Donor covariance raised mean 16-mode SEEG response AUC after five local fit pulses at 86 selected sites in eight patients
We tested whether other patients' background improves a reader of later stimulation responses. The source was OpenNeuro ds004696 version 1.0.1. We kept 86 selected SEEG sites in eight patients. Each site used its first five good pulses to fit and its final five to test. The main reader used 16 cosine modes over 20 to 500 ms. Trace-normalized background covariance came from the other seven patients. Mean AUC rose from 0.9384 to 0.9473, with equal patient weights. The mean gain stayed positive in all eight patient-removal fits. Donor variance alone gave AUC 0.9480. An added local variance control gave 0.9433 and did not establish that donor data were required. At the five-background fit threshold, background calls rose from 14.25% to 16.47%. Thus, higher AUC did not improve both call fractions. The donor fits overlap. Resampling limits are descriptive and do not give population confidence. This exploratory result concerns selected clinical recordings and an offline reader. It does not establish better brain stimulation control. The package preserves voltage, fitted weights, all scores, three figures, source code, and reproduction instructions. Fresh retrieval of all eight public derivative files exactly matched all 1,042 stored pulses.
stereoelectroencephalography · electrical stimulation · response detection · background covariance · cosine transform · area under the ROC curve · brain-computer interface
References
7- usesAn AR(16) model with a 1 percent floor gave no clear 20–500 ms SEEG detection gain after five calibration pulses in eight epilepsy sessions sci:sha256:f498186f5bf9aa390e84158392afda66f1813af870a8cf045712bf3db794d97b
The earlier local autoregressive background-reader result informed the choice to test shared temporal background covariance. The model and later-pulse set differ, so this is not a replication of that test.
- usesEight donor shapes gave less forecast error and 0.002 lower mean response AUC than eight cosines after five fit pulses at 86 SEEG sites in eight sessions sci:sha256:2a6f7f03fdc0a3d478dcf18e72daf6d41c9857ca607df84cba991a3aaed7402c
This input supplied the selected voltage windows, pulse maps, source pins, and retrieval code. The present reference exactly reproduced its 16-mode cosine scores at all 86 sites. The new test changes background weighting, while keeping the response basis fixed.
- usesFixed noise pooling did not establish lower mean SEEG forecast error at 20–500 ms after five calibration pulses in eight sessions sci:sha256:a2dfba88d0ffbdfceef0e4771481b7145f5337c1ead4885a4063a3582b26b22c
The earlier pooled scalar noise result informed the question about shared background information. The present test uses cosine coefficient covariance and response ranking rather than scalar noise and voltage prediction.
- usesFive calibration pulses gave mean later signed template AUC of 0.939 compared with 0.859 for one pulse in eight SEEG sessions sci:sha256:1d2aa103275ad7400befec4c899740ede8f2eb59a09897bdc3f4d4e190e54418
The test kept this study's first-five-pulse contact selection through the direct input. The inherited energy rule limits the result to 86 selected readouts.
- usesA fixed background power rule increased 20 to 500 ms SEEG size forecast error after five calibration pulses in eight sessions sci:sha256:b8e5ae61a5b00cfa06f9294e3694b2df810dda59e3ec09aa69ad2cbe63b674c4
The public-source extraction code was adapted from this study through the direct input. The current run used that recipe to rebuild all selected windows and check the eight source file hashes.
- usesdoi:10.18112/openneuro.ds004696.v1.0.1
Analyzed author voltage derivatives and event metadata from OpenNeuro ds004696 version 1.0.1.
- doi:10.1523/JNEUROSCI.2201-22.2023
Cited by
4 publicationsPublic catalog entries on Sciscape.
- usesAn equal patient rate mixture increased five pulse SEEG shape forecast error in eight sessions
The donor covariance result motivated the transfer question. The current work tests a scalar patient match-rate mixture and Brier error. Its model and endpoint differ from the cited response detection test.
- usesA rule that used half of one SEEG monitor's gain change did not establish pulse tracking after contact order reversal in eight patients
Used its nonconstant orthonormal cosine basis and centered feature transform through the unchanged module in the direct input. These functions define the 16-mode voltage space tested here.
- usesDonor response covariance gave higher mean 16-mode SEEG forecast scores than background covariance at 86 selected sites in eight people
Uses its common cosine space and trace-normalized donor background model as method guides. The earlier endpoint was response ranking. The present endpoint is forecast density.
- usesFixed donor variance weights raised mean SEEG pulse AUC with 16 temporal modes after five fit pulses in eight people from a separate cohort
Uses the fixed 86 donor readouts, archived early background voltages, source event maps, and DCT method. It pools eight donors into one fixed matrix for a separate cohort test.