Graph theoretical analysis of structural brain networks in autoimmune limbic encephalitis
1 Department of Epileptology, University Hospital Bonn
2 Department of Neuropathology, University Hospital Bonn
3 Institute of Experimental Epileptology and Cognition Research, University Hospital Bonn
Limbic encephalitides (LE) form a group of inflammatory brain diseases that may be characterized by detected autoantibodies against extracellular antigens such as glutamic acid decarboxylase (GAD) or intracellular antigens such as voltage gated potassium channel (VGKC). Whereas temporomesial structures lie in the focus of imaging analysis, in this study we aimed to investigate structural brain networks. 17 patients with GAD-autoantibody associated LE (GAD-LE), 14 patients with VGKC-autoantibody associated LE (VGKC-LE), 20 control patients with hippocampal sclerosis and 33 healthy control subjects underwent structural and diffusion tensor imaging at 3T. Patient images were reoriented according to pathological EEG resulting in an affected and an unaffected hemisphere. Cortical reconstruction, volumetric segmentation and automated labeling of ROIs used as nodes in the network model was performed using FreeSurfer. MRtrix3 and the Brain Connectivity Toolbox were applied to derive edge weights and perform graph theoretical analysis. Patients with GAD-LE showed significantly increased characteristic path length (ANCOVA F(3,74)=2.35, p=0.049) and significantly reduced local efficiency (ANCOVA F(3,78)=2.80, p=0.023) as compared to healthy controls and control patients with hippocampal sclerosis (post-hoc Tukey-Kramer p<0.05) using age and gender as covariates. Patients with VGKC-LE did not differ significantly from neither control group. In regression analysis across both LE groups, the affected amygdala volume was found to be a predictor for local efficiency (F(2,18)=3.35, p=0.049) adjusting for age. Our finding of distinct network affection mirrors the clinically heterogeneous presentation of LE patients. We hypothesize that reduced network efficiency in acute LE can be used as predictive marker for therapy outcome.