No abstract

The hippocampus is thought to guide navigation and has an essential contribution to learning and memory. Hippocampus is one of the brain regions impaired in Alzheimer's disease (AD), a neurodegenerative disease with progressive memory impairments and cognitive decline. Although successful treatments for AD are still not available, developing new strategies to detect AD at early stages before clinical manifestation is crucial for timely interventions. Here, we investigated in the TgF344-AD rat model the classification of AD-transgenic rats versus Wild-type littermates (WT) from electrophysiological activity recorded in the hippocampus of freely moving subjects at an early, pre-symptomatic stage of the disease (6 months old). To this end, recorded signals were filtered in two separate frequency regimes namely low frequency LFP signals and high frequency spiking activity and passed to machine learning (ML) classifiers to identify the genotype of the rats (TG vs. WT). For the low frequency analysis, we first filtered the signals and extracted the power spectra in different frequency bands known to carry differential information in the hippocampus (delta, theta, slow- and fast-gamma) while for the high frequency analysis, we extracted spike-trains of neurons and calculated different distance metrics between them, including Van Rossum (VR), Inter Spike Interval (ISI), and Event Synchronization (ES). These measures were then used as features for classification with different ML classifiers. We found that both low and high frequency signals were able to classify the rat genotype with a high accuracy with specific signals such as the gamma band power, providing an important fraction of information. In addition, when we combined information from both low and high frequency the classification was boosted indicating that independent information is present across the two bands. The results of this study offer a better insight into how different regions of the hippocampus are affected in earlier stages of AD.