Temporal and Spatial Dynamics of Different Interictal Epileptic Discharges: A Time-Frequency EEG Approach in Pediatric Focal Refractory Epilepsy

Abstract

Objective: Characterization of the spatial and temporal dynamics of interictal epileptic discharges (IED) using time-frequency analysis (TFA) and electrical-source localization (ESL). Methods: TFA was performed on IED (spikes, spike waves, and polyspike waves) recorded by high-density-EEG (HD-EEG) in 19 refractory focal epileptic children. Temporal modulations related to IEDs were analyzed in a time window around the IED peaks [-1,000 to 1,000 ms]. Spatial modulations were analyzed by ESL in the time-frequency and time domains. Results: IED were associated with complex power spectral modulations. We observed increases in power spectrum (IPS) patterns specific to IED type. For spikes, the TFA pattern consisted of an IPS (-100 to +100 ms, 4-50 Hz). For spike waves, the IPS was followed by a second IPS (+100 to +400 ms, 4-10 Hz), corresponding to the slow wave. IPS patterns were preceded (-400 to -100 ms, 4-40 Hz), and followed (+100 to +400 ms) by a decrease in the power spectrum (DPS) (n = 8). For 14 out of 19 patients, at least one ESL method was concordant with the epileptogenic area. For the remaining five patients, all of them had temporal epilepsies. ESL in the time-frequency domain (DPS/IPS) provided concordant (n = 6) or complementary (n = 4) information to the ESL in the time domain concerning the epileptogenic zone. ESL in time-frequency domain (DPS/IPS) was the only method to provide concordant information concerning the epileptogenic zone in three patients. Significance: TFA demonstrates complex time-frequency modulations of the neuronal networks around IED, suggesting that the pathological mechanisms are initiated well before onset of the classical hyper-synchronization of the IED. Combining time and time-frequency analysis of the ESL provides complementary information to define the epileptogenic zone in refractory focal epilepsy.

Keywords: electrical source imaging; high-density EEG; interictal epileptic spikes; refractory focal epilepsy; time-frequency analysis.

Figure 1

Diagram summarizing the various steps of the study (i.e., time-frequency analysis processing and source localization in time and time-frequency domains).

Figure 2

Results of the time-Frequency analysis (TFA) according to interictal epileptic discharge type and under control conditions. (A) Typical IPS patterns specific to IED type. For spikes, the TFA pattern consisted of a single IPS (−100 to 100 ms, 4–50 Hz, Patient 17). For spike-waves, this IPS was followed by a second IPS in the range of 4–10 Hz (100–400 ms), corresponding to the slow-wave, resulting in a “boot shape” (Patient 5). For polyspike waves (Patient 2), successive IPS, corresponding to the successive spikes were followed by a boot-shaped pattern concomitant with the slow-wave. A decrease in the power spectrum (DPS) symmetrically surrounded the spike-wave peaks and related IPS (−400 to 400 ms, 4–40 Hz) was observed (patient 5). (B) Control condition, vertex spike waves. An increase in spectral power concomitant to the vertex sharp wave in the time domain was observed (−100 to 200 ms, 4–40 Hz) without any other power spectral perturbations. (C) Island of atypical TFA patterns. IPS related to IEDs were preceded by a localized IPS pattern (pattern 1) (−200 to −100 ms, 15–25 Hz, Patient 6). IPS related to IEDs were associated with isolated IPS islands (pattern 2) (−100 to 100 ms, 100–150 Hz), likely corresponding to high-frequency oscillations (HFO) (Patient 4). The red lines corresponded to the baseline periods.

Figure 3

Results of the time-Frequency analysis (TFA) according to the baseline period considered. (A) Decrease in the power spectrum (DPS) surrounding IEDs. DPS symmetrically surrounded the IED negative peaks and related IPS (−400 to 400 ms, 4–40 Hz) was observed independently to the baseline period considered [−2,000 ms; 1,500 ms], [−1,000 ms; −500 ms] (patient 4) (p < 0.001). (B) Typical IPS patterns related to spike-wave. Independently to the baseline period considered, a typical IPS related to the first negative peak of the spike was followed by a second IPS in the range of 4–10 Hz (100–400 ms), corresponding to the slow-wave, resulting in a “boot shape” (Patient 3) (p < 0.001). The red lines corresponded to the baseline periods.

Figure 4

Concordance between ESL in the time and time-frequency domains and comparison with the epileptogenic sub-lobar area. ESL results were first compared between IPS and DPS (A) in the time-frequency domain and then spatial concordance with ESL in the time domain was analyzed (A,B). Finally, ESL in the time and time-frequency domains were compared to the epileptogenic sub-lobar area (C). IPS_ESL= DPS_ESL: ESL of IPS and DPS were localized to the same sub-lobar area. IPS_ESL ≠ DPS_ESL: ESL of IPS and DPS were localized to different sub-lobar areas. T_ESL = TF_ESL: ESL results in time and time-frequency domains were localized to the same sub-lobar area. T_ESL ≠ TF_ESL: ESL results in time and time-frequency domains were localized to different sub-lobar areas. Spatial concordance was considered to be total when ESL in the time and time-frequency domains were localized to the same sub-lobar area and the epileptogenic sub-lobar area (in green) and complementary when they were localized to different sub-lobar areas, including the epileptogenic sub-lobar area (in gray). The spatial concordance was considered to be partial (in blue) when only one methodology gave results concordant with the epileptogenic sub-lobar area. In orange, no spatial concordance between the ESL methodologies. *Only ESL in the time domain was concordant with the epileptogenic sub-lobar area. ^§Only ESL in the time-frequency domain (IPS) was concordant with the epileptogenic sub-lobar area.

Figure 5

Illustration of spatial concordance between ESL in the time and time-frequency domains and comparison with the epileptogenic sub-lobar area. (A) Total spatial concordance. The spatial concordance was total between T_ESL, IPS_ESL, and DPS_ESLand with the epileptogenic area and the ischemic lesion (balck circle, MRI) (patient 5). (B) Complementary spatial concordance. The spatial concordance was complementary between T_ESL, right medial occipital (M occ) and IPS_ESL/DPS_ESL (lateral occipital (L occ) with the epileptogenic area and cortical occipital malformation (black circle, MRI) (patient 6). (C) Partial spatial concordance. The spatial concordance was partial between T_ESL and IPS_ESL/DPS_ESL. T_ESL was not localized in the epileptogenic area whereas IPS_ESL/DPS_ESL were concordant with the epileptogenic area and the cortical dysplasia in the lateral right occipital region (balck circle, MRI) (patient 17). (D) No spatial concordance. No spatial concordance was observed between the ESL methods and the epileptogenic area localized in the left mesio-anterior temporal region (black circle, MRI) (patient 7).

Figure 6

TFA and significant statistical results (p < 0.001) for time frequency analysis and ESL concordance (patient 12). (A) TFA and significant statistical results (p < 0.001) of time frequency analysis [4–80 Hz] for all channels. The channel C5 is indicated by the black square. (B) Spike-wave averaged IEDs (on the top) and significant statistical results (p < 0.001) of time-frequency analysis [4–80 Hz] (channel C5). (C) ESL in time (dipole fit) and time-frequency (DPS/IPS) domains.

Figure 7

ESL using the dipole fitting and distributed methods (CLARA, SSLOFO) in the time domain and beamformer in the time-frequency domain. (A) In the time domain, both dipole fit and distributed methods (CLARA) localized IED origin in the right prefrontal area. ESL in the time domain was concordant with ESL in time frequency domain (IPS) and with the epileptogenic area (patient 1). (B) In the time domain, both dipole fit and distributed methods (SSLOFO) localized IED origin in the right lateral temporal area. ESL in the time domain was concordant with ESL in the time-frequency domain (DPS/IPS) and with the epileptogenic area (patient 13).