Abstracts

Rationale: Previous group-level studies demonstrate abnormal functional connectivity common to the focal epilepsies. However, it remains important to further understand the individual variance of functional brain connectivity in focal epilepsy, given the clinical heterogeneity from patient to patient.We introduce Adjusted Local Connectivity (ALC), a univariate statistical approach that accounts for the phase coherence among neighboring fMRI voxels to quantify regional functional connectivity differences between single epilepsy patients and the healthy population. Methods: We quantified ALC individually in 25 refractory focal epilepsy patients (30.2 ± 12.4 yr) against a group of 25 healthy control subjects (29.5 ± 9.7 yr). Frontal (11/25) and temporal (8/25) lobe epilepsies were most common. 8/25 patients were MRI-positive and 3/25 patients had previous surgery.We calculated Dynamic Regional Phase Synchrony (DRePS) from 15 mins of task-free fMRI data (repetition time = 3000ms). DRePS quantifies synchrony of nearby voxels using their instantaneous phase (following a Hilbert transform). This phase dispersion measure then constitutes a time series, whose mean and dispersion reflect local functional connectivity within a narrow frequency band (0.03-0.07 Hz). Here, we extract four different temporal features of DRePS (mean, standard deviation, inverse entropy and average power spectral density).ALC is a z-test quantifying voxel-wise differences in DRePS between a patient and the control population (N = 25). We used Random Field Theory to threshold ALC maps at voxel-wise threshold of p < 0.001 and cluster-wise threshold of p < 0.05. Results: As presented in Figure 1A, most patients had ALC voxel clusters that survived our set threshold whereas healthy controls hardly displayed any suprathreshold clusters. This is an important test as we use healthy controls as a null model in our ALC algorithm – i.e., we expect epilepsy patients to have several significant ALC clusters, but we do not expect the ALC of a single healthy control to be majorly different from other healthy controls. Figure 1B demonstrates that ALC clusters were most commonly observed in the ipsilateral hemisphere.ALC was clinically interesting in many cases. Specific examples include two cases with increased ALC proximate to the confirmed seizure focus (Figure 2A and B - arrows). Histology reports confirmed type-II focal cortical dysplasia in these two patients. Several patients also displayed high ALC close to their maximal EEG abnormality (three examples are shown in Figure 2C and E - arrows). Conclusions: ALC is a novel framework able to statistically quantify deviations in local brain connectivity between a single clinical subject and the healthy population. Although we detected elevated ALC proximate to the seizure onset zone in some patients (Fig. 1A-B), we want to emphasize that ALC will also (inevitably) delineate more diffuse network features potentially reflecting a patient’s wider ‘epileptic network’. This aligns with the contemporary view of epilepsy as a network disease and functional brain imaging data should be interpreted accordingly.We believe that ALC is another step towards reaching the ambitious aim of precision medicine: using novel scientific tools to obtain reliable clinical information, at the level of single patients rather than groups. We have released ALC as an easy-to-use toolbox to encourage clincial validation of our results. Funding: This study was supported by the National Health and Medical Research Council (NHMRC) of Australia (#628952). GJ is supported by an NHMRC practitioner's fellowship (#1060312). MP was supported by the University of Melbourne scholarships (MIRS & MIFRS).