Interictal spike networks predict surgical outcome in patients with drug-resistant focal epilepsy

Abstract

Objective: To determine if properties of epileptic networks could be delineated using interictal spike propagation seen on stereo-electroencephalography (SEEG) and if these properties could predict surgical outcome in patients with drug-resistant epilepsy.

Methods: We studied the SEEG of 45 consecutive drug-resistant epilepsy patients who underwent subsequent epilepsy surgery: 18 patients with good post-surgical outcome (Engel I) and 27 with poor outcome (Engel II-IV). Epileptic networks were derived from interictal spike propagation; these networks described the generation and propagation of interictal epileptic activity. We compared the regions in which spikes were frequent and the regions responsible for generating spikes to the area of resection and post-surgical outcome. We developed a measure termed source spike concordance, which integrates information about both spike rate and region of spike generation.

Results: Inclusion in the resection of regions with high spike rate is associated with good post-surgical outcome (sensitivity = 0.82, specificity = 0.73). Inclusion in the resection of the regions responsible for generating interictal epileptic activity independently of rate is also associated with good post-surgical outcome (sensitivity = 0.88, specificity = 0.82). Finally, when integrating the spike rate and the generators, we find that the source spike concordance measure has strong predictability (sensitivity = 0.91, specificity = 0.94).

Interpretations: Epileptic networks derived from interictal spikes can determine the generators of epileptic activity. Inclusion of the most active generators in the resection is strongly associated with good post-surgical outcome. These epileptic networks may aid clinicians in determining the area of resection during pre-surgical evaluation.

Figure 1

Propagation network construction. (A) A sample SEEG recording from a patient with three channels. Spikes are denoted by asterisks. (B) Using a spike detection algorithm, we detect the total number of spikes at all channels. (C) Taking turns, we treat each channel as a reference. In this example we only show Ch1 as a reference. The spikes in the reference channel are called initial spikes (denoted by red asterisks in Fig. 1A). We then count the number of spikes in other channels that fall within 120 ms before or after each initial spike; these spikes are called propagating spikes (denoted by black asterisks in Fig. 1A). (D) For each channel, we list the latency (ms) between the propagating spikes and initial spikes. The sign test is used to determine whether spikes on a given channel occur with consistent positive or negative time delay with respect to spikes on the reference channel (null hypothesis). The positive sign test between Ch1 and Ch2 suggests that there is directional propagation between these channels. The average latency between Ch1 and Ch2 (9.8 ms) suggests that spikes in Ch2 tend to occur after spikes in Ch1. There is no propagation relationship between Ch1 and Ch3. (E) Propagation map showing the significant propagation from Ch1 to Ch2, and the lack of propagation between Ch1 and Ch3. In this example Ch1 is a source node (an area from which spikes propagate to other regions but does not receive propagation), and Ch2 is a terminal node (an area that receives propagation from other regions but does not propagate spikes further). The relationship between Ch2 and Ch3 is not explored in this example

Figure 2

Calculation of concordance measures. (A) Total number of spikes detected for an example patient with five SEEG channels. Source node channels are green. (B) The calculation of our three measures: general spike concordance (GSC), which measures the proportion of spikes in the resection (without considering propagation); source node concordance (SNC), which measures the proportion of source nodes in the resection; and source spike concordance (SSC), which measures the proportion of source spikes (spikes detected at source nodes) in the resection

Figure 3

Propagation network characteristics. Characteristic features of epileptic networks. (A) Mean number of significant channel pairs, denoting a pathway of propagation; there was no significant difference between the good and poor outcome groups (p = 0.955). (B) Mean number of source nodes; patients in the good outcome group had significantly fewer source nodes than patients in the poor outcome group (p = 0.423). (C) Mean number of intermediate nodes; there was no significant difference between patients in the good and poor outcome groups (p = 0.897). (D) Mean number of terminal nodes; there was no significant difference between patients in the good and poor outcome groups (p = 0.776). Data shown is from the first 1‐h epoch for all patients. Each white dot represents group median, and gray bars represent interquartile range

Figure 4

Comparing concordance measures between good and poor outcome patients. Comparison of the three concordance measures between patients in the good outcome group and patients in the poor outcome group. (A) General spike concordance compared between the good outcome groups and poor outcome group (p < 0.001). (B) Source node concordance compared between the good outcome group and poor outcome group (p < 0.001). (C) Source spike concordance compared between the good outcome group and poor outcome group (p < 0.001). Data shown are from the first 1‐h epoch for all patients. Each white dot represents group median, and gray bars represent interquartile range

Figure 5

Patient‐specific imaging with an overlay of network nodes. Only the SEEG electrode contacts involved in the patient's spike propagation network are shown; spiking regions that failed to demonstrate statistically significant propagation patterns are not displayed. (A) Seizure‐free patient (Engel I) with two source nodes, both included in the resection, for a total source spike concordance value of 100%. (B) Seizure‐persistent patient (Engel IIIB) with five source nodes, none being included in the resection, for a total source spike concordance value of 0%. Interictal spikes were present in the frontal lobe, but they did not demonstrate statistically significant propagation patterns