Abstracts

Rationale: Epilepsy is a disorder of connectivity that taps into established brain networks. Resting-state fMRI (rs-fMRI) functional connectivity (FC) methods may be able to lateralize/localize the seizure focus. Efforts to use rs-fMRI connectivity methods to identify the seizure focus are few and lacking validation, especially in children. No study has implemented connectivity analysis in task-based fMRI (e.g. language or motor fMRI) to achieve seizure focus identification, or to examine if the influence of seizure focus on brain connectivity remains the same across different cognitive states (resting-state vs. task). We aimed to explore this question with a set of 3D voxelwise fMRI measures spanning from connectivity to time frequency analysis in our heterogenous presurgical pediatric patients (focus lobe, side & etiology) at a hemispherical level as a first step given 1) task state might modulate certain task related region's FC pattern and 2) different etiology of seizure focus might exhibit different direction of abnormality for different measures. We hypothesized that patients will demonstrate lateralized fMRI measures at a hemispherical level at rest and their degree of lateralization will be correlated with language and motor cognitive states.  Methods: 41 focal epilepsy patients (mean age 12.9+/-3.4 yr (6-19), 21 female, mean age onset 7.4+/- 4.7yr (0.1-17), 30 left side focus) underwent pre-operative language (N=38) or motor (N=25) fMRI and resting state fMRI (N=41). Language task, auditory description decision, typically is left activation; motor task, finger tapping, is usually bilateral. 3T scans are 5 min with block designs for tasks and eyes open for resting state. Seizure focus was confirmed through ictal vEEG and/or MRI. To increase the reproducibility of fMRI findings, we used fmriprep (https://github.com/poldracklab/fmriprep) for preprocessing. We generated the following voxelwise 3D maps, indicated in recent studies that might be abnormal in epilepsy patients: 1) local FC (Euclidean distance <14mm), 2) distant FC (Euclidean distance >14mm), 3) global FC (to rest of the brain), 4) Intrahemisphere FC (IntraFC), 5) Interhemisphere FC (InterFC), 6) Ipsi/contralateral FC hemispheric contrast(FCHC), by computing for each voxel the difference of number of voxels connected to it ipsilaterally vs. contralaterally , as well as 7) voxelwise frequency power spectrum(FQPS) map. All maps were generated in native patient space and in MNI space. We then calculated the laterality index of each of the above 3D maps for each patient at each state at the hemispheric level (L-R/L+R) and computed the Pearson Correlation between Rest vs ADT, and Rest vs Motor.  Results: At resting state, within each fMRI measure most patients did not demonstrate a hemispherical level of lateralization, especially distant FC and InterFC (Table 1). However, majority patients demonstrated lateralization (30/41 for |LI|>0.2 and 40/41 for |LI|>0.1) for at least one of the 7 measures. LIs at rest are positively correlated with those at ADT for FQPS, global FC and distant FC (rs>0.4, ps<0.02, N=38), and correlated with Motor for FQPS, local FC, distant FC (rs>0.44, ps<0.03, N=25) regardless of native or MNI space.  Conclusions: We demonstrated that hemispherical level of lateralization is present in most patients (at least one fMRI measure) and consistent across rest and task states for some of the voxelwise fMRI measures. Our findings suggested the potential of adding language/motor fMRI to rest fMRI run, which could be helpful for pediatric patient to have longer data to achieve stable network structure for lateralization of seizure focus.  Funding: UL1TR001876 National Center for Advancing Translational Sciences