A scale with template error reduced log loss against T/4 for four averaged SEEG responses in five patients
We tested probability scales for four averaged SEEG responses in five patients with epilepsy. The public data came from OpenNeuro ds004457, version 1.0.2. Three pulses per site set each template. A fourth pulse set the single-response scale, T. A Gaussian model with finite template error gives 7T/16 for four responses. Against T/4, this fixed rule reduced log loss in all five patients. The equal-patient mean fell from 2.891 to 2.592 bits per group. All site predictions stayed the same. The 30–100 ms window and probabilities without the uniform mixture gave the same direction of change. Keeping T had lower loss in three patients. The small selected site bank also favored T on mean loss. The study reused test data and used a retrospective split. It did not test live control or meaningful perception. The package preserves figures, probabilities, code, input data, source hashes, and completed reproduction checks.
SEEG · intracranial EEG · electrical stimulation · probability calibration · log loss · brain-computer interface · template uncertainty · repeated responses · covariance
References
5- usesA joint SEEG shape decoder did not show a clear mean gain over one selected pair in four patient sessions sci:sha256:81f77d4f52365d5b14fed7d3d6db7b430a2cf478b9a6a6062540c4e90bbd8e7c
This study used the original MIT retrieval, MEF extraction, and data-reading code through the cited reader package. The new raw rebuild matched all archived input arrays and event tables.
- usesA fixed covariance rule gave higher SEEG site accuracy than a diagonal rule in offline tests from five patient sessions sci:sha256:1407cd9ac712a4bcd551a3738cc90c3fab309b871572cbd6fc644b41555f7886
This study used the shared half-full, half-diagonal covariance rule and inherited contact and event eligibility definitions.
- usesEight selected sites raised accuracy in all five SEEG sessions but raised log score in four sci:sha256:6ea14d902f7540805304f1e86e0d9539a65d008e19a638b20889ef52c0a3d361
This study reused the voltage means, source metadata, pulse split, site selection, and single-response reader. It added fixed scale comparisons for repeated responses and reproduced the original single-response scores.
- usesdoi:10.18112/openneuro.ds004457.v1.0.2
The study used OpenNeuro ds004457 version 1.0.2; its pinned retrieval recipe and source hashes are preserved.
- doi:10.1523/JNEUROSCI.1325-22.2023
Background source study reporting the electrical stimulation and SEEG recordings used in this analysis.
Cited by
3 publicationsPublic catalog entries on Sciscape.
- Student t scores reduced mean probability loss with simulated fourfold voltage gain in five SEEG tests at 100–500 ms
[Legacy relation: reproduces] Repeats the finite-template versus exact-template Gaussian probability contrast in a cohort that excludes recording shafts and at 100 to 500 ms. This is a restricted-cohort repeat, not an exact replay of the prior full-cohort test.
- usesFive time templates reduced SEEG site log loss with simulated grid errors up to 10 ms in five patients
This probability-scale study informed the use of three template pulses, one separate scale pulse, and later test pulses. Our new test keeps individual test pulses and compares rules for uncertain pulse times.
- usesFinite template scaling increased pulse use without reducing site error at a 90% threshold in five offline SEEG tests
This work uses the exact voltage inputs, pulse split, reader helpers, and finite template scale. It reproduces the fixed-four probability loss and extends the analysis to repeated-pulse stopping.