Total information extracted from MEG measurements

The total information derived from Shannon's theory of communication provides a completely new index to characterize multichannel MEG data. The measured data and sensor noises are first transformed into independent virtual channels. Thereafter, the total information is obtained by summing up the power signal-to-noise ratios of the virtual channels without any assumptions of the source currents. We consider both theoretical aspects and examples of the total information obtained from MEG subject measurements. In addition, selection of the virtual channels can be optimized by maximizing the total information. The results show that the optimal number of virtual channels is about 80, which supports the suggestion that there are less than 100 degrees of freedom in state-of-the-art multichannel MEG data.