Low dimensional representations of neuronal activity in Parkinson’s disease
Abstract: This project has been concerned with developing methods for dimensionality reduction and feature extraction of brain activity in the basal ganglia in parkinsonian brains. Dimensionality reduction of local field potential activity was based on feature vectors produced from the discrete Fourier transform of activity. A heuristic-motivated visualization of these feature vectors using the k-Means algorithm for prediction and classification was used. The feature vectors were also subject to principal component analysis as a further means of feature extraction and analysis. Dimensionality reduction of spiking activity was based on spiking rates, joint rate distributions, serial correlation coefficients, power spectral density, and spectral entropy. The methods for dimensionality reduction and feature extraction developed were used to show similarities in simultaneous brain activity, notable characteristics of brain activity, and notable similarities and differences in the brain activity when comparing activity from different sub-regions of the basal ganglia and when comparing brain activity from different animals. We believe that the methods developed in this project show promise in further research of Parkinson’s disease.
AT THIS PAGE YOU CAN DOWNLOAD THE WHOLE ESSAY. (follow the link to the next page)