Identifying the neural networks subserving specific neural processes

In this paper we propose some approaches to the problem of identifying large scale physiological neural networks (LSNN). These approaches combine the well-known high temporal resolution of electric/magnetic measurements with what seems to be a rector principle of brain functioning: the synchronic activation of areas that participate in the processing of information. The time series within the brain obtained from the application of a distributed linear inverse solution (ELECTRA) to measured evoked responses are assumed to belong to a LSNN if they exhibit a significantly high correlation in the time, the frequency or the time-frequency domain. Measures of correlation among time series independent of their relative amplitudes are selected in order to deal with one of the basic limitations of linear inverse solutions, the underestimation of the source strength. This condition is essential to be able to detect in the network cortical and deeper brain areas. One of the proposed approaches is illustrated in the analysis of measured electric responses evoked by moving or colored visual stimuli