SleepSEEG: automatic sleep scoring using intracranial EEG recordings only.

Area of Science:

• Neuroscience
• Sleep Medicine
• Computational Biology

Background:

• Traditional sleep scoring relies on scalp EEG, EOG, and EMG, which are challenging in intracerebral studies.
• Automatic sleep scoring is crucial for understanding sleep disorders, epilepsy, and their interactions.

Purpose of the Study:

• To develop and validate an automatic sleep scoring algorithm using only intracranial EEG (iEEG).
• To enable sleep analysis in patients with intracranial electrodes, particularly during epilepsy presurgical evaluation.

Main Methods:

• Developed an algorithm using oscillatory and non-oscillatory spectral features from iEEG data.
• Employed unsupervised channel clustering followed by a two-step classification (multiclass tree, then binary trees for N1).
• Tested the algorithm on 11 patients, combining channel-wise classifications for epoch-level scoring.

Main Results:

• Achieved 78% overall agreement with human expert scoring on the test set.
• Demonstrated excellent performance for Wake (W), N2, and N3 sleep stages, and good performance for REM (R) sleep.
• Showed performance for stage N1 comparable to scalp EEG-based algorithms, with high confidence epochs identified (>80% specificity).

Conclusions:

• The iEEG-based automatic sleep scoring algorithm is effective for long-term recordings without traditional polysomnography signals.
• This tool facilitates sleep evaluation in clinical epileptology and neuroscience research.
• It enables hypothesis generation regarding localized sleep aspects in patients with intracranial monitoring.