Automated real-time EEG sleep spindle detection for brain-state-dependent brain stimulation

Area of Science:

• Neuroscience
• Sleep Science
• Computational Neuroscience

Background:

• Sleep spindles are key electroencephalographic events in non-rapid eye movement sleep.
• They are hypothesized to play a crucial role in memory consolidation.
• Real-time detection of sleep spindles is challenging but necessary for causal investigation.

Purpose of the Study:

• To develop and validate a robust, real-time algorithm for detecting sleep spindles.
• To enable phase-specific triggering of stimulation during sleep spindles.
• To facilitate research into the causal role of sleep spindles in cognitive functions.

Main Methods:

• Development of a multi-channel electroencephalographic signal-processing algorithm for real-time spindle detection.
• Validation using pre-recorded sleep EEG datasets streamed to a real-time computer system.
• Performance comparison against state-of-the-art offline detection algorithms.

Main Results:

• The real-time spindle detector achieved high Sensitivity (~83%), Precision (~78%), and F1-Score (~81%).
• Detection performance was robust across naps and full nights, largely independent of sleep scoring.
• Detected spindles exhibited typical characteristics in terms of frequency, duration, amplitude, topography, and phase-locking.

Conclusions:

• The developed algorithm provides a reliable method for real-time sleep spindle detection.
• This tool empowers researchers to target sleep spindles for experimental manipulation during sleep.
• Enables new avenues for investigating the causal role of sleep spindles in memory and other cognitive processes.