Signal-space projections of MEG data characterize both distributed and well-localized neuronal sources

We describe the use of signal-space projection (SSP) for the detection and characterization of simultaneous and/or sequential activation of neuronal source distributions. In this analysis, a common signal space is used to represent both the signals measured by an array of detectors and the underlying brain sources. This presents distinct advantages for the analysis of EEG and MEG data. Both highly localized and distributed sources are characterized by the components of the field patterns which are measured by the detectors. As a result, a unified description of arbitrary source configurations is obtained which permits the consistent implementation of a variety of analysis techniques. The method is illustrated by the application of SSP to auditory, visual and somatosensory evoked-response MEG data. Single-trace evoked responses obtained by SSP of spontaneous activity demonstrate that a considerable discrimination against both system noise and uncorrelated brain activity may be achieved. Application of signal-space projections determined in the frequency domain to spontaneous activity illustrates the possibility of including temporal relationships into the analysis. Finally, we demonstrate that SSP is particularly useful for the description of multiple sources of distributed activity and for the comparison of the strengths of specific neuronal sources under a variety of different paradigms or subject conditions.