Means of two responses lowered log score per pulse for calibrated site readers in five SEEG sessions at 20 to 100 ms

We tested whether two responses to one stimulation site improve a calibrated reader enough to raise its probability log score per pulse. We used 171 sites in five SEEG sessions from OpenNeuro ds004457 version 1.0.2. Three early pulses per site fit each template. Two other early pulses set probability scales. The last four pulses supplied single responses and two non-overlapping response means. At 20 to 100 ms, equal-patient accuracy was 39.82% for one pulse and 45.96% for a two-response mean. Mean log score was 2.235 bits per pulse for one response and 1.238 for the mean. The paired change was -0.996 bits per pulse, with a descriptive 95% patient bootstrap interval of -1.248 to -0.799. The schedule alone predicted site labels with 81.55% mean accuracy. These results describe offline readers in block sessions. They do not establish brain-channel capacity or a causal neural site code. The package includes compact data, source hashes, code, four figures, fitted probabilities, and reproduction checks.

stimulation response · SEEG · site classification · probability log score · pulse cost · offline reader

Admitted 2026-11-09 19:24 UTC5 citing publications

References

usesA joint SEEG shape decoder did not show a clear mean gain over one selected pair in four patient sessions sci:sha256:81f77d4f52365d5b14fed7d3d6db7b430a2cf478b9a6a6062540c4e90bbd8e7c
The inherited MIT retrieval, extraction, and read-data code traces to this publication. We preserved its license and credit.
usesA fixed covariance rule gave higher SEEG site accuracy than a diagonal rule in offline tests from five patient sessions sci:sha256:1407cd9ac712a4bcd551a3738cc90c3fab309b871572cbd6fc644b41555f7886
We used its hash-pinned MEF retrieval and extraction recipe and fixed half-full covariance rule. A fresh raw extraction matched the compact inputs exactly.
usesEight selected sites raised accuracy in all five SEEG sessions but raised log score in four sci:sha256:6ea14d902f7540805304f1e86e0d9539a65d008e19a638b20889ef52c0a3d361
We used its compact voltage bins, source metadata, and fitted-reader definitions. We used a new nine-pulse rule, two calibration pulses, and a one-response versus two-response pulse-cost comparison.
usesCalibration chose a mean 56 ms endpoint with no lower observed eight-site SEEG accuracy in five offline sessions with input from 20 ms sci:sha256:f55500da2fb76bf20ee3da626763e981457ad648bb7a742feeb6cb5b1f1caf59
Its reported time-only result in these block sessions motivated our added nearest-calibration-time control. We implemented and tested the control on our full eligible site set.
usesdoi:10.18112/openneuro.ds004457.v1.0.2
The analysis used OpenNeuro ds004457 version 1.0.2; its DOI and CC0 license are recorded in the source metadata.
doi:10.1523/JNEUROSCI.1325-22.2023

Cited by

5 publications

Public catalog entries on Sciscape.

usesA fixed Gaussian template correction reduced four-pulse site log loss in five offline SEEG sessions at 20 to 100 ms
Its pulse-cost analysis motivated reporting the number and duration of responses used by an episode. Our test concerns probability scale with an unchanged site decision. The site set and fitting budget differ, so we make no direct score comparison with that report.
usesA common time split supported at most four stimulation sites in five SEEG sessions with four calibration and four test pulses per site
This publication supplied the full compact event arrays, pulse times, input validity flags, source metadata, and pinned raw source manifest. We used those inputs for the common time split audit and verified them with a fresh raw extraction. We applied an eight pulse site rule to its full saved events rather than its nine pulse reader cohort.
usesPooled site sets included the correct site for 93.6 percent of later SEEG pulses within five sessions at 20 to 100 ms
We used the exact compact voltage inputs, source metadata, retrieval and extraction code, fixed contact and site rules, and Gaussian reader. We split the last four pulses into two cutoff pulses and two test pulses, then measured candidate-set coverage and size. The patient cohort is shared. A fresh raw reconstruction matched the compact inputs.
usesA fixed uncertainty rule gave no clear gain over random repeat requests for known neighboring sites in five offline 6 mA SEEG sessions
The prior two-response comparison helped define this test. The present test asks whether a fixed uncertainty rule selects useful repeats at the same repeat count as random requests. The recordings overlap.
usesA fixed uncertainty rule gave no clear gain over random repeat requests for known neighboring sites in five offline 6 mA SEEG sessions
The prior two-response comparison helped define this test. The present test asks whether a fixed uncertainty rule selects useful repeats at the same repeat count as random requests. The recordings overlap.