Brain Signal Analysis

Researchers: Toni Auranen, Iiro Jääskeläinen, Jouko Lampinen, Aapo Nummenmaa, Mikko Sams, Aki Vehtari

We have studied analyses of time-frequency representations (TFRs). TFR is a way of representing the power of a continuous time signal as a function of time and frequency simultaneously. It is calculated by convolving a (Morlet) wavelet with the signal and taking the squared norm of the result. We conducted an MEG study, in which the event-related desynchronization (ERD) and synchronization (ERS) of real and imagery finger movements were evaluated using TFRs. ERD/ERS describe the relative decrease and increase in the signal respectively. A nonparametric statistical method was developed for the statistical analysis of TFRs. The method is composed of a Quade test and a Bonferroni type of correction for testing multiple hypotheses. In Figure 40 some of the results are plotted on a topographic plot on an MEG helmet that is placed on the head of our artificial person. The figure is derived from TFRs and the results of the statistical tests done on them.

Figure 40

Figure 40: The figure shows the statistically significant (p<0.05) ERD/ERS TFR values of 20-Hz activity when the subjects lifted the right index finger (at 1000 ms). The geometry is that of Neuromag Vectorview helmet and bilinear interpolation was used to determine the interior colour of each polygon. There is a clear bilateral 40% ERD (blue colors) during the actual movement and a later emerging contralateral strong ERS (rebound, red colours) after the movement has ended.

We have also started collaboration with Massachusetts General Hospital - Harvard Medical School NMR Center in developing combined fMRI/MEG/EEG analysis methods to increase in spatiotemporal imaging resolution. We currently investigate the possibility of employing Bayesian Markov-Chain Monte-Carlo sampling methods in characterizing the modes of possible cortical current source configurations underlying brain signal recordings.