Estimating correlation for a real-time measure of connectivity

Area of Science:

• Neuroscience
• Computational Neuroscience
• Signal Processing

Background:

• Connectivity analysis of fMRI and EEG time-series is of significant interest.
• Estimating connectivity using pair-wise correlation (PWC) is computationally intensive.
• Real-time estimation of PWC for multiple time-series poses a significant challenge.

Purpose of the Study:

• To develop and evaluate a parallel algorithm for real-time PWC computation.
• To assess the algorithm's performance on streaming data from multiple channels.
• To determine if the implemented algorithm meets real-time constraints for neuroimaging data.

Main Methods:

• Developed a parallel algorithm for computing pair-wise correlation (PWC) on streaming data.
• Implemented the algorithm on Intel Xeon™ and IBM Cell Broadband Engine™ platforms.
• Evaluated execution time using signals recorded with different acquisition parameters.

Main Results:

• The parallel algorithm enables real-time PWC computation for multiple time-series.
• Execution times were evaluated against real-time constraints.
• Efficient implementations met real-time constraints in most tested scenarios.

Conclusions:

• The developed parallel algorithm effectively addresses the computational challenges of real-time connectivity analysis.
• This approach facilitates real-time processing of fMRI and EEG data for connectivity estimation.
• The findings support the feasibility of real-time neuroimaging data analysis using efficient parallel computing.