**1. Introduction**

The assessment of patients with focal epilepsy using a combination of structural data derived from diffusion MRI (dMRI) and functional data from electroencephalography (EEG) is gaining increased appeal [1–3]. In the brain, structural connectivity refers to an anatomical link between two or more brain regions. Connnectomes generated from diffusion MRI, can represent the strength of structural connectivity between specific brain regions. Functional connectivity is inferred from the spatio-temporal relationship between electrophysiological signals from two or more structurally discrete regions [4]. Structural connectivity is believed to give rise to functional and network behaviour [5]. In a mechanistic sense, the composition of white matter can be expected to influence the flow of activity and connectivity between neuronal populations. Therefore, if EEG functions as a tool to observe the flow of activity, the connectivity measurements from EEG can be presumed to closely resemble connectivity measurements from structural MRI. In epilepsy, structure-function coupling is proposed to have a role in identifying seizure propagation patterns [6,7], seizure generalisation [1] and predicting post-surgery seizure freedom [8,9]. Diffusion MRI derived tractography,

**Citation:** Maher, C.; D'Souza, A.; Barnett, M.; Kavehei, O.; Wang, C.; Nikpour, A. Structure-Function Coupling Reveals Seizure Onset Connectivity Patterns. *Appl. Sci.* **2022**, *12*, 10487. https://doi.org/ 10.3390/app122010487

Academic Editors: Alexander E. Hramov and Alexander N. Pisarchik

Received: 15 September 2022 Accepted: 12 October 2022 Published: 18 October 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

in conjunction with EEG, can enable the quantification of structural connectivity between different brain regions. However, in epilepsy, dMRI is held to be in the experimental realm [2]. Therefore, though there is consensus on the significance of structural connectivity information in patient diagnosis, the utility of dMRI as a routine clinical test has not been realised. Additional research is needed to investigate the value of dMRI in combination with routinely collected data such as EEG. Further, the feasibility of user-friendly tools for deploying dMRI pipelines must be assessed.

Several works employ functional MRI (fMRI) to represent functional connectivity [10–12] alongside structural connectivity from dMRI. However, given EEG is routinely collected in epilepsy clinics, it may be a more accessible and practical alternative to fMRI, which has inherent, poor temporal resolution relative to EEG. White matter connectivity and information flow between specific brain regions has been linked to scalp EEG characteristics in healthy populations [13,14]. Further, EEG has been used to produce an individualised connectivity fingerprint that is robust across recordings [15], rendering its utility as a patient-specific, analytical network measure that can address the heterogeneous nature of focal epilepsy.

Discerning the seizure onset pattern and epileptogenic zone has been shown to improve the prognosis of post-surgical outcomes [16], and EEG and dMRI can aid this goal. A study on the role of scalp EEG in predicting post-surgical seizure outcomes showed abnormal MRI was valuable in ambiguous cases containing bilateral interictal epileptiform discharges [17], suggesting MRI may enhance prediction of seizure freedom. In another study of seven patients being evaluated for epilepsy, lesional and non-lesional MRIs were combined with high and low frequency bands from high density EEG (HDEEG) [18]. The Authors showed that the absence of structural support was related to significantly reduced functional connectivity in high frequency bands. Moreover, high frequency oscillations observed on scalp EEG are increasingly recognised as a hallmark of lesional epilepsy [19]. These works highlight the advantages of combining dMRI with EEG to detect aberrations that may typically only be partly revealed by one modality.

The majority of works that blend multimodal information from dMRI and EEG focus on source localisation techniques [20–22], using a digitiser to map electrode coordinates to the scalp which can be time-consuming. Others produced an automated, individualised localisation tool to map electrodes from high density EEG (HDEEG) to the scalp only, without extending the mapping to the cortex [23]. Many prior works favoured the combination of stereo EEG with dMRI [6,9,24,25], or only explored normal (non-ictal) awake EEG data with dMRI [26].

Several methods for electrical source localisation, which utilise a range of forward and inverse solutions, have been proposed and evaluated [27–29]. Thus the current study is distinguished from those prior works for the following reasons. We aimed to understand whether a patient-specific, structure-function coupling pattern could be observed without requiring manual digitisation of electrodes or applying one of the several forward and inverse solutions. We sought to apply our existing model [30], which maps cortex regions to individual electrodes, to a larger cohort. We specifically examined the seizure onset period (regardless of wakefulness state). Lastly, we aimed to validate the feasibility of our model as a clinically translatable method to leverage the potential of dMRI, with the view of elevating it to the established state currently held by structural MRI (i.e., T1) [31]. The dMRI component of our tool was designed to be deployed on a clinician's computer, allowing straightforward data processing from new patients (with ethics approval).

The contribution of this work is twofold: a. We extend the application of our spatial mapping model to a new patient cohort, highlighting consistent between-patient variance in region to electrode mapping, and b. We add to the growing body of research showing that connectivity data derived from structural MRI may augment scalp EEG observations for certain patients; acting as an additional tool during the diagnosis stage.
