Abstract
Our study aimed to demonstrate time-dependent declarative memory changes and its associated brain regions after status epilepticus (SE) using structural imaging techniques and machine learning methods. Pilocarpine was administrated to establish the SE model. At four different time points after SE (1, 2, 3, and 4 months, respectively), rats were subjected to structural imaging acquisition as well as contextual fear conditioning for the measurement of brain structural changes and declarative memory. Voxel-based morphometry (VBM) analysis were performed. Those significantly different regions were selected as features for training support vector machine (SVM). A linear kernel was chosen for regression of declarative memory. Leave-one-out cross-validation was applied to ensure generalization. Our results showed that the pilocarpine groups displayed the most severely impaired declarative memory at 2 months after SE and improved afterward, but failed to recover to the normal condition at 4 months after SE. The pilocarpine groups showed lower gray matter volumes and larger cerebrospinal fluid (CSF) volumes. After controlling for the total brain volumes, ANOVA demonstrated gray matter volume changes in the CA1 subfield of the hippocampus, primary somatosensory cortex, entorhinal cortex, etc. The combination of VBM and SVM identifies the somatosensory cortex and entorhinal cortex as the correlated brain regions for declarative memory dysfunctions after SE. Our study indicates that compensational mechanisms might be triggered to help with the recovery of memory functions after SE. Structural changes of the somatosensory cortex and entorhinal cortex might be involved in memory impairment after SE.