OSCR

Shared latent representations of speech production for cross-patient speech decoding.

Code ↔ Paper

23 matches between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

15 blocks of code here have a permanent identifier (oscr:b:…), listed with the matches below: citable for good, with a page of its own. Select a match and press Cite to copy it. What a block identifier is.

The 23 matches
  1. [1] § Methods › Pitch subsampling ↔ aligned_decoding/processing_utils/poisson_disk_sampling.py, lines 9 to 82 · score 0.82 · Poisson disk sampling, desired pitch, subsampled electrodes, select electrodes · block oscr:b:ZXV2AT04MC9WG-V · Report a mismatch
  2. [2] § Methods › Connectionist temporal classification recurrent neural network (CTC-RNN) decoding ↔ aligned_decoding/realtime_sim/realtime_nn_model.py, lines 101 to 150 · score 0.76 · stacked RNN model, sliding window, CTC loss, blank, stride, classification · block oscr:b:0BAF3GQMSDQZD-N · Report a mismatch
  3. [3] § Results › Cross-patient alignment improves speech decoding in a simulated real-time environment ↔ aligned_decoding/scripts/train_ctc_rnn.py, lines 1 to 55 · score 0.72 · phoneme error rate, CTC RNN model, neural recording, phoneme sequences, speech decoding, loss · block oscr:b:W97GPQ2KT88SC-Z · Report a mismatch
  4. [4] § Methods › Simulated real-time data generation ↔ aligned_decoding/realtime_sim/realtime_processing.py, lines 10 to 39 · score 0.72 · Bandpass filtering, High gamma, RMS, bins, power, bands · block oscr:b:GQSDSS1GKEDX2-F · Report a mismatch
  5. [5] § Methods › Alignment of latent dynamics ↔ aligned_decoding/alignment/AlignCCA.py, lines 11 to 119 · score 0.71 · canonical correlations, target space, patient space, CCA alignment, latent dynamics, target patient · block oscr:b:QTGS8FP2VXMY0-A · Report a mismatch
  6. [6] § Methods › Cross-patient projection ↔ aligned_decoding/alignment/JointPCA.py, lines 13 to 162 · score 0.70 · transformation matrices, PCA transform, PCA decomposition, latent space, optionally, components · block oscr:b:ADJACFH442H80-B · Report a mismatch
  7. [7] § Results › High-resolution and broad coverage are critical to alignment ↔ aligned_decoding/processing_utils/poisson_disk_sampling.py, lines 9 to 82 · score 0.70 · Poisson disk sampling, sampled electrodes, pitch subsampled, mm, decoding · block oscr:b:ZXV2AT04MC9WG-V · Report a mismatch
  8. [8] § Methods › Connectionist temporal classification recurrent neural network (CTC-RNN) decoding ↔ aligned_decoding/realtime_sim/realtime_datamodule.py, lines 578 to 724 · score 0.67 · dimensionality reduction, latent space, source patient, post, truncated, aligned cross patient · block oscr:b:WAAXV288YKKGE-7 · Report a mismatch
  9. [9] § Methods › High-gamma extraction ↔ aligned_decoding/realtime_sim/realtime_processing.py, lines 10 to 39 · score 0.66 · bandpass filtered signals, high gamma, Channels, neural, decoding · block oscr:b:GQSDSS1GKEDX2-F · Report a mismatch
  10. [10] § Results › Cross-patient alignment preserves spatial articulator maps ↔ aligned_decoding/figure_analyses/supp/supp_fig_9.ipynb, lines 1241 to 1271 · score 0.64 · unaligned reconstruction, aligned reconstructed, cross patient speech, cross patient decoding, decoding accuracies, TME · Report a mismatch
  11. [11] § Methods › Cross-patient projection ↔ aligned_decoding/scripts/aligned_decode_svm.py, lines 74 to 115 · score 0.63 · bagged SVM classifier, fold cross validation, PCA, Cross patient, accuracy, trained · block oscr:b:KN23V2W22E3TE-1 · Report a mismatch
  12. [12] § Methods › Alignment of latent dynamics ↔ aligned_decoding/nn_models/data_utils/AlignCCA.py, lines 11 to 105 · score 0.62 · canonical correlations, target space, CCA alignment, latent dynamics, target patient, mapping · block oscr:b:28RQ3BDZMPEEK-L · Report a mismatch
  13. [13] § Methods › Connectionist temporal classification recurrent neural network (CTC-RNN) decoding ↔ aligned_decoding/scripts/tune_ctc_rnn.py, lines 638 to 685 · score 0.61 · Hyperparameter optimization, CTC RNN, cross validation, target patient, fit, aligned decoding · block oscr:b:AMHSSF8X51FN6-5 · Report a mismatch
  14. [14] § Methods › Connectionist temporal classification recurrent neural network (CTC-RNN) decoding ↔ aligned_decoding/scripts/tune_ctc_rnn_align.py, lines 581 to 628 · score 0.61 · Hyperparameter optimization, CTC RNN, cross validation, target patient, fit, aligned decoding · block oscr:b:SBXN84WZY1MKZ-Z · Report a mismatch
  15. [15] § Methods › Single-electrode articulator tuning ↔ aligned_decoding/scripts/aligned_decode_svm_ncv.py, lines 323 to 394 · score 0.59 · nested cross validation, linear discriminant, splitting, predict, fold, classify · block oscr:b:SM3GBF9W8AYHD-9 · Report a mismatch
  16. [16] § Methods › Single-electrode articulator tuning ↔ aligned_decoding/scripts/aligned_decode_grid_subsample.py, lines 271 to 340 · score 0.58 · nested cross validation, linear discriminant, splitting, predict, fold, classify · block oscr:b:6XHBXDAJFJ70E-9 · Report a mismatch
  17. [17] § Results › Cross-patient alignment improves speech decoding in a simulated real-time environment ↔ aligned_decoding/realtime_sim/realtime_nn_model.py, lines 101 to 150 · score 0.57 · sliding windows, error rate, RNN model, loss, classification, CTC · block oscr:b:0BAF3GQMSDQZD-N · Report a mismatch
  18. [18] § Methods › Cross-patient projection ↔ aligned_decoding/scripts/aligned_decode_spatialAvg_subsample.py, lines 239 to 312 · score 0.56 · class weight, cross validation, bagged, kernel, predictions, folds · block oscr:b:7P30HEJSXPME2-I · Report a mismatch
  19. [19] § Methods › Cross-patient phoneme decoding ↔ aligned_decoding/decoders/cross_pt_decoders.py, lines 89 to 180 · score 0.56 · latent dimensionality, cross patient sources, truncated, concatenated, variance, target patient · block oscr:b:HK80DVPJM8FXJ-Z · Report a mismatch
  20. [20] § Results › Cross-patient alignment preserves spatial articulator maps ↔ aligned_decoding/alignment/AlignCCA.py, lines 11 to 119 · score 0.56 · extract latent dynamics, CCA alignment, latent space, optionally, correlation, target patient · block oscr:b:QTGS8FP2VXMY0-A · Report a mismatch
  21. [21] § Methods › Cross-patient projection ↔ aligned_decoding/figure_analyses/fig_3.ipynb, lines 693 to 763 · score 0.55 · ROC AUC, aligned reconstructions, compatibility, flattened, components, unaligned · Report a mismatch
  22. [22] § Methods › Latent dynamics extraction ↔ aligned_decoding/figure_analyses/fig_2.ipynb, lines 1037 to 1131 · score 0.55 · articulatory features, t-SNE, perplexity, latent dynamics, Silhouette, Clustering · Report a mismatch
  23. [23] § Methods › Latent dynamics extraction ↔ aligned_decoding/figure_analyses/supp/supp_fig_6_7.ipynb, lines 1044 to 1139 · score 0.55 · articulatory features, t-SNE, perplexity, latent dynamics, Silhouette, Clustering · Report a mismatch

Paper

Loaded from Europe PMC by your browser; OSCR does not store it: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The authors' code

Python · 255 lines · 9.1 KB · MIT · 2 matches

  1. """Poisson disk sampling for spatially uniform electrode subsampling."""
  2. import numpy as np
  3. import scipy.io as sio
  4. import matplotlib.pyplot as plt
  5. import time
  6. def pitch_subsample_sig_channels(pt, pitch, data_path):
  7. """Subsamples electrodes at a given pitch and returns significant indices.
  8. Uses Poisson disk sampling to select electrodes at the specified pitch
  9. (inter-electrode spacing in mm), then identifies which of the sampled
  10. electrodes are significant. Recursively retries if no significant
  11. channels are sampled.
  12. Args:
  13. pt (str): Subject identifier string.
  14. pitch (float): Desired inter-electrode pitch in millimeters.
  15. data_path (str): Root directory containing subject data folders.
  16. Returns:
  17. ndarray: Indices of significant channels within the subsampled set.
  18. """
  19. # load in channel map
  20. chanMap = sio.loadmat(f'{data_path}/{pt}/{pt}_channelMap.mat')['chanMap']
  21. # load in significant channel data
  22. sigChan = np.squeeze(
  23. sio.loadmat(f'{data_path}/{pt}/{pt}_sigChannel.mat')['sigChannel'])
  24. # trim nan edges if necessary
  25. if chanMap.shape[1] == 24:
  26. chanMap = chanMap[:,1:-1]
  27. # to preserve pitch when sampling across different grid sizes, calculate
  28. # number of electrodes to sample based on the desired pitch
  29. if pt in ['S14', 'S22', 'S23', 'S26']:
  30. mmX = 11.3
  31. mmY = 22.5
  32. maxElec = 128
  33. elif pt in ['S33', 'S39', 'S58', 'S62']:
  34. mmX = 37.8
  35. mmY = 20.6
  36. maxElec = 256
  37. nElec = round(mmX * mmY / pitch**2)
  38. if nElec >= maxElec:
  39. # just sample all electrodes if we're sampling more than the max
  40. elecPt = np.arange(1, maxElec+1)
  41. else:
  42. # parameters for poisson disk sampling
  43. gridX, gridY = chanMap.shape
  44. domain = (gridX, gridY)
  45. spacing = np.floor(np.sqrt(gridX * gridY / nElec))
  46. # do poisson disk sampling and -1 to get 0-indexed
  47. elecIdx = poisson_disk_sampling(domain, spacing, nElec)
  48. elecIdx = np.round(elecIdx).astype(int) - 1
  49. # convert 2D coordinates to channel numbers
  50. elecPt = chanMap[elecIdx[:, 0], elecIdx[:, 1]]
  51. elecPt = np.nan_to_num(elecPt, nan=-1).astype(int)
  52. # check if we need to sample more electrodes
  53. if elecPt.shape[0] < nElec and spacing == 1:
  54. nRemaining = nElec - elecPt.shape[0]
  55. # get unsampled electrodes
  56. remainingElecs = np.setdiff1d(np.arange(1, gridX * gridY+1), elecPt)
  57. # uniformly sample remaining electrodes
  58. extraSampPt = np.random.choice(remainingElecs, nRemaining,
  59. replace=False)
  60. elecPt = np.concatenate((elecPt, extraSampPt))
  61. # get indices of significant channels in the subsampled set
  62. _, sigIdx, _ = np.intersect1d(sigChan, elecPt, return_indices=True)
  63. # do sampling over if we don't sample at least 1 significant channel
  64. if len(sigIdx) == 0:
  65. return pitch_subsample_sig_channels(pt, nElec, data_path)
  66. return sigIdx
  67. def poisson_disk_sampling(domain, spacing, nPoints, threshold=60,
  68. showIter=False, maxIter=1000):
  69. """
  70. Poisson Disk Sampling algorithm. Adapted from MATLAB code by Mohak Patel
  71. (Brown University, 2016). Follows the algorithm in Bridson 2007
  72. (https://www.cs.ubc.ca/~rbridson/docs/bridson-siggraph07-poissondisk.pdf).
  73. """
  74. ##### Initialize the grid #####
  75. ndim = len(domain)
  76. cellSize = spacing / np.sqrt(ndim)
  77. # Construct grid
  78. sGrid = [np.arange(1, s + 1, cellSize) for s in domain]
  79. sGrid = np.meshgrid(*sGrid, indexing='ij')
  80. sizeGrid = sGrid[0].shape
  81. # Flatten grid points into array of coordinates
  82. sGrid = np.column_stack([g.ravel() for g in sGrid])
  83. emptyGrid = np.ones(sGrid.shape[0], dtype=bool)
  84. nEmptyGrid = np.sum(emptyGrid)
  85. scoreGrid = np.zeros_like(emptyGrid, dtype=int)
  86. ##### Dart-throwing #####
  87. ptsCreated = 0
  88. pts = []
  89. iter = 0
  90. start = time.time()
  91. while ptsCreated < nPoints and nEmptyGrid > 0:
  92. if iter > maxIter:
  93. print(f'Reached max iterations with {ptsCreated} points. Trying sampling again.')
  94. return poisson_disk_sampling(domain, spacing, nPoints, threshold)
  95. availGrid = np.where(emptyGrid)[0]
  96. dataPts = np.minimum(nEmptyGrid, nPoints)
  97. # sample nPoints from available grid points
  98. sampPts = np.random.choice(availGrid, dataPts, replace=False)
  99. # dart throws
  100. tempPts = sGrid[sampPts] + cellSize * np.random.rand(dataPts, ndim)
  101. ### Find good dart throws ###
  102. if len(pts) > 0:
  103. allPts = np.vstack((pts, tempPts))
  104. else:
  105. allPts = tempPts
  106. # get distance to nearest neighbor
  107. neighDist = min_neighbor_distance(allPts, tempPts)
  108. # check which points are valid
  109. inDomain = np.all(tempPts < domain, axis=1) # within domain
  110. goodSpacing = neighDist > spacing # far enough from other points
  111. validPts = inDomain & goodSpacing
  112. scorePts = tempPts[~validPts, :] # keep scores from bad throws
  113. tempPts = tempPts[validPts, :] # save good throws
  114. ### update tracking grids ###
  115. # update empty grid
  116. emptyPts = np.floor((tempPts + cellSize - 1) / cellSize).astype(int)
  117. # convert to linear index
  118. emptyPtIdx = np.ravel_multi_index(emptyPts.T-1, sizeGrid)
  119. # emptyPtIdx = np.ravel_multi_index((emptyPts[:,0], emptyPts[:,1]), sizeGrid)
  120. emptyGrid[emptyPtIdx] = False
  121. # update score grid
  122. scorePts = np.floor((scorePts + cellSize - 1) / cellSize).astype(int)
  123. scorePtIdx = np.ravel_multi_index(scorePts.T-1, sizeGrid)
  124. # scorePtIdx = np.ravel_multi_index((scorePts[:,0], scorePts[:,1]), sizeGrid)
  125. scoreGrid[scorePtIdx] += 1
  126. # update empty grid if score grid has exceeded threshold
  127. emptyGrid = emptyGrid & (scoreGrid < threshold)
  128. # update quantities for next iteration
  129. nEmptyGrid = np.sum(emptyGrid)
  130. pts.extend(tempPts)
  131. ptsCreated += tempPts.shape[0]
  132. iter += 1
  133. if showIter:
  134. elapsed = time.time() - start
  135. print(f"Iteration: {iter} Points Created: {ptsCreated} "
  136. f"EmptyGrid: {nEmptyGrid} Total Time: {elapsed:.3f}")
  137. # trim points to nPoints if more are created in last iteration
  138. pts = np.vstack(pts)
  139. if ptsCreated > nPoints:
  140. ptIdxs = np.random.choice(pts.shape[0], nPoints, replace=False)
  141. pts = pts[ptIdxs]
  142. return pts
  143. def min_neighbor_distance(pts, newPts):
  144. """Computes the distance from each new point to its nearest neighbor.
  145. Because ``newPts`` is a subset of ``pts``, the true nearest neighbor is
  146. the second closest point (the first is the point itself).
  147. Args:
  148. pts (ndarray): All existing points with shape (n, ndim).
  149. newPts (ndarray): Query points with shape (m, ndim), must be a
  150. subset of ``pts``.
  151. Returns:
  152. ndarray: Array of length m with the nearest-neighbor distance for
  153. each query point.
  154. """
  155. # find distances to nearest neighbors
  156. _, D = knn_search(pts, newPts, 2)
  157. # since pts and newPts will include the same points, nearest neighbor
  158. # distance will be the second index, since the first will be the point
  159. # itself
  160. return D[:, 1]
  161. def knn_search(pts, newPts, k):
  162. """Brute-force k-nearest-neighbor search.
  163. Args:
  164. pts (ndarray): Reference points with shape (n, ndim).
  165. newPts (ndarray): Query points with shape (m, ndim).
  166. k (int): Number of nearest neighbors to return.
  167. Returns:
  168. tuple: (I, D) where I is an (m, k) integer array of neighbor indices
  169. into ``pts`` and D is an (m, k) float array of corresponding
  170. Euclidean distances, both sorted by ascending distance.
  171. """
  172. m = newPts.shape[0]
  173. D = np.zeros((m, k))
  174. I = np.zeros((m, k), dtype=int)
  175. for i in range(m):
  176. dist = np.sqrt(np.sum((pts - newPts[i])**2, axis=1))
  177. I[i] = np.argsort(dist)[:k]
  178. D[i] = np.sort(dist)[:k]
  179. return I, D
  180. if __name__ == '__main__':
  181. # testing the grid sampling
  182. gridX = 8
  183. gridY = 16
  184. for nElec in [10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 100]:
  185. print('### Sampling for nElec =', nElec, '###')
  186. # nElec =
  187. domain = (gridX, gridY)
  188. spacing = np.floor(np.sqrt(gridX * gridY / nElec))
  189. # print(spacing)
  190. n_grids = 3
  191. saved_grids = np.zeros((n_grids, gridX, gridY))
  192. for i in range(n_grids):
  193. points = poisson_disk_sampling(domain, spacing, nElec)
  194. points = np.round(points).astype(int) - 1
  195. print(f'Sampled {points.shape[0]} points')
  196. grid = np.zeros((gridX, gridY))
  197. grid[points[:, 0], points[:, 1]] = 1
  198. # plt.figure(figsize=(8, 8))
  199. plt.imshow(grid.T, cmap='gray', origin='lower')
  200. plt.show()
  201. saved_grids[i] = grid
  202. # check that the grids are different
  203. n_unique_grids = np.unique(saved_grids, axis=0).shape[0]
  204. print(n_unique_grids, n_unique_grids == n_grids)

poisson_disk_sampling.py at commit a70f1ae, under MIT · at the source

Overview

Authors: Z. Spalding1, S. Duraivel1,2, S. Rahimpour3,4, C. Wang1, K. Barth1, C. Schmitz1, S. P. Lad3, A. H. Friedman3, D. G. Southwell3,5,6, J. Viventi1,3,5,6, G. B. Cogan1,3,6,7,8
  1. Department of Biomedical Engineering, Duke University,Durham, NC USA
  2. Present Address: McGovern Institute for Brain Research, Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology,Cambridge, MA USA
  3. Department of Neurosurgery, Duke School of Medicine,Durham, NC USA
  4. Present Address: Department of Neurosurgery, University of Utah,Salt Lake City, UT USA
  5. Department of Neurobiology, Duke School of Medicine,Durham, NC USA
  6. Duke Comprehensive Epilepsy Center, Duke University,Durham, NC USA
  7. Department of Neurology, Duke School of Medicine,Durham, NC USA
  8. Center for Cognitive Neuroscience, Duke University,Durham, NC USA
Institutions: Duke University (United States); Massachusetts Institute of Technology (United States); University of Utah (United States)
Journal: Nature communications, volume 17, issue 1, article 8716
Dates: received 1 September 2025; accepted 1 July 2026; published online 16 July 2026
Type: Research article · Language: English
License: CC BY-NC-ND
Identifiers: DOI 10.1038/s41467-026-75455-1 · PMID 42463660 · PMCID PMC13490415 · OpenAlex W4413425109
Open access: gold, a free copy (OpenAlex)
Status: code verified (how the code is found and verified)
Categories: intracranial EEG (iEEG / ECoG / SEEG) (modality), human (organism), cognitive (subfield)
Methods: Spectral & time-frequency, Connectivity, Statistics, Smoothing, state filtering, decompositions, Machine learning, Preprocessing
Keywords: Brain-machine interface, Motor cortex, Language
MeSH: Brain-Computer Interfaces*, Speech*, Adult, Brain, Electrocorticography, Female, Humans, Male (* major topic)
Topic: EEG and Brain-Computer Interfaces (Cognitive Neuroscience, Neuroscience), according to OpenAlex
Citations: cited by 1 paper (Europe PMC); 55 references in the paper

Abstract

The abstract is not reproduced here: the paper's license (CC BY-NC-ND) does not allow it. Read it in the paper, at the publisher or on Europe PMC.

Repositories

Its files are read in the Code ↔ Paper reader above, with 23 matches between paragraphs and lines of code.

gamaleldin/TME

License: none: the authors keep all their rights
State: the link answers, verified on 17 November 2026
Evidence: files inventoried
Commit: ad1adf835e72dbba012406b5a3af30701adc8993, 8 September 2017
Languages: MATLAB (13), C/C++ (8), C++ (8)
Size: 40 files, 29 scripts
Software Heritage: not archived
Found in: the text, “Generation of surrogate data”
Holds: README, tests
Not found: license file, CITATION.cff, environment file, continuous integration, documentation
Availability: 2 checks, the latest on 17 November 2026: the link answers
  • 17 November 2026: the link answers
  • 27 September 2026: the link answers
30 files

coganlab/cross_patient_speech_decoding

License: MIT
State: the link answers, verified on 27 September 2026
Evidence: files inventoried
Commit: a70f1ae711cf3627cf9357d2a72868a74f2d774f, 3 June 2026
Languages: Python (45), Jupyter (29)
Size: 83 files, 74 scripts
Software Heritage: not archived
Found in: “Code availability”
Holds: README, license file, environment (environment.yml, requirements.txt), 29 notebooks
Not found: CITATION.cff, tests, continuous integration, documentation
Tools: NumPy (63 files), scikit-learn (37 files), SciPy (37 files), Matplotlib (33 files), seaborn (28 files), pandas (27 files), Plotly (25 files), statsmodels (22 files), h5py (18 files), PyTorch (15 files), PyTorch Lightning (9 files), Keras (1 file), TensorFlow (1 file)
Availability: 1 check, the latest on 27 September 2026: the link answers
  • 27 September 2026: the link answers
76 files

Zenodo 20531684

License: MIT
State: the link answers, verified on 27 September 2026
Evidence: files inventoried
Size: 1 file
Software Heritage: not checked
Found in: the references
Not found: README, license file, CITATION.cff, environment file, tests, continuous integration, documentation
Tools: NumPy (63 files), scikit-learn (37 files), SciPy (37 files), Matplotlib (33 files), seaborn (28 files), pandas (27 files), Plotly (25 files), statsmodels (22 files), h5py (18 files), PyTorch (15 files), PyTorch Lightning (9 files), Keras (1 file), TensorFlow (1 file)
Availability: 1 check, the latest on 27 September 2026: the link answers (HTTP 200)
  • 27 September 2026: the link answers (HTTP 200)
76 files
At the source:

Code availability statement

The paper has a code availability statement. Its license (CC BY-NC-ND) does not allow reproducing it here; in short, from what the harvester recognized in it:

Read it in the paper: doi.org/10.1038/s41467-026-75455-1.

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

  • 3 repositories of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
  • 177 scripts, each with its path and the digest of its content;
  • 23 matches between paragraphs of the paper and lines of the code (method lexical-v1);
  • neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

No dataset and no data link were found in the paper.

Data availability statement

The paper has a data availability statement. Its license (CC BY-NC-ND) does not allow reproducing it here; in short, from what the harvester recognized in it:

  • no repository, dataset or request procedure was recognized in it

Read it in the paper: doi.org/10.1038/s41467-026-75455-1.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 27 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 11 authors, 3 keywords, 8 MeSH terms, 3 funders, 52 references.

Cite

This paper

Spalding, Z., Duraivel, S., Rahimpour, S., Wang, C., Barth, K., Schmitz, C., Lad, S. P., Friedman, A. H., Southwell, D. G., Viventi, J., & Cogan, G. B. (2026). Shared latent representations of speech production for cross-patient speech decoding. Nature communications, 17(1), 8716. https://doi.org/10.1038/s41467-026-75455-1

BibTeX

@article{spalding2026shared,
author = {Spalding, Z. and Duraivel, S. and Rahimpour, S. and Wang, C. and Barth, K. and Schmitz, C. and Lad, S. P. and Friedman, A. H. and Southwell, D. G. and Viventi, J. and Cogan, G. B.},
title = {{Shared latent representations of speech production for cross-patient speech decoding}},
journal = {Nature communications},
year = {2026},
month = jul,
volume = {17},
number = {1},
pages = {8716},
publisher = {Nature Portfolio},
issn = {2041-1723},
doi = {10.1038/s41467-026-75455-1},
url = {https://doi.org/10.1038/s41467-026-75455-1},
pmid = {42463660},
pmcid = {PMC13490415}
}

RIS

TY - JOUR
AU - Spalding, Z.
AU - Duraivel, S.
AU - Rahimpour, S.
AU - Wang, C.
AU - Barth, K.
AU - Schmitz, C.
AU - Lad, S. P.
AU - Friedman, A. H.
AU - Southwell, D. G.
AU - Viventi, J.
AU - Cogan, G. B.
TI - Shared latent representations of speech production for cross-patient speech decoding
T2 - Nature communications
J2 - Nat Commun
PY - 2026
DA - 2026/07/16
VL - 17
IS - 1
SP - 8716
SN - 2041-1723
PB - Nature Portfolio
DO - 10.1038/s41467-026-75455-1
UR - https://doi.org/10.1038/s41467-026-75455-1
LA - en
ER -

CSL-JSON

{
"id": "10.1038/s41467-026-75455-1",
"type": "article-journal",
"title": "Shared latent representations of speech production for cross-patient speech decoding",
"container-title": "Nature communications",
"author": [
{
"family": "Spalding",
"given": "Z."
},
{
"family": "Duraivel",
"given": "S."
},
{
"family": "Rahimpour",
"given": "S."
},
{
"family": "Wang",
"given": "C."
},
{
"family": "Barth",
"given": "K."
},
{
"family": "Schmitz",
"given": "C."
},
{
"family": "Lad",
"given": "S. P."
},
{
"family": "Friedman",
"given": "A. H."
},
{
"family": "Southwell",
"given": "D. G."
},
{
"family": "Viventi",
"given": "J."
},
{
"family": "Cogan",
"given": "G. B."
}
],
"container-title-short": "Nat Commun",
"volume": "17",
"issue": "1",
"page": "8716",
"DOI": "10.1038/s41467-026-75455-1",
"PMID": "42463660",
"PMCID": "PMC13490415",
"ISSN": "2041-1723",
"publisher": "Nature Portfolio",
"URL": "https://doi.org/10.1038/s41467-026-75455-1",
"language": "en",
"issued": {
"date-parts": [
[
2026,
7,
16
]
]
}
}

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Similar papers

The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.

[1] doi:10.1038/s41467-023-42555-1 [code]
High-resolution neural recordings improve the accuracy of speech decoding.
Journal: Nature communications
In common: Statistics and Machine Learning Toolbox, intracranial EEG (iEEG / ECoG / SEEG), cognitive, 15 references, 4 authors
[2] doi:10.1038/s41467-026-71234-0 [code]
Decoupling simultaneous motor imagination and execution via orthogonal ECoG neural representations.
Journal: Nature communications
In common: h5py, statsmodels, PyTorch, 6 other tools, intracranial EEG (iEEG / ECoG / SEEG), 10 references
[3] doi:10.1088/1741-2552/ace8be [code]
Direct speech reconstruction from sensorimotor brain activity with optimized deep learning models.
Journal: Journal of neural engineering
In common: TensorFlow, h5py, PyTorch, 6 other tools, intracranial EEG (iEEG / ECoG / SEEG), 9 references
[4] doi:10.1038/s41467-022-33611-3 [code]
Generalizable spelling using a speech neuroprosthesis in an individual with severe limb and vocal paralysis.
Journal: Nature communications
In common: Keras, TensorFlow, h5py, 7 other tools, cognitive, 7 references
[5] doi:10.1088/1741-2552/ad88a4 [code]
Reducing power requirements for high-accuracy decoding in iBCIs.
Journal: Journal of neural engineering
In common: PyTorch Lightning, h5py, PyTorch, 6 other tools, 8 references
[6] doi:10.1038/s41467-021-27725-3 [code]
Imagined speech can be decoded from low- and cross-frequency intracranial EEG features.
Journal: Nature communications
In common: statsmodels, seaborn, Statistics and Machine Learning Toolbox, 5 other tools, intracranial EEG (iEEG / ECoG / SEEG), cognitive, 8 references
[7] doi:10.3390/brainsci14030196 [code]
Adaptive LDA Classifier Enhances Real-Time Control of an EEG Brain-Computer Interface for Decoding Imagined Syllables.
Journal: Brain sciences
In common: h5py, scikit-learn, SciPy, 2 other tools, cognitive, 10 references
[8] doi:10.1038/s41598-024-60277-2 [code]
Online speech synthesis using a chronically implanted brain-computer interface in an individual with ALS.
Journal: Scientific reports
In common: h5py, PyTorch, scikit-learn, 4 other tools, intracranial EEG (iEEG / ECoG / SEEG), cognitive, 8 references
[9] doi:10.7554/elife.89421 [code]
An emerging view of neural geometry in motor cortex supports high-performance decoding.
Journal: eLife
In common: Keras, TensorFlow, h5py, 7 other tools, 7 references
[10] doi:10.1038/s42003-025-07464-7 [code]
Learning to operate an imagined speech Brain-Computer Interface involves the spatial and frequency tuning of neural activity.
Journal: Communications biology
In common: Statistics and Machine Learning Toolbox, cognitive, 12 references

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Add the badge to the code's README

OSCR: the badge of this paper

If you are an author or a maintainer of this paper's code, you can add this badge. It says what OSCR knows of the paper (its code, its validated tracing map) and links the code to this page. Paste one of these snippets under the README's title. Keep the two marks around it, so that the badge can be replaced in place and is never added twice. OSCR changes nothing in your repository.

Markdown
<!-- oscr-badge:start -->
[![OSCR](https://openscicode.org/badge/doi/10.1038/s41467-026-75455-1.svg)](https://openscicode.org/badge/doi/10.1038/s41467-026-75455-1)
<!-- oscr-badge:end -->
reStructuredText
.. oscr-badge:start

.. image:: https://openscicode.org/badge/doi/10.1038/s41467-026-75455-1.svg
   :target: https://openscicode.org/badge/doi/10.1038/s41467-026-75455-1
   :alt: OSCR

.. oscr-badge:end
HTML
<!-- oscr-badge:start -->
<a href="https://openscicode.org/badge/doi/10.1038/s41467-026-75455-1"><img src="https://openscicode.org/badge/doi/10.1038/s41467-026-75455-1.svg" alt="OSCR"></a>
<!-- oscr-badge:end -->

On GitHub, open the README, choose Edit (the pencil), paste the snippet under the title, then "Commit changes".

If you choose to, you can have the badge kept up to date for you, with oscr badge --write from the command line or with the GitHub Action in your repository's workflows. Every other form and option (a card, AsciiDoc, LaTeX, shields.io) is on the badge page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: code errors and mismatches between the code and the paper about this paper and its code, from readers and its authors, beside the paper and its tracing map.

Report a mismatch between the code and the paper · Report a code error (signed in)

Reproductions: reports from readers who ran the authors' code and did not get the paper's result, what they ran, with which environment, commit and data.

Report a reproduction failure (signed in)

Versions of its code: the releases of its code tied to a version of this paper (the preprint, the accepted manuscript, the version of record…), each with the tracing map versioned with it.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.