Table 1.
EEG Features.
| Signal Descriptor | References | Brief description |
|---|---|---|
| Complexity features | Degree of randomness or irregularity | |
| Shannon entropy | (22) | Additive measure of signal stochasticity |
| Tsalis entropy (n = 10) | (23) | Non-additive measure of signal stochasticity |
| Information quantity (δ, α, θ, β, γ) | (24) | Entropy of a wavelet decomposed signal |
| Cepstrum coefficients (n = 2) | (25) | Rate of change in signal spectral band power |
| Lyapunov exponent | (26) | Separation between signals with similar trajectories |
| Fractal embedding dimension | (27) | How signal properties change with scale |
| Hjorth mobility | (28) | Mean signal frequency |
| Hjorth complexity | (28) | Rate of change in mean signal frequency |
| False nearest neighbor | (29) | Signal continuity and smoothness |
| ARMA coefficients (n = 2) | (30) | Autoregressive coefficient of signal at (t-1) and (t-2) |
| Continuity features | Clinically grounded signal characteristics | |
| Median frequency | The median spectral power | |
| δ band power | Spectral power in the 0–3 Hz range | |
| θ band power | Spectral power in the 4–7 Hz range | |
| α band power | Spectral power in the 8–15 Hz range | |
| β band power | Spectral power in the 16–31 Hz range | |
| γ band power | Spectral power above 32 Hz | |
| Standard deviation | (31) | Average difference between signal value and it's mean |
| α/δ ratio | (14) | Ratio of the power spectral density in α and δ bands |
| Regularity (burst-suppression) | (14) | Measure of signal stationarity/spectral consistency |
| Voltage < (5, 10, 20 μ) | Low signal amplitude | |
| Diffuse slowing | (32) | Indicator of peak power spectral density <8 Hz |
| Spikes | (32) | Signal amplitude exceeds μ by 3σ for 70 ms or less |
| Delta burst after spike | (32) | Increased δ after spike, relative to δ before spike |
| Sharp spike | (32) | Spikes lasting <70 ms |
| Number of bursts | Number of amplitude bursts | |
| Burst length μ and σ | Statistical properties of bursts | |
| Burst band powers (δ, α, θ, β, γ) | Spectral power of bursts | |
| Number of suppressions | Segments with contiguous amplitude suppression | |
| Suppression length μ and σ | Statistical properties of suppressions | |
| Connectivity features | Interactions between EEG electrode pairs | |
| Coherence – δ | (14) | Correlation in 0–4 Hz power between signals |
| Mutual information | (18) | Measure of dependence |
| Granger causality – All | (33) | measure of causality |
| Phase lag index | (34) | Association between the instantaneous phase of signals |
| Cross-correlation magnitude | (35) | Maximum correlation between two signals |
| Cross-correlation – lag | (35) | Time-delay that maximizes correlation between signals |
The 58 EEG features fell into three EEG signal property domains: Complexity features (25 in total), Category features (27 in total), Connectivity features (six in total).