Pipeline EEG

• Acquisition
• Pre-processing
• Spindle detection
• Spindle analysis
• Reconstruction

Acquisition (CoRe)

• Saturation of extra ECG fixed
• Checking and saving impedance measures before and after each task- and sleep-EEG recordings -> quality of EEG signal for later analysis
• Size and reference of used EEG caps saved -> building cap templates for each size using database from Paris

Preprocessing:

MRI + registration

• Freesurfer
• Non-linear registration T1 to MNI152
• Inverse warp a cap template to subject space

EEG pre-processing

• Gradient artifact removal (FASTR -  Niazy et al., 2005, NeuroImage)
• ECG QRS-peak detection: automatic channel selection
• Downsampling (250Hz)
• BCG correction (FASST; ICA-based algorithm)
• Bandpass filter (0.5-25Hz)
• Re-referencing to averaged mastoids (M1-M2)
• Extra Fast Kernel ICA (removal of movement-related artifacts)
• Bad interval detection

• ​Automatic script (Matlab):
  • Wavelet-based algorithm (Warby et al. 2014 Nature Methods) ​Github: https://github.com/Mensen/swa-matlab
  • Detect Spindles
  • Remove spindles detected during Bad Intervals
  • Add sleep scoring to spindle detection output
  • Write VMRK (marker file - brainvision)
    • Extract specific spindles depending on sleep stage
    • Add sleep scoring
  • Github: https://github.com/arnaudbore/spindlesDetection

Spindle detection

Matlab structure

•     Ref_Region: electrodes
•     Ref_TypeName: Fast or Slow spindle
•     Ref_Start: Beginning of the spindle
•     Ref_End: End of the spindle
•     Ref_NegativePeak
•     Ref_PositivePeak
•     Ref_Peak2Peak: spindle amplitude
•     Ref_Length: spindle duration
•     Ref_NumberOfWaves: determination of spindle frequency
•     scoring: Sleep Stage

Spindle detection - Outputs

Spindle detection - Outputs

Manual validation

Spindle analysis

• Spindle features
  • Type (slow vs. fast, NREM2 vs. NREM3)
  • Number
  • Amplitude
  • Duration
  • Frequency
  • Density
• Topographic distribution and propagation
• Connectivity (coherence & phase-locking )
• Hemispheric differences in spindle features and power (e.g., C3-C4; Nishida & Walker, 2007, PloSOne)

Reconstruction

• Brainstorm  (Tadel et al., 2011, Comp. Int. & Neuro.)
• BEM head model
• Import EEG epochs based on spindle markers (-600 to 2000ms)
• Remove DC offset (-600 to -100ms)
• Noise covariance estimation
• Bandpass filtering (spindle range: 11-17Hz)
• Source estimation (minimum norm)
• Signal projection and power estimate
• Dynamic visualization (movie)
• Deep sources?

Single spindle-event detected on Pz

Time-frequency maps (detected on Pz)

Pz

Fz

C4

% increase in the specific frequency band

Time-frequency maps (detected on Pz)

% increase in the delta frequency band

% increase in the spindle frequency band

Connectivity - Coherence (detected on Pz)

Pz - spindle band

Fz - spindle band

Fz - theta band

Pz - theta band

Connectivity - Coherence (detected on Pz)

Pz - spindle band

Fz - spindle band

Fz - theta band

Pz - theta band

T7 - spindle band

All channels