Figure 3. .

Influence of seizure onset zone on sensitivity of network statistics to subsampling. Each nodal and global metric is shown. All data shown are for the EEC of first seizure and high gamma coherence. (A) Association between metric agreement and distance of ignored electrode contacts from the seizure onset zone. Each point represents the patient-specific Spearman rank correlation coefficient between the metric agreement and the distance of the ignored electrodes from the seizure onset zone. The metric agreement is defined for nodal metrics as the Spearman rank correlation coefficient between the original metric and the metric obtained from subsampling, and for global metrics as the negative absolute value of the relative difference between the original and subsampled metric. Horizontal lines show the average distance-agreement association across patients. No distance-agreement association was significantly different from 0 (two-sided one-sample t test, α = 0.05/8 for Bonferroni correction), signifying that all metrics are equally vulnerable to incomplete sampling near versus distant from the seizure onset zone. (B) Difference in metric agreement between seizure onset zone-sparing and seizure onset zone-targeted subsampling. Each point represents the patient-specific difference in metric agreement between subsampling that spares the seizure onset zone electrode contacts and subsampling that targets the seizure onset zone electrode contacts. Positive values indicate that sparing the seizure onset zone from removal yields a higher agreement with the original network statistic than does targeting the seizure onset zone for removal. Horizontal lines show the average difference in metric agreement across patients. No difference in agreement was significantly different from 0 (two-sided one-sample t test, α = 0.05/8 for Bonferroni correction), signifying that metrics are equally vulnerable to subsampling that spares versus targets the seizure onset zone. However, there was nonsignificantly higher transitivity metric agreement when the seizure onset zone was spared. The horizontal position of individual points was determined by random jitter to improve readability.