Talk January 25, 2001, Functional Brain Imaging Center of Washington University in St. Louis

Blind Decomposition Reveals Hemodynamic
and Electrophysiologic Response Features

Scott Makeig
The Salk Institute
La Jolla CA

Both current primary human neuroimaging techniques, fMRI and high-density EEG/MEG, record highly complex records of changes in brain activity through time. Unfortunately, in both modalities it is difficult to separate the data into functionally distinct sources that must in both cases include several types of artifacts. Analysis techniques for functional magnetic resonance imaging (fMRI) data typically require a priori knowledge of the time course of the hemodynamic response (HR) and assume the homogeneity of both HR and noise across different brain regions and even across different subjects. These approaches may be problematic when the expected time course is unknown or the hemodynamic assumptions are invalid. Source decomposition methods for EEG/MEG typically assume the data to sum synchronous activity within several dipole-like cortical areas. Unfortunately, the EEG/MEG source inversion problem is undercomplete, and it is difficult to know how to combine the low-resolution spatial and high-resolution temporal information contained in EEG data.

Since 1995 my colleagues and I at Salk Institute have been exploring the applications of blind source separation methods based on Independent Component Analysis (ICA) for detecting unforeseen HRs in event-related fMRI experiments and/or unforeseen spatiotemporal phenomena in event-related EEG experiments.

Since its practical development about 1995, ICA has rapidly become a widely explored technique in modern signal processing, with new or potential applications in a wide range of fields. This is occurring because the idea behind ICA is in fact quite simple. Rather than finding phenomena in multi-dimensional data that most closely match some a priori model, ICA blindly separates the given data into maximal independent information sources (in a technical sense). This proves to be possible under a range of conditions, and is after all (in a more general sense) what one needs to be able to separately examine phenomena of interest in the data.

I will show some current results of analysis of EEG and fMRI data that combine ICA and other signal processing methods. The fMRI results, in particular, suggest that HRs from fRMI experiments may be site dependent, subject dependent and task dependent and may vary widely from time to time or trial to trial. The causes of single-trial HR variability, while unknown, probably include variations in subject cognitive strategy or state. Since HRs in brain areas involved in higher cognitive functions may be expected to be more variable than HRs in primary sensory and motor areas, use of data-driven analysis methods such as ICA appears essential to identify and explore their individual, time-varying features. Likewise, the EEG results are inspiring new attempts to understand the functional significance of macroscopic event-related brain dynamics.

Return to Current Abstracts