University of Houston

Recent advances in acquisition hardware design allow the collection of electroencephalography (EEG) and magnetoencephalography (MEG) time series at an ever increasing temporal and spatial resolutions. At the same time, computationally complex analysis methods have also become very popular. Therefore, efficient implementation of such methods that take advantage of the newest computational hardware and the latest developments in algorithms is of paramount importance to analyze the EEG and MEG signals efficiently.

In this dissertation, we present computational methods and optimizations that enable efficient computation of large multivariate models for time series analysis. We focus on the efficient implementation of Granger causality algorithm on computer clusters . We explore the applicability of both shared and distributed memory programming paradigms to enable the analysis of arbitrarily large data sets. Our solutions are validated using clinical datasets of varying sizes. Computation of brain activation profiles using our implementation of Granger causality has resulted in reduction of computation time by up to 98.5% over existing versions.

In addition to Granger causality, we parallelize an iterative artifact removal algorithm based on independent component analysis that we have previously developed. Parallelization has resulted in a speedup of an order of magnitude and can now provide quasi real time performance.