Interpretation of digital EEG
Ki-Young Jung, MD, PhD Dept. of Neurology, Korea University Medical Center, Korea University College of Medicine, Seoul, Korea KU Computational Neuroscience Research Lab (KUCNL) http://eeg.re.kr KEP June 7, 2012
Digital EEG instrument Memory
Filter G1 G2
Amplifier
ADC
Microprocessor
Monitor
digitization
Pen write unit
archives
Software: record & interpretation
Analog-to-digital conversion
Resolution (bits) Sampling rate (Hz)
Analog EEG
Structure of digital EEG data • A matrix consisting of – Columns #: Number of channels – Rows #: sampling rate (Hz) * time (msec)
Aliasing signal • In order to reliably digitize a signal of a given frequency, F, the signal must be sampled at a rate of more than two times (Nyquist theorem) • False representation of the source signal occurs when a signal of a certain frequency is sampled too slowly to resolve it frequency content aliasing (i.e., false name) • ACNS guideline (2006) – Acquisition of EEG data onto a digital storage medium should occur at a minimum sampling rate three times the high-frequency filter setting (e.g. 200 Hz for 70 Hz high frequency filter) – Digitization should use a resolution of at least 11 bits per sample be able to resolve EEG down to 0.5 uV
Advantage of DEEG • • • •
montage reformatting, virtual reference Flexible setting of gain, filter, and time base Easy measurement for amplitude and duration Convenient data storage and retrieval, networking
• • • • •
Voltage topography Power spectrum and map Long-term monitoring by QEEG Spike and seizure detection Source localization: dipole, current distribution
Referential recording • Digital EEG machine use the principle of “referential recording” (cf. analog EEG using bipolar recording) – a separate electrode (called reference) is used as the second input into the differential amplifier for each channel (Fp1-ref, F7-ref, T7-ref, etc) – Allow users to convert montage simply by subtracting or adding channels appropriately (montage reformatting) • Bipolar Fp1-F7 = (Fp1-ref) – (F7-ref) • virtual reference: linked ears (A1A2), common average reference (AVR)
Digital filtering • Digital filter do not cause any delay in the signal • EEG removed by analog filters during acquisition cannot be recovered • Typical filter sequence for routine EEG review – 1) use broad-pass analog filter setting for acquisition (scalep 0.1-70 Hz, intracranial 0.1-100 Hz) – 2) Review EEG initially using broad-pass digital filter (0.370 Hz) – 3) apply notch filter (60 Hz) if line noise present – 4) brief apply HFF as needed during clinical events – 5) Apply continuous low and high filter only for extremely agitated or uncooperative patients
Sources of EEG • Current sources – Synchronous excitation of brain tissue
• Volume conduction – Current flows trough brain tissues
• Measurement by electrode – Potential difference
Characteristics of EEG signal • Intracellular current by neuronal interaction • Synaptic potential of neurons (pyramidal cell, glial cell, interneurons) voltage change over time: time domain • Oscillation of neuronal assembly and their networks (short and long association fibers) oscillatory characteristics: frequency domain • Multichannel recording from multiple brain regions: spatial domain
Computational analysis
Quantification Feature extraction Prediction Classification Noise reduction
Common methods of EEG analysis • Time series analysis – Linear analysis: correlation, prediction – Nonlinear analysis: dimension, entropy, nonlinear prediction • Frequency analysis – Power spectral density – Time-frequency analysis • Spatial analysis – Topographic map: voltage, power spectrum – Functional connectivity: coherence, synchrony – Source localization: current source, oscillatory source
Application of EEG analysis • Quantitative comparison of functional brain state – Comparison between patient and healthy control – Correlation between clinical disease stage and EEG features – Effect of CNS acting drugs, Depth of anesthesia – Classification or characterization of CNS drugs – Continuous EEG monitoring • Spike detection, Seizure detection and prediction • Topographic mapping and source localization – Interictal, ictal activity – Specific ERP components • EEG modulation by internal or external stimulation • Brain-computer interface (BCI)
EEG localization principles A-B A-E B-C B-E C-D
C-E
D-E
D-E
Negative
Positive
Bipolar montage
A>B>D>E>C Referential montage
Radial vs. Tangential sources
Dipole
J Clin Neurophysiol 2007
Bipolar - Longitudinal, Transverse chain Referential – Midline, Contralateral, Common average
JKSCN 2003
Potential distribution (voltage topography) • Regardless of point of reference, the features of the terrain do not change (reference independent) – Shape of isopotential contour lines – Voltage gradients (voltage slopes) – Topographic distribution of negative and positive peak • Different voltage distribution indicates different source in the brain
Assuming source from voltage topography • 1) find two voltage field maxima (i.e., negative and positive) • 2) draw a three-dimensional line between two maxima identify the orientation of the field • 3) amplitude and gradient of the field maxima determines the source location depth of the source • 4) the source should be proportionately closer to the field maxima of greater amplitude
Methods for source localization • Model-independent – Voltage topographic map – Current source derivation (Laplacian) • Model-dependent – Cortical potential image – Source localization • Discrete method • Distributed method
Frequency analysis • Delta, theta, alpha, beta, gamma • Frequency and brain state – Sleep and wakefulness – Level of consciousness • Frequency and cognition – Brain maturation – Intelligence – Higher cortical function
Frequency band Delta (0.5-4 Hz) Theta (4-8 Hz)
Function Signal detection, Decision making Selective attention, Orienting, working memory Alpha (8-13 Hz) Sensory, Movement, Memory Beta (13-30 Hz) Movement, Response inhibition Gamma (30-50 Hz) Early sensory, Attention, Perceptual switching Infraslow (