MSIF-Net: Automated EEG Sleep Staging for REM Sleep Behavior Disorder

An automated sleep-staging framework for REM sleep behavior disorder (RBD) using EEG data. The Multi-Stream Imaging Fusion Network (MSIF-Net) was validated for epoch-level Wake/NREM/REM classification and whole-night REM architecture estimation. The dataset, authored by Huanyu Li and last updated in May 2026, contains results from this validation study.

Use Cases

• Training or benchmarking automated sleep-staging models based on the described MSIF-Net architecture.
• Analyzing correlations between automated REM% metrics and clinical symptom scales like PSQI and RBDSQ.
• Validating EEG-only sleep staging methods against manual polysomnography summaries in a clinical RBD cohort.

Strengths

• Model achieved class-wise F1 scores of 0.84 for Wake, 0.94 for NREM, and 0.80 for REM.
• Whole-night REM% and NREM% showed close agreement with manual clinical summaries.
• Automated REM% demonstrated statistically significant inverse correlations with PSQI and RBDSQ symptom scales.

Limitations

• Dataset scale is limited to 22.9 KB, suggesting it contains summary results rather than raw EEG time-series data.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for certain analyses.

Provenance

Source

figshare

Collection Method

Developed using a Multi-Stream Imaging Fusion Network (MSIF-Net) integrating raw EEG, spectrograms, and handcrafted descriptors, validated on polysomnography data from an RBD cohort.

Freshness

Last updated 2026-05-19 04:21:06