Seizure detection using the phase-slope index and multichannel ECoG

Abstract

Detection and analysis of epileptic seizures is of clinical and research interest. We propose a novel seizure detection and analysis scheme based on the phase-slope index (PSI) of directed influence applied to multichannel electrocorticogram data. The PSI metric identifies increases in the spatio-temporal interactions between channels that clearly distinguish seizure from interictal activity. We form a global metric of interaction between channels and compare this metric to a threshold to detect the presence of seizures. The threshold is chosen based on a moving average of recent activity to accommodate differences between patients and slow changes within each patient over time. We evaluate detection performance over a challenging population of five patients with different types of epilepsy using a total of 47 seizures in nearly 258 h of recorded data. Using a common threshold procedure, we show that our approach detects all of the seizures in four of the five patients with a false detection rate less than two per hour. A variation on the global metric is proposed to identify which channels are strong drivers of activity in each patient. These metrics are computationally efficient and suitable for real-time application.

Fig. 1

Simulation scenario for evaluating sensitivity of PSI to segment length. v[n] and w[n] are uniform [−100, 100] independent random variables, while y[n] is a sequence of Gaussian distributed independent random variables with zero mean and variance 0.1. The bandpass filter of 1–11 Hz passband is a 12th-order IIR filter and the allpass filter is a 3rd order IIR filter, both assuming a sampling frequency of 100 Hz. The allpass filter introduces a nonlinear phase shift between z₁ [n] and z₂ [n].

Fig. 2

Graphical depiction of causal interaction between channels using the PSI metric Ψ_ij with i, j =1, 2,…, 88 during 60-s segments for patient 2. Only values with Ψ_ij > 2 are shown. Yellow indicates smaller PSI values while red indicates larger PSI values. (a) Interictal segment. (b) Ictal segment.

Fig. 3

PSI metric Λ^k [(7) computed with segment length N =20 s] over time plotted with the threshold Γ^k [(9) computed with L =90 and c =8] for 1600 min of data for patient 1. All four seizures were identified and one event is falsely declared as seizure.

Fig. 4

Seizure detection percentage (solid line) and false detections per hour (dashed line) for a range of values of c in (9). Twenty-second segments were used in (4) and L =90 segments (30 min) was set in (9) to calculate the moving threshold Γ^k. (a) Patient 2. (b) Patient 5.

Fig. 5

Shown are two 10 s segments of ECoG representative of PSI false detections, both taken from patient 3 s record. Panel A is an example of abnormal activity that occasionally organizes, but does not rise to the level of a clinical seizure and may therefore be considered a subclinical event. Panel B shows equipment artifact.

Fig. 6

Average delay in detecting seizures using 20-s segments with a threshold Γ^k that employs L =90 (30 min) in (9). Error bars denote one standard deviation. The value of the constant c in (9) was 4.8 for the PSI method and 2.1 for the power method.

Fig. 7

Plot of the average seizure evolution PSI ${\tilde{\Lambda }}_i^l$ defined in (13) using 20-s segment lengths for configurations with at least five seizures. (a) Patient 1. (b) Patient 2. (c) Patient 3. (d) Patient 4: second electrode configuration. (e) Patient 5: first electrode configuration. (f) Patient 5: second electrode configuration. The horizontal long dashed lines are the clinically identified leading channels.