Using correlated resting BOLD signal fluctuations to delineate seizure networks
Abstract number :
3.269
Submission category :
5. Neuro Imaging
Year :
2011
Submission ID :
15335
Source :
www.aesnet.org
Presentation date :
12/2/2011 12:00:00 AM
Published date :
Oct 4, 2011, 07:57 AM
Authors :
C. Keller, S. Bickel, S. Jain, D. Groppe, L. Entz, C. Kelly, P. Kingsley, S. Hwang, C. Harden, F. Lado, A. Mehta
Rationale: It is increasingly being recognized that correlated fluctuations of the blood oxygen level dependent signal (BOLD ) recorded at rest underlies the brain s functional intrinsic network architecture. The epileptogenic zone may involve ictal onset areas as well as areas where seizures spread. We investigated seizure networks using resting state functional connectivity analysis of the BOLD signal in patients undergoing invasive electrode monitoring for epilepsy surgery. Specifically, using seeding and clustering approaches we computed predictive values for parts of the seizure onset area to localize the entirety of the seizure network as defined with intracranial EEG.Methods: We recorded functional magnetic resonance imaging (fMRI) at rest on a 3T GE scanner in 6 patients with medically intractable focal epilepsy prior to implantation of subdural electrodes. The region underlying electrodes defined as ictal onset and seizure spread were used as seeds for functional connectivity analysis (RSFC). Additionally, a clustering approach was used to delineate the possible local spatial extent of seizure onset areas. Specifically, a similarity index for whole brain correlation maps of neighbouring voxels was subjected to k-means clustering to identify regions with similar whole brain connectivity patterns. Furthermore, sensitivity, specificity, likelihood ratios, negative and positive predictive values were computed for these methods to identify seizure onset and spreading areas.Results: Parts of the seizure networks showed correlated BOLD signal correlations. Particularly negative predictive values showed promising results. Clustering identified distinct regions around seizure onset electrodes that showed similar connectivity patterns.Conclusions: Seizure networks showed correlated BOLD signal fluctuations at rest. High negative predictive values of parts of the seizure network to identify other electrodes overlying epileptic brain tissue suggests that this approach may be useful in helping localization of non-pathological brain areas. Further studies taking into account clinical outcomes following resection are needed to validate this approach.
Neuroimaging