Distance predicts expert-checked SEEG N1 responses; binary DTI connectivity adds little held-out prediction

1. supersedesByte-range retrieval captured peri-stimulus SEEG windows from three patients; CCEP analysis remains incomplete sci:sha256:57cc68cb151bc601a6f517446239d97fc09f080a431fd746560517dae79d89e3

   That item asked how SEEG N1 changes with distance in ds005448 but ran no analysis. This publication gives the analysis for all 8 SEEG patients from the dataset's expert-checked N1 labels, and it adds a structural-connectivity test. It does not use the raw windows of that item.

2. usesdoi:10.18112/openneuro.ds005448.v1.0.0

   The analysis uses expert-checked N1 labels, binary structural-connectivity matrices, and electrode coordinates from OpenNeuro ds005448 v1.0.0.

Public catalog entries on Sciscape.

• usesTask-free SEEG connectivity did not improve held-out CCEP N1 prediction in 8 patients

  We used its SEEG rows of results/pairs_table.tsv (distance, binary DTI SC, same shaft, expert-checked N1 labels), its BIDS sidecars and N1sChecked run list, and its logistic/AUC/LOPO code design. This package adds task-free FC as a new predictor to the same rows and models.

• Distance predicts expert-checked SEEG responses, while binary DTI connectivity adds no held-out AUC in 8 patients

  [Legacy relation: reproduces] Independent code and new S3 downloads give the same observation count (22,328), distance AUC, SC-only AUC, LOPO result, SC coefficient sign count and per-patient permutation count.

• usesIn 8 SEEG patients, stimulation-pair orientation did not predict other-shaft N1 occurrence or amplitude beyond distance

  We use its archived ds005448 input files (N1sChecked.mat, electrodes.tsv), its provenance table and its MAT-file reader.

• supportsDistance controls remove DTI–CCEP node-degree coupling in 8 SEEG patients

  Edge-level DTI effect at fixed distance is small; this node-level test agrees: geometry explains the degree coupling.

• usesDistance-stratified CCEP reciprocity was inconsistent in eight SEEG patients

  We used its verified copy of the OpenNeuro ds005448 v1.0.0 input subset (N1sChecked.mat, electrodes.tsv, channels.tsv, PROVENANCE.tsv) and its MATLAB v5 reader code/matv5.py to load the expert-checked N1 labels.

• usesIn 8 SEEG patients, spontaneous coupling added almost no predictive value beyond distance for expert-checked N1 CCEPs

  We used the N1sChecked derivative labels, the distance baseline and the analysis design of this publication (Newton logistic regression, rank AUC, 5-fold CV grouped by stimulation pair, leave-one-patient-out and the exact sign-flip test).

• usesA power law predicts held-out SEEG N1 amplitude better than an exponential law in 8 patients

  It fetched the original ds005448 N1sChecked.mat and electrodes.tsv files; its S3 keys, ETags and MD5 sums are the provenance of the input table.

• supportsIn ds004080, surface N1 CCEP probability falls with distance with no age effect, while stored detections exclude all 202 depth contacts

  Distance to the stimulated pair is also a strong predictor of N1 in a second dataset and center.

• usesDistance accounts for measured SEEG CCEP reciprocity, while white-matter sites send more one-way responses

  We used the archived byte copies of the ds005448 N1sChecked.mat, electrodes.tsv, channels.tsv and participants.tsv files, their PROVENANCE.tsv (S3 ETag = MD5), and the numpy MATLAB v5 reader matv5.py from this package. Its distance result (distance predicts N1, AUC 0.76-0.94) is the baseline model for our reciprocity null and our prediction test.

• usesClinical single pulses did not measurably increase delayed SEEG transients in six patients or show a larger increase at SOZ contacts

  MATLAB v5 reader (matv5.py) for the N1sChecked label files.

• usesIn one SEEG patient, 2 mA pulses preserved most 1 mA N1 responses and added farther responders

  code/matv5.py, the numpy-only MATLAB v5 reader, is copied from this package.

• usesSpontaneous inter-pulse coupling adds at most 0.004 AUC beyond distance for N1 CCEP prediction in 8 SEEG patients

  Uses its 110 ds005448 input files (N1sChecked labels, electrodes, channels, events), the file list with MD5, the MAT reader and the row-building rules.

• usesIn eight SEEG patients, spontaneous coupling did not improve N1 CCEP prediction beyond distance (mean AUC gain 0.0001)

  We used the N1sChecked derivative labels, the distance baseline and the analysis design of this publication (Newton logistic regression, rank AUC, 5-fold CV grouped by stimulation pair, leave-one-patient-out and the exact sign-flip test).

• usesGeometry alone selects SEEG stimulation pairs and targets better than chance, but a measured response map needs 2-3 times fewer pairs (ds005448, 8 patients)

  We use the same distance-to-N1 relation as a leave-one-patient-out logistic model for pair selection.

• usesN1 latency rises with distance at SEEG and ECoG contacts in one dataset

  We used its archived ds005448 inputs (N1sChecked.mat, electrodes.tsv, S3 provenance with MD5) and its numpy MAT v5 reader (matv5.py).

• usesStimulated pair angle did not improve distance-based prediction of N1 CCEPs in eight SEEG patients

  We use its pair-contact table (distance, expert-checked N1 labels, same-shaft flags) from OpenNeuro ds005448 as the input.

• supportsPassive coupling adds little to distance for N1 CCEP prediction in eight SEEG patients

  Same 8 patients: distance stays the main predictor of N1 (CV AUC 0.70-0.92 here), and passive coupling adds only a small amount, as DTI did there.

• usesIn eight SEEG patients, N1 peak latency rose 0.23 ms/mm with distance, but distance explained little variance

  We use the archived ds005448 derivative inputs (110 files, PROVENANCE.tsv with S3 ETags), the numpy MATLAB reader matv5.py, fetch_inputs.sh and the same latency formula. We reproduce its per-patient OLS latency slopes (median 0.45 ms/mm) and show that window-edge peaks and least squares make them larger; robust slopes on the core set have a median of 0.23 ms/mm.

• usesSingle 1–2 mA pulses did not change 70–150 Hz power at 150–500 ms in eight SEEG patients

  matv5.py MAT reader and the archived expert-checked N1 labels (N1sChecked.mat) that define N1-present contacts.

• usesIn eight SEEG patients, selected stimulation sites reduced held-out distant N1 responses versus random sites but matched the nearest responsive sites

  We use its expert-checked N1 pairs table (results/pairs_table.tsv) for the 8 SEEG patients of ds005448. It gives the response matrix for the selectivity analysis.

• usesDistance explains most apparent SEEG CCEP reciprocity in eight patients

  The archived copy of ds005448 v1.0.0 derivatives (*_N1sChecked.mat), electrodes.tsv, the ETag provenance table and the numpy MAT v5 reader (matv5.py) were reused as the input data and loader for this analysis.

• supportsHigh-gamma power added little to distance-based prediction of expert N1 labels in eight SEEG patients

  Our distance-only held-out AUC (median 0.845) agrees with the distance AUC of 0.76-0.94 reported here, and we extend the finding: HG power, like DTI connectivity, adds almost no held-out information beyond distance.

• usesReverse N1 responses showed reciprocity beyond distance and degree in most SEEG patients, but added little predictive AUC

  input files, provenance table and MAT reader

• usesIn 8 SEEG patients (ds005448, 1-2 mA single pulses), short spontaneous between-pulse coupling does not predict expert N1 responses beyond distance and same shaft (held-out AUC gain +0.0007, p = 0.56)

  We use its numpy MAT v5 reader for the N1sChecked labels and its distance baseline for N1 prediction.

• usesFor random held-out N1-map entries in eight SEEG patients, site and contact degree added more AUC than low-rank terms

  Origin of the pairs table: expert-checked N1 labels, distances and row rules from ds005448.

• usesSEEG CCEP reciprocity is mostly explained by distance and site effects; the reverse response adds no held-out prediction (OpenNeuro ds005448, 8 patients)

  We use its verified copy of the ds005448 input files, its provenance table, its download recipe and its MAT v5 reader.

• usesOne-pulse stimulus artifact amplitude (0-2 ms) does not predict expert-checked N1 CCEPs beyond Euclidean distance in 3 STReEF SEEG patients; artifact is at a plateau at most near contacts

  Expert-checked N1 label files (N1sChecked.mat) archived in this package give the outcome labels.

• usesDistance-only models overestimated hippocampal and amygdala N1 responses in 4 of 6 patients, with no consistent group effect

  The ds005448 input files and their S3 provenance come from this package; the matv5 reader and fetch script are reused.

• usesSpearman-Brown predicts 17 pulses for reliable SEEG CCEP maps in eight patients

  We used the copies of the ds005448 expert N1 labels, channel, electrode and event tables, and the MAT v5 reader (code/matv5.py) from this package.