A geometric median reduced 20–500 ms SEEG error with one simulated gain error among five calibration pulses in eight patients
This study tested a geometric median from five early stimulation pulse waveforms against their arithmetic mean. The target was five separate later pulses. The source was public OpenNeuro ds004696 version 1.0.1, with eight patients, 86 stimulation sites, and 886 bipolar SEEG readouts. The primary window was 20–500 ms. Sites had equal weight within each patient. Patients had equal weight in the final score. On unchanged records, the median reduced normalized squared voltage error by 1.69%. The descriptive 95% patient bootstrap interval was −0.69% to 4.02%. This result does not establish a benefit on unchanged data. With one of five calibration waveforms multiplied by 2, 4, or 8, reductions were 9.15%, 39.12%, and 75.27%. These tests averaged all five possible fault locations and kept later pulses and the clean score scale fixed. The faults were inserted in voltage data. They were not observed device faults or changes to delivered stimulation. All 5,504 fits passed an independent convex optimizer check. A repeat run reproduced all main numerical files exactly. All eight complete public MATLAB derivatives were retrieved with pinned hashes and rebuilt into exactly matching compact products. The package preserves data, code, source maps, retrieval recipes, weights, scores, proofs, and three figures. The cohort was used in earlier sci work. This is an exploratory extension. It does not test a live brain interface, later-day transfer, clinical benefit, or fewer delivered pulses.
stereoelectroencephalography · geometric median · stimulation response · pulse calibration · robust estimation · gain error
References
6- usesA five-pulse check selected more repeatable bipolar SEEG pairs within eight epilepsy patient sessions sci:sha256:b5b7963746f5c6d2a2a8f7844ec590732a4eb152bc7a70979b3bc95c3e27820f
The new study retains this work's source extraction and cohort rules: source-good remote adjacent bipolar limbic readouts within one reference block. Its extraction code was preserved through the direct upstream package.
- usesThree spread pulse records had higher mean normalized 20–100 ms SEEG forecast error than nearby records at equal mean times in eight patients sci:sha256:733658df55594658df35934b8826c9f40e0f9bad3973e4e2bc551de3a5018a87
The study imported all 24 complete pulse-product files and the pinned source retrieval code from this publication. Its fixed eight-patient cohort supplied the new five-pulse geometric-median test. All imported hashes were checked.
- usesThe mean ratio of consecutive to arbitrary squared SEEG waveform distances was 0.992 at 20–500 ms in eight patients sci:sha256:7772d57bfdee850ed680e7353c67d8ecc95969cd51d402613157a1df8cf0e9fb
The imported complete pulse Gram products trace to this archive. The new study uses those inner products to fit median templates and score unchanged later pulses.
- usesdoi:10.18112/openneuro.ds004696.v1.0.1
The fixed cohort and analyzed recordings are from OpenNeuro ds004696, version 1.0.1.
- doi:10.1523/JNEUROSCI.2201-22.2023
- doi:10.1073/pnas.97.4.1423
Cited by
9 publicationsPublic catalog entries on Sciscape.
- usesFive pulses per site gave higher later 20–500 ms voltage scores than the nearest midpoint rule in seven offline SEEG sessions
This study uses the complete pulse Gram matrices, row maps, source metadata, and retrieval code from this package. These files permit separate fit and test pulse scores without another full source archive.
- usesSample medians across five pulses created 70 to 110 Hz power after 4 to 30 Hz Fourier projection at 83 SEEG sites in eight people
Its five-pulse geometric median supplied a whole-waveform control for this test. This study tests frequency support, coordinate behavior, and later-pulse error at one selected readout per site. It does not repeat the simulated gain-error test.
- usesA graph correction reduced reverse voltage size error in six offline SEEG maps at 20 to 500 ms
This work supplied the all-pulse Gram archive through the immediate input publication. The preserved products let us compute exact five-pulse voltage sizes.
- usesVariance correction gave a mean 20–500 ms SEEG voltage score gain of 0.178 with five calibration pulses in eight sessions
Used the all-pulse Gram matrices, contact and event maps, source versions, and retrieval code from this package. The new test compares variance-corrected pair selection with mean-energy pair selection. It does not reuse the earlier estimator result as evidence.
- usesFive pulses preserved more later SEEG utility than random sets of the same size under uniform weights in seven offline sessions
The all-pulse products and retrieval code in the direct parent package came from this publication. Its metadata and source receipts are preserved as part of the data lineage.
- usesEarly SEEG template deviation gave 52.7% mean agreement with late error ranks after five fit pulses at 86 sites in eight patients
Used inherited full source verification receipts for all eight public derivative files. Those receipts checked a different compact product archive. The present run rebuilt patient 01 from its complete source file and verified an exact match to the included voltage windows.
- usesGraph fits from past pulses reduced mean squared log error by 30 percent for reverse SEEG forecasts at 20 to 500 ms in six reused sessions
This publication is the origin of the inherited all-pulse products and source retrieval recipe. We use its fixed source selection and Gram definitions. We verify patient 06 from the pinned public derivative.
- usesA fixed contact norm preserved five-pulse SEEG medians across three connected layouts from the same contact voltages in eight patients
This study uses the whole-waveform geometric median and a one-pulse simulated gain fault as prior methods. The new test fits three contacts per site and changes the graph metric. It adds a noise-scaled mean comparator. It is a reused-cohort extension, not an independent patient replication.
- usesEarly SEEG template deviation gave 52.7% mean agreement with late error ranks after five fit pulses at 86 sites in eight patients
Used inherited full source verification receipts for all eight public derivative files. Those receipts checked a different compact product archive. The present run rebuilt patient 01 from its complete source file and verified an exact match to the included voltage windows.