INTERICTAL EPILEPTIFORM DISCHARGE MAPPING BASED ON LOCAL SYNCHRONIZATION OF FMRI SIGNAL
Abstract number :
1.183
Submission category :
5. Neuro Imaging
Year :
2013
Submission ID :
1742866
Source :
www.aesnet.org
Presentation date :
12/7/2013 12:00:00 AM
Published date :
Dec 5, 2013, 06:00 AM
Authors :
J. V. Liu, E. J. Kobylarz, G. P. Thomas, B. Jobst
Rationale: Epilepsy diagnosis and treatment benefit from the mapping of interictal epileptiform discharges (IED), which is achieved at much higher spatial resolution with functional MRI than with scalp EEG. In previous studies that use either simultaneous EEG-fMRI (Pittau et al., Neurology 78:1479) or fMRI alone (Khatamian et al., Epilepsy Res 94:177), fMRI signal analysis is based on amplitude increases in each voxel, which represent increased neural activities during IED. However, IED are also accompanied by increased neural synchronization across a local area (e.g. 6-20 cm2) surrounding the IED (Wennberg et al., Clin Neurophys 122:1295). This encouraged us to develop new fMRI analysis paradigms based on the increase in synchronization of fMRI signals across voxels within a local area during IED. Such synchronization can be measure by local degree centrality (LDC), a network metric of local functional connectivity (Sepulcre et al., PLoS Comp Biol 6:e1000808).Methods: One-hour simultaneous scalp EEG (32 channels) and whole-brain fMRI (resolution: 3 mm, 2.5 s) data were acquired from 5 patients with focal epilepsy, and pre-processed using steps similar to that in previous studies. For IED mapping that uses EEG, fMRI data of 20 s duration around each IED event identified from EEG were stitched together into an IED-containing time segment, and IED-absent time segments of the same duration were constructed for comparison. LDC was computed for each time segment and each voxel as the sum of the fMRI signal correlation coefficients between this voxel and all voxels within 1.4 cm distance (i.e. an area of > 6 cm2). The statistical significance (z-score) of LDC difference between IED-containing and IED-absent time segments was mapped. For IED mapping without EEG, LDC was computed for segments of 40 s duration slid across time, effectively creating a time series of LDC values for each voxel. Using cluster analysis (Khatamian et al.), voxel clusters that have the maximum LDC value at the same time were isolated and then inspected visually to identify the cluster most likely associated with IED.Results: For mapping that utilizes EEG, focal regions with significant LDC increases during IED were identified in all 5 patients, and their locations were concordant with EEG. Figure 1 shows the LDC z-score map in a patient with EEG spikes in left temporal lobe. The LDC-based approach performed better than the conventional EEG-fMRI method, which failed to identify concordant activation regions in 2 of the 5 patients who had relatively fewer events. For mapping without EEG, the LDC-based approach identified concordant voxel clusters in 3 patients, again with better performance than the conventional method. The resultant map (Figure 2) is often less specific than when using EEG and involves resting-state networks unrelated to IED, consistent with previous studies.Conclusions: The new method based on local fMRI signal synchronization is at least as effective in mapping interictal activities as the conventional methods based on signal amplitude, both when using simultaneous EEG to identify IED events and when without EEG.
Neuroimaging