**2. Materials and Methods**

*2.1. Participants and Data*

Nine adults with focal epilepsy were recruited from the Comprehensive Epilepsy Centre at the Royal Prince Alfred Hospital (RPAH, Sydney, Australia), and MRI was performed at the Brain and Mind Centre (Sydney, Australia). Inclusion criteria were adults diagnosed with focal epilepsy, aged 18–60, presenting without surgery, with a minimum of two recorded seizures, and who were willing and able to comply with the study procedures for the duration of their participation. Exclusion criteria were pregnant women and individuals with intellectual disabilities. Ethical approval was obtained from the RPAH Local Health District (RPAH-LHD) ethics committee (see Institutional Review Statement in Section 5). The entire data processing and analysis consisted of several consecutive steps, depicted in Figure 1.

**Figure 1.** Schematic of data processing and analysis steps. To obtain the structural connectomes ("SC", Step 1, a), the dMRI was processed, and anatomically-constrained probabilistic tractography was conducted as outlined in Section 2.3. Cortical regions of interest were based on the Desikan-Killiany (DK) [32] atlas. To obtain the functional connectomes ("FC"), the first 5 s of a given seizure were selected using one-second windows. Each one-second window was processed using Curry's sensor coherence algorithm (Step 1, b, ii), producing a 21 × 21 coherence matrix. The reference electrodes were removed before statistical analysis, resulting in a 19 × 19 coherence matrix which was used as the functional connectome. In Step 2, the ANTs non-linear registration tool was used to warp electrodes in the MNI template space to the subject space, creating a subject-specific electrode warp (a, i–iii). To produce a subject-specific, one-to-one map of each cortical region to its nearest electrode, we applied our inverse square method (b, i). The inputs were each subject's electrode warp and cortical region labels from Step 1. The result was a structural connectome with an electrode name corresponding to each of the 70 regions, i.e., F7/L.LOFG (b, ii). In Step 3 (a), the 70 × 70 structural connectome was condensed to match the dimensions of the functional matrix (19 × 19). Specifically, the values of all regions corresponding to a given electrode pair were summed, and the total value was used as the connectivity value for that same electrode pair in the new condensed structural connectome. Lastly, z-scores were computed for all connectivity values in the structural and functional connectomes (b, i,ii) and the statistical analysis was conducted.
