Physiological mechanisms of hippocampal memory processing : experiments and applied adaptive filtering

Abstract

The hippocampus is necessary for the formation and storage of episodic memory, however, the computations within and between hippocampal subregions (CA1, CA3, and dentate gyrus) that mediate these memory processing functions are not completely understood. We investigate by recording in the hippocampal subregions as rats execute an augmented linear track task. From these recordings, we construct ensemble rate representations using a point process adaptive filter to characterize single-unit activity from each subregion. We compared the dynamics of these rate representations by computing average max rate and average rate modulation during different experimental epochs and on different segments of the track. We found that the representations in CA3 were modulated most when compared to CAl and DG during the first 5 minutes of experience. With more experience, we found the average rate modulation decreased gradually across all areas and converged to values that were not statistically different. These results suggest a specialized role for CA3 during initial context acquisition, and suggest that rate modulation becomes coherent across HPC subregions after familiarization. Information transfer between the hippocampus and neocortex is important for the consolidation of spatial and episodic memory. This process of information transfer is referred to as memory consolidation and may be mediated by a phenomena called "replay." We know that the process of replay is associated with a rise in multi-unit activity and the presence of ripples (100-250 Hz oscillations lasting from 75ms to 100ms) in CAl. Because ripples result from the same circuits as replay activity, the features of the ripple may allow us to deduce the mechanisms for replay induction and the nature of information transmitted during replay events.

(cont.) Because ripples are relatively short events, analytical methods with limited temporal-spectral resolution are unable to fully characterize all the structure of ripples. In the thesis, we develop a framework for characterizing, classifying, and detecting ripples based on instantaneous frequency and instantaneous frequency modulation. The framework uses an autoregressive model for spectral-temporal analysis in combination with a Kalman filter for sample-to-sample estimates of frequency parameters. We show that the filter is flexible in the degree of smoothing as well as robust in the estimation of frequency. We demonstrate that under the proposed framework ripples can be classified based on high or low frequency, and positive or negative frequency modulation; can combine amplitude and frequency information for selective detection of ripple events; and can be used to determine the number of ripples participating in "long ripple" events.