Abstracts

Rationale: The epilepsy brain commonly exhibits a network of dysfunctional activity thought to underlie seizures. Methods that can identify this network can be clinically informative. Several MRI methods can contribute to this, including resting state functional (rsfMRI) connectivity and spectroscopic imaging (MRSI). In this report we studied n=10 patients and n=8 controls using rsfMRI and MRSI, assessing the local connectivity using the regional homogeneity approach in brain regions that were metabolically abnormal as identified by MRSI. Methods: Patients were recruited from the Univ. Pittsburgh Comprehensive Epilepsy Center. Data were collected on a Siemens 3T Trio, 32 channel head coil. The whole brain rsfMRI data were collected with MB=3, 2.5mm3 EPI sequence with TE/TR 35/2000, 6.8min duration; MRSI data using a fast rosette encoded spectroscopic image to longitudinally cover 65mm (frontal-parietal-mid-temporal lobes, TE/TR 40/2000, two components each 9.5min). rsfMRI preprocessing included motion correction, within-subject registration between T1 and rsfMRI imaging data and time series linear detrending. To maintain anatomical detail, native brain space was maintained throughout without smoothing. Analysis was performed using a regional homogeneity (ReHo) approach, calculating the Kendall’s coefficient for concordance over a 27voxel neighborhood (1). The MRSI reconstruction was performed as previously described (2).For tissue referencing, structural MP2RAGE scans were segmented/parcellated using Freesurfer and co-registered with SPM. Automated LCM analysis was used to generate metabolite ratios and Cramer Rao lower bounds (CRLB). As Cr/NAA linearly varies with gray matter content, pixels of metabolic dysfunction were determined by significance testing on its Cr/NAA value as predicted by linear regression from segmented anatomical data and control regression statistics. For each patient, abnormal voxels determined by MRSI were assessed for their overlapping mean ReHo value and compared as a fraction to the overall gray matter ReHo value. This was compared to controls, applying a temporal gray matter filter (Table 1). Results: Fig. 1 shows MRSI and ReHo data from two patients. For patient B, the region of MRSI abnormality largely spans this patient’s known site of right parietal hemorrhage and overlaps with the decreased ReHo. Over n=10 patients, the MRSI defined abnormal voxels exhibited a much smaller amount of local connectivity, at 22.1?7.7% of the signal of overall gray matter. In temporal gray matter, the patients did not exhibit a different local ReHo in comparison to control (72.1? 5.4 patients; controls 70.8?5.5). Conclusions: Deoxyhemoglobin serves as an endogenous contrast signal from which rsfMRI can provide a measure of low frequency brain activity. Based on this, the ReHo analysis provides a model-free assessment of tissue gray matter. Co-registration between MRSI and rsfMRI data finds that metabolically abnormal regions exhibit lower local rsfMRI coherence in comparison to overall gray matter or temporal regions. This is consistent with the view that metabolically dysfunctional areas are less locally coherent, most likely reflecting abnormal local neurotransmission. Funding: Funding support provided from NIH NS090417 and EB011639