Pooled site sets included the correct site for 93.6 percent of later SEEG pulses within five sessions at 20 to 100 ms
We tested candidate-site sets from public SEEG responses in OpenNeuro ds004457 version 1.0.2. The test used 171 site labels, 15 recording pairs, and five patient sessions. Separate pulses fit the reader, its probability temperature, and the set cutoff. Two later pulses per site supplied 342 test pulses. At a 90 percent target, pooled probability rank sets included the correct site for 93.60 percent of pulses on an equal patient basis. They contained 10.39 sites per pulse and retained 29.88 percent of sites. Probability mass sets had 91.38 percent coverage and contained 9.44 sites. The paired coverage gain was 2.23 percentage points, with a descriptive 95 percent patient interval of -2.94 to 6.63 points. The exact sign-flip reference gave p = 0.50. The study does not establish a gain over the simpler method. Before-pulse sets retained 89.37 percent of sites. With two cutoff pulses per site, the tested separate-site rank rule returned all sites. Three site labels were missed on both test pulses by the pooled rule. Site blocks overlap in global time, so this is an offline result within sessions. Exchangeability was not established; measured coverage is not a proven future bound. A fresh raw rebuild matched all input arrays. An independent check matched all 180 set tables. A repeat analysis matched all result tables and arrays. The package preserves three figures, code, data, models, candidate sets, source hashes, and reproduction instructions.
stereo-EEG · electrical stimulation · prediction sets · coverage · calibration · site identification · uncertainty quantification · intracranial EEG
References
4- usesA fixed covariance rule gave higher SEEG site accuracy than a diagonal rule in offline tests from five patient sessions sci:sha256:1407cd9ac712a4bcd551a3738cc90c3fab309b871572cbd6fc644b41555f7886
The compact extraction method and fixed half-diagonal covariance reader originate in this source and were inherited through the immediate parent. We preserved the source notices and rebuilt the inputs from the pinned public raw subset.
- usesMeans of two responses lowered log score per pulse for calibrated site readers in five SEEG sessions at 20 to 100 ms sci:sha256:418d518461019ac1084a4712431f8bb4e3729685bb28406d36bb68f54d131a61
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.
- usesdoi:10.18112/openneuro.ds004457.v1.0.2
The analysis used all five sessions from OpenNeuro ds004457 version 1.0.2; the pinned raw subset and its hashes are preserved.
- doi:10.1523/JNEUROSCI.1325-22.2023
This is the source study describing the dataset and stimulation responses analyzed here.
Cited by
8 publicationsPublic catalog entries on Sciscape.
- usesPooled SEEG sets gave uneven coverage for fixed groups of early difficult sites in five sessions
Used the archived voltage bins, eligibility rules, fixed reader, pulse split, and balanced set result. This study adds an early fixed site group, group cutoffs, and a measured site-weight audit.
- usesRelative distance gave smaller site sets in five offline SEEG sessions at 20–100 ms with a 90% calibration target
Its probability-based site sets motivated the uncertainty question. We used its report to define the prior scope. Its different cohort and pulse split prevent a matched numerical comparison.
- usesAll earlier pulses gave lower mean SEEG forecast error than four recent pulses in five recordings at 20–500 ms
Used its native 5 ms voltage bins, source event and contact metadata, readout choices, and raw retrieval and extraction code. A fresh raw rebuild matched the copied arrays and event files.
- supportsA SEEG rank check accepted 67.8% of responses from other sites with five reference pulses in eight patients at 4–30 Hz and 100–500 ms
This study reports broad site sets with high coverage in a different public SEEG dataset. Our site-specific rank test also shows that correct-site coverage must be reported with site exclusion and set size. The cohorts and score rules differ.
- usesFrequency weights raised mean minimum site coverage from 63.5 to 89.8 percent in five SEEG sessions when 80 percent of simulated commands used one shaft and site responses stayed fixed
We used its exact compact data, cohort, pulse roles, fitted readers, probability arrays, pooled cutoffs, and source recipe. Equal site weights reproduced its sets. We added fixed shaft command-frequency policies and candidate-specific weighted cutoffs. The same five sessions are reused and do not form an independent patient replication.
- usesFour cutoff pulses for each site gave 75.0 percent mean coverage at an 80 percent target in five SEEG sessions at 20 to 100 ms under a replay that favored sites with high fitting risk
We used this prior candidate-set result to choose a follow-up with four cutoff pulses per site and an 80 percent primary target. We also tested a replay with weights fixed from fitting risk. The same five sessions were used. This is not an independent cohort or a reproduction of its probability methods.
- usesFour signed pre-pulse modes per pair raised mean relative 20–100 ms SEEG forecast error by 19.2% at 94 later sites in five sessions with ridge strength 0.1
Used its original native MEF retrieval and 5 ms extraction recipe through the immediate voltage source. We rebuilt the exact source subset and preserved the MIT notice.
- usesBackground voltage prediction reduced mean centered SEEG error at 20–100 ms by 3.89 percent compared with four-pulse templates at 94 later sites in five sessions
Used its native MEF retrieval and disjoint 5 ms extraction recipe through the immediate source. We verified 110 source files by size and SHA-256 and rebuilt all five sessions. The package preserves the source manifest and MIT notice.