Patient score offsets changed pooled STE HFO AUC from 0.239 to 0.787 with fixed channel ranks in 18 SEEG patients
In this fixed public sample of 18 SEEG patients, patient score offsets changed pooled seizure onset zone (SOZ) AUC for first hour STE high-frequency oscillation (HFO) counts from 0.238818 to 0.786620 while preserving every patient’s channel ranks, ties, and AUC. The original pooled AUC was 0.738237 and the equal-patient mean was 0.718833. The separate-block endpoints are extrema only within that construction, which uses observed labels; they are not limits for all offsets or prediction results for new patients. Analyses used 1,285 channels and 181 SOZ-positive channels from OpenNeuro ds005398 v1.1.3. Results describe released detector marks, not verified neural events. This retrospective audit shows that pooled AUC can depend strongly on between-patient score levels when the target is within-patient channel ranking.
seizure onset zone · high-frequency oscillations · AUC · intracranial EEG · SEEG · patient-level evaluation · channel localization
References
6- uses20 ms detector matches gave an uncertain SOZ rank gain over combined counts in the first hour of 18 SEEG sleep records sci:sha256:e2fb21fc2b8904e6e332f4adef0382d28035541049ea83d70d8da9f15b9adbb1
The unchanged retrieval code has this upstream origin. It checks pinned public metadata, S3 event versions, and EDF header ranges.
- usesPatient mean log STE count subtraction raised mean SOZ prediction loss in 18 SEEG patient exclusion tests at ridge penalty 0.01 sci:sha256:631136a35995fb1ade3b3b76300c98d8686544259d359fb9d8782daa98a9e702
This source package supplied the exact ds005398 event bytes, metadata, source manifest, parser, and retrieval code. We reused no model fits or predictions.
- usesThe interval for added STE HFO rank gain included zero in fixed contact order models for 17 SEEG patients sci:sha256:74196c384bfb7cfbe2317f8fd7c692d13fd50f2b78786436fee6371c62b31fc3
The reused source parser has this upstream origin. It defines the channel selection and first-hour event counts used here.
- usesOffsets from labels changed pooled gamma AUC with separate patient score ranges and fixed local ranks in 35 SEEG patients sci:sha256:11b45c1ef044fd86e4be7997ed8107c67953a2013ab2fd4a40e9c09eb5ce2638
We extend its separate patient score block audit from gamma power to released HFO counts. Its unchanged audit module confirms our decomposition, ranks, and endpoints. We do not claim a new general theorem.
- usesdoi:10.18112/openneuro.ds005398.v1.1.3
The analyzed released event tables and metadata are from this exact dataset version, reported as CC0.
- doi:10.1111/epi.18545
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.