Event Abstract Back to Event Using a temporal decomposition technique to investigate dynamic changes in subnetworks derived from resting state fMRI during generalized spike-and-wave discharges: an EEG-fMRI pilot study Chayanin Tangwiriyasakul1*, Suejen Perani1, Siti N. Yaakub1 and Mark P. Richardson1 1 King's College London, Basic and Clinical Neuroscience, United Kingdom Introduction: Many recent studies suggest that generalized seizures arise from abnormality in whole brain networks (Richardson et al. 2012). Most evidence shows abnormalities in static networks (e.g. reduced default mode network connectivity). In this study, we are interested to investigate dynamic changes in the level of synchrony during generalized spike and wave discharge (GSW) period using simultaneous EEG-fMRI recordings. Methods: Three patients with newly diagnosed epilepsy underwent four resting state scans (each scan lasts ~10 minutes). Pre-processed fMRI data were band-pass filtered between 0.04-0.07 Hz (Glerean et al. 2011) and a Hilbert transform was applied. The AAL atlas was used to divide the brain into 90 cortical regions, with each region represented by the first principal component of the instantaneous phase of all voxels in the region. We estimated a binary synchronization matrix at each time-point with regions considered to be synchronized if the phase difference was less than π/6, giving 296 binary synchronization matrices for each subject (Ponce-Alvarez et al. 2014). Tensor decomposition technique was then used to extract sub-networks, resulting in 9 major components (90x90 matrices). GSW periods were marked by an experienced neurologist. Results: A total of 32 GSW occurrences were found in this study. Figure 1 shows an example of temporal evolution of nine subnetworks in patient-1. From visual inspection (see Figure 1B/1C), 20 out of 32 SWD occurrences (63%) were found during a minimum of network synchrony (or turning point, labelled as T in Fig 1B/C). One the other hand, 9 events (27%) were found while the net contribution from all subnetworks was in a maximum stable-and-synchronous state (or plateau state, labelled as P in Fig 1B/C). The remaining 3 GSW occurrences (10%) were unclassified. Figure 1: Subnetwork structure of spatiotemporal synchronization patterns in patient-1. A) The 9 detected subnetworks. B) Temporal activation strength of each subnetwork. C) Total activation strength of all nine subnetworks (which represents a level of total network synchrony). D) Temporal evolution of phase difference throughout the ~10 minute scan. E) Band-pass filtered fMRI raw signals in 25 brain regions. The red stripes indicate the period during GSW. Note that T = turning and P = Plateau. Conclusion: Up to now, the mechanism underlying seizure generation is not yet clear. The results from this study may suggest two possible causes of SWD generations: (1) transition between two states, which may cause instability in the brain networks or (2) over synchronization of the brain networks. In the future, a larger study population is needed to address these two hypotheses. Figure 1