Activities and Seminars

Pietro Guccione. Multivariate Signal Processing in fMRI.
Date: May 27, 2014

What: Multivariate Signal Processing in fMRI.

Where: BCBL auditorium

Who: Pietro Guccione, PhD, Assistant Professor in Communication Systems and Signal Processing at Dipartimento di Ingegneria Elettrica e Informazione, Politecnico di Bari; Italy.

When: 12 noon

Multivariate Signal Processing is a modern and very exploited way to look at data. Data can say more than we imagine on their origins: data are able to say where they come from, how they are joined together, which relation they have with other set of data, how they can predict future behavior of systems and so on. Our research group has been interested in Signal Processing since 1999. Main interests have been on Remote Sensing, with application on Synthetic Aperture Radar design, parameter optimization and processing techniques. With this knowledge, a new line of research has started two years ago on multivariate data, with particular application on biomedical imaging. The presentation of our research activities is on two case studies both focused on the analysis and grouping of subjects based on their functional Magnetic Resonance Imaging data acquired during a task completion. We followed a linear data-driven and group analysis approach, the M-CCA that, compared to model-driven approaches, makes no assumptions on the response to recover. In the second study, instead, we followed a nonlinear approach. There is a large assumption in literature to consider physiological and biological systems as complex processes whose dynamics are constantly being influenced by nonlinear interactions. Following such models we tried to characterize fMRI signal using nonlinear invariant properties. The works presented perhaps provide us with more questions than answers; on the other hand these works are still open and not all the research lines have been investigated or explored enough. We hope anyway that the approach and the potentials of such applications will be of interest.