Single voxel autocorrelation reflects hippocampal function in temporal lobe epilepsy

We have previously shown that autocorrelation analyses of BOLD signal applied to single voxels in healthy controls can identify behaviorally-relevant gradients of temporal dynamics throughout the hippocampal long-axis. A question that remains is how changes in the brain's functional and structural integrity affect single-voxel autocorrelation. In this study, we investigate how hippocampal autocorrelation is affected by hippocampal dysfunction by investigating a population of patients with unilateral temporal lobe epilepsy (TLE). Many patients with TLE have mesial temporal sclerosis (MTS), characterized by scarring and neuronal loss, particularly in the anterior hippocampus. Here, we compared patients with left and right TLE, some with and without MTS, to healthy controls. We applied our single-voxel autocorrelation method and data-driven clustering approach to segment the hippocampus based on the autocorrelation of resting-state fMRI. We found that patients with left TLE had longer intrinsic timescales (i.e., higher autocorrelation) compared to controls, particularly in the anterior-medial portion of the hippocampus. This was true for both the epileptogenic and non-epileptogenic hemispheres. We also evaluated the extent of cluster preservation (i.e., spatial overlap with controls) of patient autocorrelation clusters and the relationship to verbal and visuospatial memory. We found that patients with greater cluster preservation in the anterior-medial hippocampus had better memory performance. Surprisingly, we did not find any effect of MTS on single-voxel autocorrelation, despite the structural changes associated with the condition. These results suggest that spatiotemporal dynamics of activity can be informative regarding the functional integrity of the hippocampus in TLE.