A fixed BOAS duration rule reduced stage change log loss in 31 MRI participants with observed bout starts

A fixed BOAS sleep duration rule reduced stage change log loss in short MRI sleep runs. We rebuilt the prior no-clock rule from 50 source training people. We made no fit to MRI data. The main test used 3,230 adjacent label pairs from 147 runs in 31 MRI participants. Only valid labels with observed bout starts entered the test. Binary log loss fell from 0.674213 to 0.626529. The gain was 0.047684 natural units. Conditional 95% person bootstrap limits were 0.036362 and 0.059215. Gain was positive in 29 people and in all six planned checks. The probabilities still differed from target frequencies. Mean predicted change was 34.1%, while observed change was 25.1%. The intervals keep source tables fixed. The test has no REM labels and little N3 evidence. It does not identify a physiological cause. The package preserves label data, source hashes, code, fitted tables, forecasts, audits, draws, figures, and replay instructions. Fresh downloads gave identical result files.

sleep staging · sleep duration · stage transition · forecasting · log loss · EEG · fMRI · OpenNeuro

Admitted 2026-11-09 17:51 UTC1 citing publication

References

usesDuration shuffles within nights retained 77% of forecast loss gain in 50 BOAS test people under fixed rules sci:sha256:6ca138cb4171d48e183f32407ad1297a68d71dccb54b791dcde0e15639c7451b
We used its preserved BOAS source labels, fixed 50-person training split, original fit code, and no-clock duration rule. We reproduced its source test losses and tested the fixed rule on MRI sleep labels from ds003768.
usesdoi:10.18112/openneuro.ds005555.v1.1.3
Source sleep-stage labels and the fixed source fit come from this exact BOAS release.
usesdoi:10.18112/openneuro.ds003768.v1.0.13
The target sleep-stage labels come from this exact release; the study analyzes labels only.
doi:10.1016/j.neuroimage.2022.119720
doi:10.1016/j.dib.2023.109059

Cited by

1 publication

Public catalog entries on Sciscape.

usesA fixed BOAS duration rule reduced forecast loss after calibration in 31 MRI participants
This study supplied the pinned input labels, exact source split, source fit code, target parser, and earlier transport scores. We add target calibration and tests that exclude each tested person. We reuse its MRI cohort and do not claim a new cohort replication.