*2.8. Statistical Analysis of Structure-Function Coupling*

To test whether the laterality of the strong connections matched each patient's diagnosis, we first split the structural and functional connectomes into left and right hemispheres (Figure 1, Step 3, b) and removed cross-hemisphere electrode pairs. For example, if an electrode pair contained two electrodes in the left hemisphere (i.e., FP1-F3) or one left hemisphere and one central electrode (FP1-Fz), the electrode pair was kept. All electrode pairs that crossed from one hemisphere to the other (i.e., F3-F4) were removed. To test whether the highly connected electrode pairs from the structural and functional connectomes were congruent, we first computed the z-scores for all electrode pairs from all connectomes (Figure 1, Step 3, b, i). The z-score arrays were: a. 1 *×* 1D array per hemisphere for the structural connectome and b. 5 *×* 1D arrays (for each 1 s time window) per seizure, per hemisphere for the functional connectomes. Lastly, the structural and functional zscores from each hemisphere were displayed in parallel format for analysis (Figure 1, Step 3, b, ii).

To preserve only the most robust connections to represent high coherence between two electrodes, a z-score threshold of 2 (i.e., two standard deviations from the mean) was chosen for the structural connectome, and a threshold of 1.8 was chosen for the functional connectomes. Next, the z-scores from both connectome types were compared per onesecond window from each seizure. If the same electrode pair from the structural and functional connectomes contained a z-score between 1.8 and 2 (or greater than 2), that electrode pair was classified as showing high structure-function coupling (termed "coupled electrode pairs").

The senior neurologist provided a "laterality" score for each patient based on whether the most frequently observed seizure onset zone was consistently restricted to one hemisphere. A laterality score of zero represented poorly lateralised seizures, whilst highly lateralised seizures received a score of three. If overall, the patient had late-lateralising seizures, they were classified as being non-lateralised at onset (i.e., a score of 0–1). Each patient was also assigned an "expected onset zone", predicated on the most frequently observed onset zone observed in all of a patient's recorded seizures, including seizures that were poorly or non-lateralised. The neurologist reviewed the raw EEG to confirm whether the electrode pair with high structure-function coupling was congruent with the expected seizure onset zone. All z-scores and statistical analyses were produced in SPSS v28 (Armonk, NY, USA: IBM Corp).
