites, and quantify how model mismatch at the target affects the outcome. Without this, the statement that the linear model is 'accurate enough' over the control horizon remains an unsupported assumption.
[Section 4, Figure 3d and Section 5] The tradeoff between control horizon and prediction horizon is described only qualitatively. Figure 3d shows that the R2 score for the uncontrolled delay-space model declines rapidly after roughly a few timesteps, but the paper does not provide the corresponding R2 at the control horizon used in the control simulations, nor does it report how the required control energy grows with T. A quantitative analysis of this tradeoff is essential because the entire method depends on the finite-dimensional linear model being reliable over the time needed to reach the target.
[Section 3 and Figure 4a] The two-step workflow's first step—hub identification via Causal Flow—is not demonstrated in the simulations. The paper relies on a prior publication [25] for the Causal Flow method, but the RNN used in Figure 4 is said to have 'hub nodes inferred via Causal Flow' without any details of how Causal Flow was applied to the recurrent network or whether the identified hubs were validated. To support the integrated feasibility claim, the authors should either provide the Causal Flow analysis for the same RNN or explicitly state that the control demonstration assumes known hubs and relegate the hub-identification validation to future work.

minor comments (5)

[Section 1.1] There is a typo in 'T ransfer Entropy'; it should read 'Transfer Entropy'.
[Annotated Bibliography, [30]] The word 'chactic' should be 'chaotic'.
[Section 5] The expression for the energy cost is garbled in the text: 'E = Pt=T t=0 |u(t)|2' should be formatted as a proper summation, e.g., E = sum_{t=0}^{T} ||u(t)||^2.
[Section 5, Figure 4 caption] The text refers to 'predicted trajectories from Eq(2) and Eq(3) in panel a', but only Eq. (1) is defined in the manuscript; equation numbers should be added or the cross-reference corrected.
[Figure 3d] The caption reports an 'R2-score' but does not define R2 or specify the axes of the plot; including axis labels and a definition would make the figure self-contained.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the control workflow is tested against the true nonlinear RNN, and the hub-identification step cites independent, externally falsifiable prior work.

full rationale

The paper's claimed chain is: (i) identify hubs from spontaneous activity via Causal Flow, citing [25] for the method and its experimental validation; (ii) fit a delay-space linear model A, B from a training period (Eq. 1, DMD with control); (iii) compute a minimum-energy control input from that fitted model; and (iv) evaluate the resulting trajectory on the actual nonlinear RNN. Steps (ii)-(iv) are self-contained: the A and B matrices are estimated from data, but the central claim that the evoked xf closely approximates the desired target xf is checked against the simulating RNN, which is external to the fitted model. No equation is equal to another by construction, and no fitted parameter is renamed as a prediction. The target state is not defined as the output of the fitted model. The reliance on [25] for the hub-identification component is a citation to peer-reviewed experimental results that are externally falsifiable, not an unverified self-citation. The acknowledged divergence between predicted and evoked trajectories beyond short horizons (Section 5, Fig. 4d) and the unquantified 'they converge at the desired target state' statement are correctness and evidence limitations, not circular reductions.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claims rest on standard delay-embedding theorems, on the prior empirical hub result of reference [25], and on a linear-input modeling assumption for stimulation that is stated but not derived. The demonstration depends on fitted matrices A and B and on several simulation parameters that are not reported.

free parameters (3)

Delay embedding dimension D and lag tau = Not reported
The delay-space construction and Hankel matrix depend on choosing D and tau; the paper states the optimization is explained in reference [25], but the values used in the RNN simulations are not given.
Koopman matrices A and B (Eq. 1) = Not reported, estimated by least squares from training data
The input-driven linear model z(t+1)=A z(t)+B u(t) is fit to simulated training trajectories; all control design depends on these fitted matrices, and their accuracy is not quantified.
Control horizon T and target state z_f = Not reported
The minimum-energy control depends on the chosen horizon and target; the paper discusses the horizon tradeoff qualitatively but gives no specific values or sensitivity analysis.

assumptions (4)

domain assumption The recorded activity arises from a deterministic low-dimensional chaotic attractor, so Takens' delay embedding reconstructs the state space.
The hub identification and Koopman linearization rest on the validity of delay embedding for neural data, which the paper argues from the literature but does not establish for its own demonstrations.
ad hoc to paper Neural dynamics become approximately linear in a high-dimensional delay space (Koopman formalism).
The linearization is approximate and only valid over short horizons; the paper acknowledges prediction accuracy declines rapidly (Fig. 3d), making this a load-bearing assumption for the control step.
ad hoc to paper External stimulation inputs u(t) enter the delay-space dynamics additively and linearly, as in z(t+1)=A z(t)+B u(t).
There is no derivation that current injection maps to a linear input term in delay space for a spiking or rate network; this is a modeling assumption stated in Eq. 1 without justification.
domain assumption Hubs identified by Causal Flow from spontaneous activity are the most effective stimulation sites.
The paper relies on the prior empirical result of Nejatbakhsh et al. (reference [25]) to justify the first step; it is treated as established rather than re-derived.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Towards model-based design of causal manipulations of brain circuits with high spatiotemporal precision." pith.science (2026). https://pith.science/paper/MNKTMOQH

@misc{pith2026250524790,
  author       = {Pith},
  title        = {Pith review of: Towards model-based design of causal manipulations of brain circuits with high spatiotemporal precision},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MNKTMOQH}},
  note         = {Machine review of arXiv:2505.24790}
}

read the original abstract

Recent advancements in neurotechnology enable precise spatiotemporal patterns of microstimulations with single-cell resolution. The choice of perturbation sites must satisfy two key criteria: efficacy in evoking significant responses and selectivity for the desired target effects. This choice is currently based on laborious trial-and-error procedures, unfeasible for sequences of multi-site stimulations. Efficient methods to design complex perturbation patterns are urgently needed. Can we design a spatiotemporal pattern of stimulation to steer neural activity and behavior towards a desired target? We outline a method for achieving this goal in two steps. First, we identify the most effective perturbation sites, or hubs, only based on short observations of spontaneous neural activity. Second, we provide an efficient method to design multi-site stimulation patterns by combining approaches from nonlinear dynamical systems, control theory and data-driven methods. We demonstrate the feasibility of our approach using multi-site stimulation patterns in recurrent network models.

Figures

Figures reproduced from arXiv: 2505.24790 by the authors.

**Figure 1.** a) Schematic of a trajectory in neural state space. The goal is to design a stimulation sequence u(1), . . . , u(T) that steers the trajectory to a desired target xf . b) Schematic of an Utah array implanted into the cortex of a monkey. CCM identifies the hub electrodes from spontaneous activity recorded from this array. c) Schematic of hypothetical neural network that gives rise to the activity recorded by the Utah… view at source ↗

**Figure 2.** a) Left: Lorenz attractor in dynamical state space. Middle: x(t) and 3D Hankel matrix constructed from X(t). Right: Trajectory traced out in the delay space by the points in the Hankel matrix. b) Left: Schematic of a network with 2 subnetworks X and Y. X consists of 3 neurons that support a chaotic trajectory of a combinatorial threshold linear network (CTLN) [31]. Subnetwork X has unidirectional projections to subn… view at source ↗

**Figure 3.** a) Schematic of Koopman operator. Nonlinear dynamics of a system becomes linear in higher dimensional delay space. b) Schematic of an RNN (top) and timeseries of 3 representative neurons (bottom). c) In delay space, the Koopman operator can be approximated by the matrix which maps one column of the Hankel matrix to its adjacent column on the right. This matrix A can be obtained by solving the linear least squares pr… view at source ↗

Figures from the paper (1 more)

**Figure 4.** Figure 4: a) Schematic of an RNN with hub nodes inferred via Causal Flow shown in light blue, where current u(t) is injected during a training period used to fit matrices A and B to observed trajectories. b) Examples of training signals used in the simulations. Trajectories of t…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

36 extracted references · 36 canonical work pages

[25]

et al.Predicting the effect of micro-stimulation on macaque prefrontal activity based on spontaneous circuit dynamics

Nejatbakhsh, A. et al.Predicting the effect of micro-stimulation on macaque prefrontal activity based on spontaneous circuit dynamics. Physical Review Research5, 043211 (2023)

work page 2023
[1]

Oehrn, C. R. et al. Chronic adaptive deep brain stimulation versus conventional stimulation in parkinson’s disease: a blinded randomized feasibility trial. Nature Medicine 1–12 (2024). 9

work page 2024
[2]

D., Britten, K

Salzman, C. D., Britten, K. H. & Newsome, W. T. Cortical microstimulation influences perceptual judgements of motion direction. Nature 346, 174–177 (1990)

work page 1990
[3]

D., Murasugi, C

Salzman, C. D., Murasugi, C. M., Britten, K. H. & Newsome, W. T. Microstimulation in visual area mt: effects on direction discrimination performance. Journal of Neuroscience12, 2331–2355 (1992)

work page 1992
[4]

M., Salzman, C

Murasugi, C. M., Salzman, C. D. & Newsome, W. T. Microstimulation in visual area mt: effects of varying pulse amplitude and frequency. Journal of Neuroscience13, 1719–1729 (1993)

work page 1993
[5]

Fisher, R. S. & Velasco, A. L. Electrical brain stimulation for epilepsy. Nature Reviews Neurology 10, 261–270 (2014)

work page 2014
[6]

& Esteky, H

Afraz, S.-R., Kiani, R. & Esteky, H. Microstimulation of inferotemporal cortex influences face categorization. Nature 442, 692–695 (2006)

work page 2006
[7]

Parvizi, J. et al. Electrical stimulation of human fusiform face-selective regions distorts face perception. Journal of Neuroscience32, 14915–14920 (2012)

work page 2012

Show all 36 references

[8]

& Tsao, D

Moeller, S., Crapse, T., Chang, L. & Tsao, D. Y. The effect of face patch microstimulation on perception of faces and objects. Nature neuroscience20, 743–752 (2017)

2017
[9]

Rigotti, M. et al. The importance of mixed selectivity in complex cognitive tasks. Nature 497, 585–590 (2013)

2013
[10]

Fusi, S., Miller, E. K. & Rigotti, M. Why neurons mix: high dimensionality for higher cognition. Current opinion in neurobiology37, 66–74 (2016)

2016
[11]

Brovelli, A. et al. Beta oscillations in a large-scale sensorimotor cortical network: directional influences revealed by granger causality. Proceedings of the National Academy of Sciences101, 9849–9854 (2004)

2004
[12]

Measuring information transfer

Schreiber, T. Measuring information transfer. Physical review letters85, 461 (2000)

2000
[13]

Rolls, E. T. & Deco, G. The noisy brain: stochastic dynamics as a principle of brain function (Oxford university press, 2010)

2010
[14]

O’Shea, D. J. et al. Direct neural perturbations reveal a dynamical mechanism for robust computation. bioRxiv 2022–12 (2022)

2022
[15]

D., Luo, T

Kim, T. D., Luo, T. Z., Pillow, J. W. & Brody, C. D. Inferring latent dynamics underlying neural population activity via neural differential equations. In International Conference on Machine Learning, 5551–5561 (PMLR, 2021)

2021
[16]

& Barab´ asi, A.-L

Liu, Y.-Y. & Barab´ asi, A.-L. Control principles of complex systems. Reviews of Modern Physics 88, 035006 (2016)

2016
[17]

Gu, S. et al. Controllability of structural brain networks. Nature communications 6, 8414 (2015)

2015
[18]

Manjunatha, K. K. H. et al. Controlling target brain regions by optimal selection of input nodes. PLOS Computational Biology20, e1011274 (2024). 10

2024
[19]

Beggs, J. M. & Plenz, D. Neuronal avalanches in neocortical circuits. Journal of neuroscience 23, 11167–11177 (2003)

2003
[20]

L., Yang, H., Petermann, T., Roy, R

Shew, W. L., Yang, H., Petermann, T., Roy, R. & Plenz, D. Neuronal avalanches imply maximum dynamic range in cortical networks at criticality.Journal of neuroscience29, 15595– 15600 (2009)

2009
[21]

Detecting strange attractors in turbulence

Takens, F. Detecting strange attractors in turbulence. In Dynamical Systems and Turbulence, Warwick 1980: proceedings of a symposium held at the University of Warwick 1979/80, 366– 381 (Springer, 2006)

1980
[22]

Sugihara, G. et al. Detecting causality in complex ecosystems. science 338, 496–500 (2012)

2012
[23]

J., Takahashi, H

Tajima, S., Mita, T., Bakkum, D. J., Takahashi, H. & Toyoizumi, T. Locally embedded presages of global network bursts. Proceedings of the National Academy of Sciences114, 9517–9522 (2017)

2017
[24]

& Toyoizumi, T

Tajima, S., Yanagawa, T., Fujii, N. & Toyoizumi, T. Untangling brain-wide dynamics in consciousness by cross-embedding. PLoS computational biology11, e1004537 (2015)

2015
[26]

& Arieli, A

Kenet, T., Bibitchkov, D., Tsodyks, M., Grinvald, A. & Arieli, A. Spontaneously emerging cortical representations of visual attributes. Nature 425, 954–956 (2003)

2003
[27]

& La Camera, G

Mazzucato, L., Fontanini, A. & La Camera, G. Dynamics of multistable states during ongoing and evoked cortical activity. Journal of Neuroscience35, 8214–8231 (2015)

2015
[28]

& Jirsa, V

Deco, G. & Jirsa, V. K. Ongoing cortical activity at rest: criticality, multistability, and ghost attractors. Journal of Neuroscience32, 3366–3375 (2012)

2012
[29]

Manley, J. et al. Simultaneous, cortex-wide dynamics of up to 1 million neurons reveal un- bounded scaling of dimensionality with neuron number. Neuron 112, 1694–1709 (2024)

2024
[30]

& Mezic, I

Arbabi, H. & Mezic, I. Ergodic theory, dynamic mode decomposition, and computation of spectral properties of the koopman operator. SIAM Journal on Applied Dynamical Systems 16, 2096–2126 (2017)

2017
[31]

& Curto, C

Parmelee, C., Moore, S., Morrison, K. & Curto, C. Core motifs predict dynamic attractors in combinatorial threshold-linear networks. PloS one 17, e0264456 (2022)

2022
[32]

& Mezi´ c, I

Korda, M. & Mezi´ c, I. Linear predictors for nonlinear dynamical systems: Koopman operator meets model predictive control. Automatica 93, 149–160 (2018)

2018
[33]

L., Brunton, S

Proctor, J. L., Brunton, S. L. & Kutz, J. N. Dynamic mode decomposition with control. SIAM Journal on Applied Dynamical Systems15, 142–161 (2016)

2016
[34]

& Bullo, F

Pasqualetti, F., Zampieri, S. & Bullo, F. Controllability metrics, limitations and algorithms for complex networks. IEEE Transactions on Control of Network Systems1, 40–52 (2014). 11

2014
[35]

Ahamed, T., Costa, A. C. & Stephens, G. J. Capturing the continuous complexity of behaviour in caenorhabditis elegans. Nature Physics 17, 275–283 (2021)

2021
[36]

W., Johnson, L

Brunton, B. W., Johnson, L. A., Ojemann, J. G. & Kutz, J. N. Extracting spatial–temporal coherent patterns in large-scale neural recordings using dynamic mode decomposition. Journal of neuroscience methods258, 1–15 (2016). 12

2016

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [25]

et al.Predicting the effect of micro-stimulation on macaque prefrontal activity based on spontaneous circuit dynamics

Nejatbakhsh, A. et al.Predicting the effect of micro-stimulation on macaque prefrontal activity based on spontaneous circuit dynamics. Physical Review Research5, 043211 (2023)

work page 2023

[2] [1]

Oehrn, C. R. et al. Chronic adaptive deep brain stimulation versus conventional stimulation in parkinson’s disease: a blinded randomized feasibility trial. Nature Medicine 1–12 (2024). 9

work page 2024

[3] [2]

D., Britten, K

Salzman, C. D., Britten, K. H. & Newsome, W. T. Cortical microstimulation influences perceptual judgements of motion direction. Nature 346, 174–177 (1990)

work page 1990

[4] [3]

D., Murasugi, C

Salzman, C. D., Murasugi, C. M., Britten, K. H. & Newsome, W. T. Microstimulation in visual area mt: effects on direction discrimination performance. Journal of Neuroscience12, 2331–2355 (1992)

work page 1992

[5] [4]

M., Salzman, C

Murasugi, C. M., Salzman, C. D. & Newsome, W. T. Microstimulation in visual area mt: effects of varying pulse amplitude and frequency. Journal of Neuroscience13, 1719–1729 (1993)

work page 1993

[6] [5]

Fisher, R. S. & Velasco, A. L. Electrical brain stimulation for epilepsy. Nature Reviews Neurology 10, 261–270 (2014)

work page 2014

[7] [6]

& Esteky, H

Afraz, S.-R., Kiani, R. & Esteky, H. Microstimulation of inferotemporal cortex influences face categorization. Nature 442, 692–695 (2006)

work page 2006

[8] [7]

Parvizi, J. et al. Electrical stimulation of human fusiform face-selective regions distorts face perception. Journal of Neuroscience32, 14915–14920 (2012)

work page 2012